RNN-T Conformer training for LibriSpeech (#143)

* Begin to add RNN-T training for librispeech.

* Copy files from conformer_ctc.

Will edit it.

* Use conformer/transformer model as encoder.

* Begin to add training script.

* Add training code.

* Remove long utterances to avoid OOM when a large max_duraiton is used.

* Begin to add decoding script.

* Add decoding script.

* Minor fixes.

* Add beam search.

* Use LSTM layers for the encoder.

Need more tunings.

* Use stateless decoder.

* Minor fixes to make it ready for merge.

* Fix README.

* Update RESULT.md to include RNN-T Conformer.

* Minor fixes.

* Fix tests.

* Minor fixes.

* Minor fixes.

* Fix tests.

.github/workflows/test.yml

@@ -103,8 +103,10 @@ jobs:
           cd egs/librispeech/ASR/conformer_ctc
           pytest -v -s
 
-          cd ..
-          pytest -v -s ./transducer
+          if [[ ${{ matrix.torchaudio }} == "0.10.0" ]]; then
+            cd ../transducer
+            pytest -v -s
+          fi
 
       - name: Run tests
         if: startsWith(matrix.os, 'macos')
@@ -120,5 +122,7 @@ jobs:
           cd egs/librispeech/ASR/conformer_ctc
           pytest -v -s
 
-          cd ..
-          pytest -v -s ./transducer
+          if [[ ${{ matrix.torchaudio }} == "0.10.0" ]]; then
+            cd ../transducer
+            pytest -v -s
+          fi

.gitignore

@@ -8,3 +8,4 @@ exp*/
 download
 *.bak
 *-bak
+*bak.py

README.md

@@ -34,8 +34,11 @@ We do provide a Colab notebook for this recipe.
 
 ### LibriSpeech
 
-We provide two models for this recipe: [conformer CTC model][LibriSpeech_conformer_ctc]
-and [TDNN LSTM CTC model][LibriSpeech_tdnn_lstm_ctc].
+We provide 3 models for this recipe:
+
+- [conformer CTC model][LibriSpeech_conformer_ctc]
+- [TDNN LSTM CTC model][LibriSpeech_tdnn_lstm_ctc]
+- [RNN-T Conformer model][LibriSpeech_transducer]
 
 #### Conformer CTC Model
 
@@ -58,6 +61,20 @@ The WER for this model is:
 
 We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model:  [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1kNmDXNMwREi0rZGAOIAOJo93REBuOTcd?usp=sharing)
 
+
+#### RNN-T Conformer model
+
+Using Conformer as encoder.
+
+The best WER with greedy search is:
+
+|     | test-clean | test-other |
+|-----|------------|------------|
+| WER | 3.16       | 7.71       |
+
+We provide a Colab notebook to run a pre-trained RNN-T conformer model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1_u6yK9jDkPwG_NLrZMN2XK7Aeq4suMO2?usp=sharing)
+
+
 ### Aishell
 
 We provide two models for this recipe: [conformer CTC model][Aishell_conformer_ctc]
@@ -125,6 +142,7 @@ Please see: [![Open In Colab](https://colab.research.google.com/assets/colab-bad
 
 [LibriSpeech_tdnn_lstm_ctc]: egs/librispeech/ASR/tdnn_lstm_ctc
 [LibriSpeech_conformer_ctc]: egs/librispeech/ASR/conformer_ctc
+[LibriSpeech_transducer]: egs/librispeech/ASR/transducer
 [Aishell_tdnn_lstm_ctc]: egs/aishell/ASR/tdnn_lstm_ctc
 [Aishell_conformer_ctc]: egs/aishell/ASR/conformer_ctc
 [TIMIT_tdnn_lstm_ctc]: egs/timit/ASR/tdnn_lstm_ctc

egs/librispeech/ASR/RESULTS.md

@@ -1,5 +1,51 @@
 ## Results
 
+### LibriSpeech BPE training results (RNN-T)
+
+#### 2021-12-17
+
+RNN-T + Conformer encoder
+
+The best WER is
+
+|     | test-clean | test-other |
+|-----|------------|------------|
+| WER | 3.16       | 7.71       |
+
+using `--epoch 26 --avg 12` during decoding with greedy search.
+
+The training command to reproduce the above WER is:
+
+```
+export CUDA_VISIBLE_DEVICES="0,1,2,3"
+
+./transducer/train.py \
+  --world-size 4 \
+  --num-epochs 30 \
+  --start-epoch 0 \
+  --exp-dir transducer/exp-lr-2.5-full \
+  --full-libri 1 \
+  --max-duration 250 \
+  --lr-factor 2.5
+```
+
+The decoding command is:
+
+```
+epoch=26
+avg=12
+
+./transducer/decode.py \
+  --epoch $epoch \
+  --avg $avg \
+  --exp-dir transducer/exp-lr-2.5-full \
+  --bpe-model ./data/lang_bpe_500/bpe.model \
+  --max-duration 100
+```
+
+You can find the tensorboard log at: <https://tensorboard.dev/experiment/PYIbeD6zRJez1ViXaRqqeg/>
+
+
 ### LibriSpeech BPE training results (Conformer-CTC)
 
 #### 2021-11-09

egs/librispeech/ASR/conformer_ctc/decode.py

@@ -428,8 +428,6 @@ def decode_dataset(
       The first is the reference transcript, and the second is the
       predicted result.
     """
-    results = []
-
     num_cuts = 0
 
     try:

egs/librispeech/ASR/local/display_manifest_statistics.py

@@ -0,0 +1,215 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+This file displays duration statistics of utterances in a manifest.
+You can use the displayed value to choose minimum/maximum duration
+to remove short and long utterances during the training.
+
+See the function `remove_short_and_long_utt()` in transducer/train.py
+for usage.
+"""
+
+
+from lhotse import load_manifest
+
+
+def main():
+    path = "./data/fbank/cuts_train-clean-100.json.gz"
+    path = "./data/fbank/cuts_train-clean-360.json.gz"
+    path = "./data/fbank/cuts_train-other-500.json.gz"
+    path = "./data/fbank/cuts_dev-clean.json.gz"
+    path = "./data/fbank/cuts_dev-other.json.gz"
+    path = "./data/fbank/cuts_test-clean.json.gz"
+    path = "./data/fbank/cuts_test-other.json.gz"
+
+    cuts = load_manifest(path)
+    cuts.describe()
+
+
+if __name__ == "__main__":
+    main()
+
+"""
+## train-clean-100
+Cuts count: 85617
+Total duration (hours): 303.8
+Speech duration (hours): 303.8 (100.0%)
+***
+Duration statistics (seconds):
+mean    12.8
+std     3.8
+min     1.3
+0.1%    1.9
+0.5%    2.2
+1%      2.5
+5%      4.2
+10%     6.4
+25%     11.4
+50%     13.8
+75%     15.3
+90%     16.7
+95%     17.3
+99%     18.1
+99.5%   18.4
+99.9%   18.8
+max     27.2
+
+## train-clean-360
+Cuts count: 312042
+Total duration (hours): 1098.2
+Speech duration (hours): 1098.2 (100.0%)
+***
+Duration statistics (seconds):
+mean    12.7
+std     3.8
+min     1.0
+0.1%    1.8
+0.5%    2.2
+1%      2.5
+5%      4.2
+10%     6.2
+25%     11.2
+50%     13.7
+75%     15.3
+90%     16.6
+95%     17.3
+99%     18.1
+99.5%   18.4
+99.9%   18.8
+max     33.0
+
+## train-other 500
+Cuts count: 446064
+Total duration (hours): 1500.6
+Speech duration (hours): 1500.6 (100.0%)
+***
+Duration statistics (seconds):
+mean    12.1
+std     4.2
+min     0.8
+0.1%    1.7
+0.5%    2.1
+1%      2.3
+5%      3.5
+10%     5.0
+25%     9.8
+50%     13.4
+75%     15.1
+90%     16.5
+95%     17.2
+99%     18.1
+99.5%   18.4
+99.9%   18.9
+max     31.0
+
+## dev-clean
+Cuts count: 2703
+Total duration (hours): 5.4
+Speech duration (hours): 5.4 (100.0%)
+***
+Duration statistics (seconds):
+mean    7.2
+std     4.7
+min     1.4
+0.1%    1.6
+0.5%    1.8
+1%      1.9
+5%      2.4
+10%     2.7
+25%     3.8
+50%     5.9
+75%     9.3
+90%     13.3
+95%     16.4
+99%     23.8
+99.5%   28.5
+99.9%   32.3
+max     32.6
+
+## dev-other
+Cuts count: 2864
+Total duration (hours): 5.1
+Speech duration (hours): 5.1 (100.0%)
+***
+Duration statistics (seconds):
+mean    6.4
+std     4.3
+min     1.1
+0.1%    1.3
+0.5%    1.7
+1%      1.8
+5%      2.2
+10%     2.6
+25%     3.5
+50%     5.3
+75%     7.9
+90%     12.0
+95%     15.0
+99%     22.2
+99.5%   27.1
+99.9%   32.4
+max     35.2
+
+## test-clean
+Cuts count: 2620
+Total duration (hours): 5.4
+Speech duration (hours): 5.4 (100.0%)
+***
+Duration statistics (seconds):
+mean    7.4
+std     5.2
+min     1.3
+0.1%    1.6
+0.5%    1.8
+1%      2.0
+5%      2.3
+10%     2.7
+25%     3.7
+50%     5.8
+75%     9.6
+90%     14.6
+95%     17.8
+99%     25.5
+99.5%   28.4
+99.9%   32.8
+max     35.0
+
+## test-other
+Cuts count: 2939
+Total duration (hours): 5.3
+Speech duration (hours): 5.3 (100.0%)
+***
+Duration statistics (seconds):
+mean    6.5
+std     4.4
+min     1.2
+0.1%    1.5
+0.5%    1.8
+1%      1.9
+5%      2.3
+10%     2.6
+25%     3.4
+50%     5.2
+75%     8.2
+90%     12.6
+95%     15.8
+99%     21.4
+99.5%   23.8
+99.9%   33.5
+max     34.5
+"""

egs/librispeech/ASR/transducer/README.md

@@ -0,0 +1,19 @@
+## Introduction
+
+The encoder consists of Conformer layers in this folder. You can use the
+following command to start the training:
+
+```bash
+cd egs/librispeech/ASR
+
+export CUDA_VISIBLE_DEVICES="0,1,2,3"
+
+./transducer/train.py \
+  --world-size 4 \
+  --num-epochs 30 \
+  --start-epoch 0 \
+  --exp-dir transducer/exp \
+  --full-libri 1 \
+  --max-duration 250 \
+  --lr-factor 2.5
+```

egs/librispeech/ASR/transducer/asr_datamodule.py

@@ -0,0 +1 @@
+../tdnn_lstm_ctc/asr_datamodule.py

egs/librispeech/ASR/transducer/beam_search.py

@@ -0,0 +1,212 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from dataclasses import dataclass
+from typing import Dict, List, Optional, Tuple
+
+import torch
+from model import Transducer
+
+
+def greedy_search(model: Transducer, encoder_out: torch.Tensor) -> List[int]:
+    """
+    Args:
+      model:
+        An instance of `Transducer`.
+      encoder_out:
+        A tensor of shape (N, T, C) from the encoder. Support only N==1 for now.
+    Returns:
+      Return the decoded result.
+    """
+    assert encoder_out.ndim == 3
+
+    # support only batch_size == 1 for now
+    assert encoder_out.size(0) == 1, encoder_out.size(0)
+    blank_id = model.decoder.blank_id
+    device = model.device
+
+    sos = torch.tensor([blank_id], device=device).reshape(1, 1)
+    decoder_out, (h, c) = model.decoder(sos)
+    T = encoder_out.size(1)
+    t = 0
+    hyp = []
+    max_u = 1000  # terminate after this number of steps
+    u = 0
+
+    while t < T and u < max_u:
+        # fmt: off
+        current_encoder_out = encoder_out[:, t:t+1, :]
+        # fmt: on
+        logits = model.joiner(current_encoder_out, decoder_out)
+        # logits is (1, 1, 1, vocab_size)
+
+        log_prob = logits.log_softmax(dim=-1)
+        # log_prob is (1, 1, 1, vocab_size)
+        # TODO: Use logits.argmax()
+        y = log_prob.argmax()
+        if y != blank_id:
+            hyp.append(y.item())
+            y = y.reshape(1, 1)
+            decoder_out, (h, c) = model.decoder(y, (h, c))
+            u += 1
+        else:
+            t += 1
+
+    return hyp
+
+
+@dataclass
+class Hypothesis:
+    ys: List[int]  # the predicted sequences so far
+    log_prob: float  # The log prob of ys
+
+    # Optional decoder state. We assume it is LSTM for now,
+    # so the state is a tuple (h, c)
+    decoder_state: Optional[Tuple[torch.Tensor, torch.Tensor]] = None
+
+
+def beam_search(
+    model: Transducer,
+    encoder_out: torch.Tensor,
+    beam: int = 5,
+) -> List[int]:
+    """
+    It implements Algorithm 1 in https://arxiv.org/pdf/1211.3711.pdf
+
+    espnet/nets/beam_search_transducer.py#L247 is used as a reference.
+
+    Args:
+      model:
+        An instance of `Transducer`.
+      encoder_out:
+        A tensor of shape (N, T, C) from the encoder. Support only N==1 for now.
+      beam:
+        Beam size.
+    Returns:
+      Return the decoded result.
+    """
+    assert encoder_out.ndim == 3
+
+    # support only batch_size == 1 for now
+    assert encoder_out.size(0) == 1, encoder_out.size(0)
+    blank_id = model.decoder.blank_id
+    sos_id = model.decoder.sos_id
+    device = model.device
+
+    sos = torch.tensor([blank_id], device=device).reshape(1, 1)
+    decoder_out, (h, c) = model.decoder(sos)
+    T = encoder_out.size(1)
+    t = 0
+    B = [Hypothesis(ys=[blank_id], log_prob=0.0, decoder_state=None)]
+    max_u = 20000  # terminate after this number of steps
+    u = 0
+
+    cache: Dict[
+        str, Tuple[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]
+    ] = {}
+
+    while t < T and u < max_u:
+        # fmt: off
+        current_encoder_out = encoder_out[:, t:t+1, :]
+        # fmt: on
+        A = B
+        B = []
+        #  for hyp in A:
+        #      for h in A:
+        #          if h.ys == hyp.ys[:-1]:
+        #              # update the score of hyp
+        #              decoder_input = torch.tensor(
+        #                  [h.ys[-1]], device=device
+        #              ).reshape(1, 1)
+        #              decoder_out, _ = model.decoder(
+        #                  decoder_input, h.decoder_state
+        #              )
+        #              logits = model.joiner(current_encoder_out, decoder_out)
+        #              log_prob = logits.log_softmax(dim=-1)
+        #              log_prob = log_prob.squeeze()
+        #              hyp.log_prob += h.log_prob + log_prob[hyp.ys[-1]].item()
+
+        while u < max_u:
+            y_star = max(A, key=lambda hyp: hyp.log_prob)
+            A.remove(y_star)
+
+            # Note: y_star.ys is unhashable, i.e., cannot be used
+            # as a key into a dict
+            cached_key = "_".join(map(str, y_star.ys))
+
+            if cached_key not in cache:
+                decoder_input = torch.tensor(
+                    [y_star.ys[-1]], device=device
+                ).reshape(1, 1)
+
+                decoder_out, decoder_state = model.decoder(
+                    decoder_input,
+                    y_star.decoder_state,
+                )
+                cache[cached_key] = (decoder_out, decoder_state)
+            else:
+                decoder_out, decoder_state = cache[cached_key]
+
+            logits = model.joiner(current_encoder_out, decoder_out)
+            log_prob = logits.log_softmax(dim=-1)
+            # log_prob is (1, 1, 1, vocab_size)
+            log_prob = log_prob.squeeze()
+            # Now log_prob is (vocab_size,)
+
+            # If we choose blank here, add the new hypothesis to B.
+            # Otherwise, add the new hypothesis to A
+
+            # First, choose blank
+            skip_log_prob = log_prob[blank_id]
+            new_y_star_log_prob = y_star.log_prob + skip_log_prob.item()
+
+            # ys[:] returns a copy of ys
+            new_y_star = Hypothesis(
+                ys=y_star.ys[:],
+                log_prob=new_y_star_log_prob,
+                # Caution: Use y_star.decoder_state here
+                decoder_state=y_star.decoder_state,
+            )
+            B.append(new_y_star)
+
+            # Second, choose other labels
+            for i, v in enumerate(log_prob.tolist()):
+                if i in (blank_id, sos_id):
+                    continue
+                new_ys = y_star.ys + [i]
+                new_log_prob = y_star.log_prob + v
+                new_hyp = Hypothesis(
+                    ys=new_ys,
+                    log_prob=new_log_prob,
+                    decoder_state=decoder_state,
+                )
+                A.append(new_hyp)
+            u += 1
+            # check whether B contains more than "beam" elements more probable
+            # than the most probable in A
+            A_most_probable = max(A, key=lambda hyp: hyp.log_prob)
+            B = sorted(
+                [hyp for hyp in B if hyp.log_prob > A_most_probable.log_prob],
+                key=lambda hyp: hyp.log_prob,
+                reverse=True,
+            )
+            if len(B) >= beam:
+                B = B[:beam]
+                break
+        t += 1
+    best_hyp = max(B, key=lambda hyp: hyp.log_prob / len(hyp.ys[1:]))
+    ys = best_hyp.ys[1:]  # [1:] to remove the blank
+    return ys

egs/librispeech/ASR/transducer/conformer.py

@@ -0,0 +1,922 @@
+#!/usr/bin/env python3
+# Copyright (c)  2021  University of Chinese Academy of Sciences (author: Han Zhu)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import math
+import warnings
+from typing import Optional, Tuple
+
+import torch
+from torch import Tensor, nn
+from transformer import Transformer
+
+from icefall.utils import make_pad_mask
+
+
+class Conformer(Transformer):
+    """
+    Args:
+        num_features (int): Number of input features
+        output_dim (int): Number of output dimension
+        subsampling_factor (int): subsampling factor of encoder (the convolution layers before transformers)
+        d_model (int): attention dimension
+        nhead (int): number of head
+        dim_feedforward (int): feedforward dimention
+        num_encoder_layers (int): number of encoder layers
+        dropout (float): dropout rate
+        cnn_module_kernel (int): Kernel size of convolution module
+        normalize_before (bool): whether to use layer_norm before the first block.
+        vgg_frontend (bool): whether to use vgg frontend.
+    """
+
+    def __init__(
+        self,
+        num_features: int,
+        output_dim: int,
+        subsampling_factor: int = 4,
+        d_model: int = 256,
+        nhead: int = 4,
+        dim_feedforward: int = 2048,
+        num_encoder_layers: int = 12,
+        dropout: float = 0.1,
+        cnn_module_kernel: int = 31,
+        normalize_before: bool = True,
+        vgg_frontend: bool = False,
+        use_feat_batchnorm: bool = False,
+    ) -> None:
+        super(Conformer, self).__init__(
+            num_features=num_features,
+            output_dim=output_dim,
+            subsampling_factor=subsampling_factor,
+            d_model=d_model,
+            nhead=nhead,
+            dim_feedforward=dim_feedforward,
+            num_encoder_layers=num_encoder_layers,
+            dropout=dropout,
+            normalize_before=normalize_before,
+            vgg_frontend=vgg_frontend,
+            use_feat_batchnorm=use_feat_batchnorm,
+        )
+
+        self.encoder_pos = RelPositionalEncoding(d_model, dropout)
+
+        encoder_layer = ConformerEncoderLayer(
+            d_model,
+            nhead,
+            dim_feedforward,
+            dropout,
+            cnn_module_kernel,
+            normalize_before,
+        )
+        self.encoder = ConformerEncoder(encoder_layer, num_encoder_layers)
+        self.normalize_before = normalize_before
+        if self.normalize_before:
+            self.after_norm = nn.LayerNorm(d_model)
+        else:
+            # Note: TorchScript detects that self.after_norm could be used inside forward()
+            #       and throws an error without this change.
+            self.after_norm = identity
+
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            The input tensor. Its shape is (batch_size, seq_len, feature_dim).
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames in
+            `x` before padding.
+        Returns:
+          Return a tuple containing 2 tensors:
+            - logits, its shape is (batch_size, output_seq_len, output_dim)
+            - logit_lens, a tensor of shape (batch_size,) containing the number
+              of frames in `logits` before padding.
+        """
+        if self.use_feat_batchnorm:
+            x = x.permute(0, 2, 1)  # (N, T, C) -> (N, C, T)
+            x = self.feat_batchnorm(x)
+            x = x.permute(0, 2, 1)  # (N, C, T) -> (N, T, C)
+
+        x = self.encoder_embed(x)
+        x, pos_emb = self.encoder_pos(x)
+        x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+
+        # Caution: We assume the subsampling factor is 4!
+        lengths = ((x_lens - 1) // 2 - 1) // 2
+        assert x.size(0) == lengths.max().item()
+        mask = make_pad_mask(lengths)
+
+        x = self.encoder(x, pos_emb, src_key_padding_mask=mask)  # (T, N, C)
+
+        if self.normalize_before:
+            x = self.after_norm(x)
+
+        logits = self.encoder_output_layer(x)
+        logits = logits.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+
+        return logits, lengths
+
+
+class ConformerEncoderLayer(nn.Module):
+    """
+    ConformerEncoderLayer is made up of self-attn, feedforward and convolution networks.
+    See: "Conformer: Convolution-augmented Transformer for Speech Recognition"
+
+    Args:
+        d_model: the number of expected features in the input (required).
+        nhead: the number of heads in the multiheadattention models (required).
+        dim_feedforward: the dimension of the feedforward network model (default=2048).
+        dropout: the dropout value (default=0.1).
+        cnn_module_kernel (int): Kernel size of convolution module.
+        normalize_before: whether to use layer_norm before the first block.
+
+    Examples::
+        >>> encoder_layer = ConformerEncoderLayer(d_model=512, nhead=8)
+        >>> src = torch.rand(10, 32, 512)
+        >>> pos_emb = torch.rand(32, 19, 512)
+        >>> out = encoder_layer(src, pos_emb)
+    """
+
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        cnn_module_kernel: int = 31,
+        normalize_before: bool = True,
+    ) -> None:
+        super(ConformerEncoderLayer, self).__init__()
+        self.self_attn = RelPositionMultiheadAttention(
+            d_model, nhead, dropout=0.0
+        )
+
+        self.feed_forward = nn.Sequential(
+            nn.Linear(d_model, dim_feedforward),
+            Swish(),
+            nn.Dropout(dropout),
+            nn.Linear(dim_feedforward, d_model),
+        )
+
+        self.feed_forward_macaron = nn.Sequential(
+            nn.Linear(d_model, dim_feedforward),
+            Swish(),
+            nn.Dropout(dropout),
+            nn.Linear(dim_feedforward, d_model),
+        )
+
+        self.conv_module = ConvolutionModule(d_model, cnn_module_kernel)
+
+        self.norm_ff_macaron = nn.LayerNorm(
+            d_model
+        )  # for the macaron style FNN module
+        self.norm_ff = nn.LayerNorm(d_model)  # for the FNN module
+        self.norm_mha = nn.LayerNorm(d_model)  # for the MHA module
+
+        self.ff_scale = 0.5
+
+        self.norm_conv = nn.LayerNorm(d_model)  # for the CNN module
+        self.norm_final = nn.LayerNorm(
+            d_model
+        )  # for the final output of the block
+
+        self.dropout = nn.Dropout(dropout)
+
+        self.normalize_before = normalize_before
+
+    def forward(
+        self,
+        src: Tensor,
+        pos_emb: Tensor,
+        src_mask: Optional[Tensor] = None,
+        src_key_padding_mask: Optional[Tensor] = None,
+    ) -> Tensor:
+        """
+        Pass the input through the encoder layer.
+
+        Args:
+            src: the sequence to the encoder layer (required).
+            pos_emb: Positional embedding tensor (required).
+            src_mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+
+        Shape:
+            src: (S, N, E).
+            pos_emb: (N, 2*S-1, E)
+            src_mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, N is the batch size, E is the feature number
+        """
+
+        # macaron style feed forward module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_ff_macaron(src)
+        src = residual + self.ff_scale * self.dropout(
+            self.feed_forward_macaron(src)
+        )
+        if not self.normalize_before:
+            src = self.norm_ff_macaron(src)
+
+        # multi-headed self-attention module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_mha(src)
+        src_att = self.self_attn(
+            src,
+            src,
+            src,
+            pos_emb=pos_emb,
+            attn_mask=src_mask,
+            key_padding_mask=src_key_padding_mask,
+        )[0]
+        src = residual + self.dropout(src_att)
+        if not self.normalize_before:
+            src = self.norm_mha(src)
+
+        # convolution module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_conv(src)
+        src = residual + self.dropout(self.conv_module(src))
+        if not self.normalize_before:
+            src = self.norm_conv(src)
+
+        # feed forward module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_ff(src)
+        src = residual + self.ff_scale * self.dropout(self.feed_forward(src))
+        if not self.normalize_before:
+            src = self.norm_ff(src)
+
+        if self.normalize_before:
+            src = self.norm_final(src)
+
+        return src
+
+
+class ConformerEncoder(nn.TransformerEncoder):
+    r"""ConformerEncoder is a stack of N encoder layers
+
+    Args:
+        encoder_layer: an instance of the ConformerEncoderLayer() class (required).
+        num_layers: the number of sub-encoder-layers in the encoder (required).
+        norm: the layer normalization component (optional).
+
+    Examples::
+        >>> encoder_layer = ConformerEncoderLayer(d_model=512, nhead=8)
+        >>> conformer_encoder = ConformerEncoder(encoder_layer, num_layers=6)
+        >>> src = torch.rand(10, 32, 512)
+        >>> pos_emb = torch.rand(32, 19, 512)
+        >>> out = conformer_encoder(src, pos_emb)
+    """
+
+    def __init__(
+        self, encoder_layer: nn.Module, num_layers: int, norm: nn.Module = None
+    ) -> None:
+        super(ConformerEncoder, self).__init__(
+            encoder_layer=encoder_layer, num_layers=num_layers, norm=norm
+        )
+
+    def forward(
+        self,
+        src: Tensor,
+        pos_emb: Tensor,
+        mask: Optional[Tensor] = None,
+        src_key_padding_mask: Optional[Tensor] = None,
+    ) -> Tensor:
+        r"""Pass the input through the encoder layers in turn.
+
+        Args:
+            src: the sequence to the encoder (required).
+            pos_emb: Positional embedding tensor (required).
+            mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+
+        Shape:
+            src: (S, N, E).
+            pos_emb: (N, 2*S-1, E)
+            mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, T is the target sequence length, N is the batch size, E is the feature number
+
+        """
+        output = src
+
+        for mod in self.layers:
+            output = mod(
+                output,
+                pos_emb,
+                src_mask=mask,
+                src_key_padding_mask=src_key_padding_mask,
+            )
+
+        if self.norm is not None:
+            output = self.norm(output)
+
+        return output
+
+
+class RelPositionalEncoding(torch.nn.Module):
+    """Relative positional encoding module.
+
+    See : Appendix B in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+    Modified from https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/embedding.py
+
+    Args:
+        d_model: Embedding dimension.
+        dropout_rate: Dropout rate.
+        max_len: Maximum input length.
+
+    """
+
+    def __init__(
+        self, d_model: int, dropout_rate: float, max_len: int = 5000
+    ) -> None:
+        """Construct an PositionalEncoding object."""
+        super(RelPositionalEncoding, self).__init__()
+        self.d_model = d_model
+        self.xscale = math.sqrt(self.d_model)
+        self.dropout = torch.nn.Dropout(p=dropout_rate)
+        self.pe = None
+        self.extend_pe(torch.tensor(0.0).expand(1, max_len))
+
+    def extend_pe(self, x: Tensor) -> None:
+        """Reset the positional encodings."""
+        if self.pe is not None:
+            # self.pe contains both positive and negative parts
+            # the length of self.pe is 2 * input_len - 1
+            if self.pe.size(1) >= x.size(1) * 2 - 1:
+                # Note: TorchScript doesn't implement operator== for torch.Device
+                if self.pe.dtype != x.dtype or str(self.pe.device) != str(
+                    x.device
+                ):
+                    self.pe = self.pe.to(dtype=x.dtype, device=x.device)
+                return
+        # Suppose `i` means to the position of query vecotr and `j` means the
+        # position of key vector. We use position relative positions when keys
+        # are to the left (i>j) and negative relative positions otherwise (i<j).
+        pe_positive = torch.zeros(x.size(1), self.d_model)
+        pe_negative = torch.zeros(x.size(1), self.d_model)
+        position = torch.arange(0, x.size(1), dtype=torch.float32).unsqueeze(1)
+        div_term = torch.exp(
+            torch.arange(0, self.d_model, 2, dtype=torch.float32)
+            * -(math.log(10000.0) / self.d_model)
+        )
+        pe_positive[:, 0::2] = torch.sin(position * div_term)
+        pe_positive[:, 1::2] = torch.cos(position * div_term)
+        pe_negative[:, 0::2] = torch.sin(-1 * position * div_term)
+        pe_negative[:, 1::2] = torch.cos(-1 * position * div_term)
+
+        # Reserve the order of positive indices and concat both positive and
+        # negative indices. This is used to support the shifting trick
+        # as in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+        pe_positive = torch.flip(pe_positive, [0]).unsqueeze(0)
+        pe_negative = pe_negative[1:].unsqueeze(0)
+        pe = torch.cat([pe_positive, pe_negative], dim=1)
+        self.pe = pe.to(device=x.device, dtype=x.dtype)
+
+    def forward(self, x: torch.Tensor) -> Tuple[Tensor, Tensor]:
+        """Add positional encoding.
+
+        Args:
+            x (torch.Tensor): Input tensor (batch, time, `*`).
+
+        Returns:
+            torch.Tensor: Encoded tensor (batch, time, `*`).
+            torch.Tensor: Encoded tensor (batch, 2*time-1, `*`).
+
+        """
+        self.extend_pe(x)
+        x = x * self.xscale
+        pos_emb = self.pe[
+            :,
+            self.pe.size(1) // 2
+            - x.size(1)
+            + 1 : self.pe.size(1) // 2  # noqa E203
+            + x.size(1),
+        ]
+        return self.dropout(x), self.dropout(pos_emb)
+
+
+class RelPositionMultiheadAttention(nn.Module):
+    r"""Multi-Head Attention layer with relative position encoding
+
+    See reference: "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+
+    Args:
+        embed_dim: total dimension of the model.
+        num_heads: parallel attention heads.
+        dropout: a Dropout layer on attn_output_weights. Default: 0.0.
+
+    Examples::
+
+        >>> rel_pos_multihead_attn = RelPositionMultiheadAttention(embed_dim, num_heads)
+        >>> attn_output, attn_output_weights = multihead_attn(query, key, value, pos_emb)
+    """
+
+    def __init__(
+        self,
+        embed_dim: int,
+        num_heads: int,
+        dropout: float = 0.0,
+    ) -> None:
+        super(RelPositionMultiheadAttention, self).__init__()
+        self.embed_dim = embed_dim
+        self.num_heads = num_heads
+        self.dropout = dropout
+        self.head_dim = embed_dim // num_heads
+        assert (
+            self.head_dim * num_heads == self.embed_dim
+        ), "embed_dim must be divisible by num_heads"
+
+        self.in_proj = nn.Linear(embed_dim, 3 * embed_dim, bias=True)
+        self.out_proj = nn.Linear(embed_dim, embed_dim, bias=True)
+
+        # linear transformation for positional encoding.
+        self.linear_pos = nn.Linear(embed_dim, embed_dim, bias=False)
+        # these two learnable bias are used in matrix c and matrix d
+        # as described in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context" Section 3.3
+        self.pos_bias_u = nn.Parameter(torch.Tensor(num_heads, self.head_dim))
+        self.pos_bias_v = nn.Parameter(torch.Tensor(num_heads, self.head_dim))
+
+        self._reset_parameters()
+
+    def _reset_parameters(self) -> None:
+        nn.init.xavier_uniform_(self.in_proj.weight)
+        nn.init.constant_(self.in_proj.bias, 0.0)
+        nn.init.constant_(self.out_proj.bias, 0.0)
+
+        nn.init.xavier_uniform_(self.pos_bias_u)
+        nn.init.xavier_uniform_(self.pos_bias_v)
+
+    def forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        pos_emb: Tensor,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        r"""
+        Args:
+            query, key, value: map a query and a set of key-value pairs to an output.
+            pos_emb: Positional embedding tensor
+            key_padding_mask: if provided, specified padding elements in the key will
+                be ignored by the attention. When given a binary mask and a value is True,
+                the corresponding value on the attention layer will be ignored. When given
+                a byte mask and a value is non-zero, the corresponding value on the attention
+                layer will be ignored
+            need_weights: output attn_output_weights.
+            attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all
+                the batches while a 3D mask allows to specify a different mask for the entries of each batch.
+
+        Shape:
+            - Inputs:
+            - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - pos_emb: :math:`(N, 2*L-1, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.
+            If a ByteTensor is provided, the non-zero positions will be ignored while the position
+            with the zero positions will be unchanged. If a BoolTensor is provided, the positions with the
+            value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.
+            - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.
+            3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,
+            S is the source sequence length. attn_mask ensure that position i is allowed to attend the unmasked
+            positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend
+            while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``
+            is not allowed to attend while ``False`` values will be unchanged. If a FloatTensor
+            is provided, it will be added to the attention weight.
+
+            - Outputs:
+            - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,
+            E is the embedding dimension.
+            - attn_output_weights: :math:`(N, L, S)` where N is the batch size,
+            L is the target sequence length, S is the source sequence length.
+        """
+        return self.multi_head_attention_forward(
+            query,
+            key,
+            value,
+            pos_emb,
+            self.embed_dim,
+            self.num_heads,
+            self.in_proj.weight,
+            self.in_proj.bias,
+            self.dropout,
+            self.out_proj.weight,
+            self.out_proj.bias,
+            training=self.training,
+            key_padding_mask=key_padding_mask,
+            need_weights=need_weights,
+            attn_mask=attn_mask,
+        )
+
+    def rel_shift(self, x: Tensor) -> Tensor:
+        """Compute relative positional encoding.
+
+        Args:
+            x: Input tensor (batch, head, time1, 2*time1-1).
+                time1 means the length of query vector.
+
+        Returns:
+            Tensor: tensor of shape (batch, head, time1, time2)
+          (note: time2 has the same value as time1, but it is for
+          the key, while time1 is for the query).
+        """
+        (batch_size, num_heads, time1, n) = x.shape
+        assert n == 2 * time1 - 1
+        # Note: TorchScript requires explicit arg for stride()
+        batch_stride = x.stride(0)
+        head_stride = x.stride(1)
+        time1_stride = x.stride(2)
+        n_stride = x.stride(3)
+        return x.as_strided(
+            (batch_size, num_heads, time1, time1),
+            (batch_stride, head_stride, time1_stride - n_stride, n_stride),
+            storage_offset=n_stride * (time1 - 1),
+        )
+
+    def multi_head_attention_forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        pos_emb: Tensor,
+        embed_dim_to_check: int,
+        num_heads: int,
+        in_proj_weight: Tensor,
+        in_proj_bias: Tensor,
+        dropout_p: float,
+        out_proj_weight: Tensor,
+        out_proj_bias: Tensor,
+        training: bool = True,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        r"""
+        Args:
+            query, key, value: map a query and a set of key-value pairs to an output.
+            pos_emb: Positional embedding tensor
+            embed_dim_to_check: total dimension of the model.
+            num_heads: parallel attention heads.
+            in_proj_weight, in_proj_bias: input projection weight and bias.
+            dropout_p: probability of an element to be zeroed.
+            out_proj_weight, out_proj_bias: the output projection weight and bias.
+            training: apply dropout if is ``True``.
+            key_padding_mask: if provided, specified padding elements in the key will
+                be ignored by the attention. This is an binary mask. When the value is True,
+                the corresponding value on the attention layer will be filled with -inf.
+            need_weights: output attn_output_weights.
+            attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all
+                the batches while a 3D mask allows to specify a different mask for the entries of each batch.
+
+        Shape:
+            Inputs:
+            - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - pos_emb: :math:`(N, 2*L-1, E)` or :math:`(1, 2*L-1, E)` where L is the target sequence
+            length, N is the batch size, E is the embedding dimension.
+            - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.
+            If a ByteTensor is provided, the non-zero positions will be ignored while the zero positions
+            will be unchanged. If a BoolTensor is provided, the positions with the
+            value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.
+            - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.
+            3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,
+            S is the source sequence length. attn_mask ensures that position i is allowed to attend the unmasked
+            positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend
+            while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``
+            are not allowed to attend while ``False`` values will be unchanged. If a FloatTensor
+            is provided, it will be added to the attention weight.
+
+            Outputs:
+            - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,
+            E is the embedding dimension.
+            - attn_output_weights: :math:`(N, L, S)` where N is the batch size,
+            L is the target sequence length, S is the source sequence length.
+        """
+
+        tgt_len, bsz, embed_dim = query.size()
+        assert embed_dim == embed_dim_to_check
+        assert key.size(0) == value.size(0) and key.size(1) == value.size(1)
+
+        head_dim = embed_dim // num_heads
+        assert (
+            head_dim * num_heads == embed_dim
+        ), "embed_dim must be divisible by num_heads"
+        scaling = float(head_dim) ** -0.5
+
+        if torch.equal(query, key) and torch.equal(key, value):
+            # self-attention
+            q, k, v = nn.functional.linear(
+                query, in_proj_weight, in_proj_bias
+            ).chunk(3, dim=-1)
+
+        elif torch.equal(key, value):
+            # encoder-decoder attention
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = 0
+            _end = embed_dim
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            q = nn.functional.linear(query, _w, _b)
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim
+            _end = None
+            _w = in_proj_weight[_start:, :]
+            if _b is not None:
+                _b = _b[_start:]
+            k, v = nn.functional.linear(key, _w, _b).chunk(2, dim=-1)
+
+        else:
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = 0
+            _end = embed_dim
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            q = nn.functional.linear(query, _w, _b)
+
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim
+            _end = embed_dim * 2
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            k = nn.functional.linear(key, _w, _b)
+
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim * 2
+            _end = None
+            _w = in_proj_weight[_start:, :]
+            if _b is not None:
+                _b = _b[_start:]
+            v = nn.functional.linear(value, _w, _b)
+
+        if attn_mask is not None:
+            assert (
+                attn_mask.dtype == torch.float32
+                or attn_mask.dtype == torch.float64
+                or attn_mask.dtype == torch.float16
+                or attn_mask.dtype == torch.uint8
+                or attn_mask.dtype == torch.bool
+            ), "Only float, byte, and bool types are supported for attn_mask, not {}".format(
+                attn_mask.dtype
+            )
+            if attn_mask.dtype == torch.uint8:
+                warnings.warn(
+                    "Byte tensor for attn_mask is deprecated. Use bool tensor instead."
+                )
+                attn_mask = attn_mask.to(torch.bool)
+
+            if attn_mask.dim() == 2:
+                attn_mask = attn_mask.unsqueeze(0)
+                if list(attn_mask.size()) != [1, query.size(0), key.size(0)]:
+                    raise RuntimeError(
+                        "The size of the 2D attn_mask is not correct."
+                    )
+            elif attn_mask.dim() == 3:
+                if list(attn_mask.size()) != [
+                    bsz * num_heads,
+                    query.size(0),
+                    key.size(0),
+                ]:
+                    raise RuntimeError(
+                        "The size of the 3D attn_mask is not correct."
+                    )
+            else:
+                raise RuntimeError(
+                    "attn_mask's dimension {} is not supported".format(
+                        attn_mask.dim()
+                    )
+                )
+            # attn_mask's dim is 3 now.
+
+        # convert ByteTensor key_padding_mask to bool
+        if (
+            key_padding_mask is not None
+            and key_padding_mask.dtype == torch.uint8
+        ):
+            warnings.warn(
+                "Byte tensor for key_padding_mask is deprecated. Use bool tensor instead."
+            )
+            key_padding_mask = key_padding_mask.to(torch.bool)
+
+        q = q.contiguous().view(tgt_len, bsz, num_heads, head_dim)
+        k = k.contiguous().view(-1, bsz, num_heads, head_dim)
+        v = v.contiguous().view(-1, bsz * num_heads, head_dim).transpose(0, 1)
+
+        src_len = k.size(0)
+
+        if key_padding_mask is not None:
+            assert key_padding_mask.size(0) == bsz, "{} == {}".format(
+                key_padding_mask.size(0), bsz
+            )
+            assert key_padding_mask.size(1) == src_len, "{} == {}".format(
+                key_padding_mask.size(1), src_len
+            )
+
+        q = q.transpose(0, 1)  # (batch, time1, head, d_k)
+
+        pos_emb_bsz = pos_emb.size(0)
+        assert pos_emb_bsz in (1, bsz)  # actually it is 1
+        p = self.linear_pos(pos_emb).view(pos_emb_bsz, -1, num_heads, head_dim)
+        p = p.transpose(1, 2)  # (batch, head, 2*time1-1, d_k)
+
+        q_with_bias_u = (q + self.pos_bias_u).transpose(
+            1, 2
+        )  # (batch, head, time1, d_k)
+
+        q_with_bias_v = (q + self.pos_bias_v).transpose(
+            1, 2
+        )  # (batch, head, time1, d_k)
+
+        # compute attention score
+        # first compute matrix a and matrix c
+        # as described in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context" Section 3.3
+        k = k.permute(1, 2, 3, 0)  # (batch, head, d_k, time2)
+        matrix_ac = torch.matmul(
+            q_with_bias_u, k
+        )  # (batch, head, time1, time2)
+
+        # compute matrix b and matrix d
+        matrix_bd = torch.matmul(
+            q_with_bias_v, p.transpose(-2, -1)
+        )  # (batch, head, time1, 2*time1-1)
+        matrix_bd = self.rel_shift(matrix_bd)
+
+        attn_output_weights = (
+            matrix_ac + matrix_bd
+        ) * scaling  # (batch, head, time1, time2)
+
+        attn_output_weights = attn_output_weights.view(
+            bsz * num_heads, tgt_len, -1
+        )
+
+        assert list(attn_output_weights.size()) == [
+            bsz * num_heads,
+            tgt_len,
+            src_len,
+        ]
+
+        if attn_mask is not None:
+            if attn_mask.dtype == torch.bool:
+                attn_output_weights.masked_fill_(attn_mask, float("-inf"))
+            else:
+                attn_output_weights += attn_mask
+
+        if key_padding_mask is not None:
+            attn_output_weights = attn_output_weights.view(
+                bsz, num_heads, tgt_len, src_len
+            )
+            attn_output_weights = attn_output_weights.masked_fill(
+                key_padding_mask.unsqueeze(1).unsqueeze(2),
+                float("-inf"),
+            )
+            attn_output_weights = attn_output_weights.view(
+                bsz * num_heads, tgt_len, src_len
+            )
+
+        attn_output_weights = nn.functional.softmax(attn_output_weights, dim=-1)
+        attn_output_weights = nn.functional.dropout(
+            attn_output_weights, p=dropout_p, training=training
+        )
+
+        attn_output = torch.bmm(attn_output_weights, v)
+        assert list(attn_output.size()) == [bsz * num_heads, tgt_len, head_dim]
+        attn_output = (
+            attn_output.transpose(0, 1)
+            .contiguous()
+            .view(tgt_len, bsz, embed_dim)
+        )
+        attn_output = nn.functional.linear(
+            attn_output, out_proj_weight, out_proj_bias
+        )
+
+        if need_weights:
+            # average attention weights over heads
+            attn_output_weights = attn_output_weights.view(
+                bsz, num_heads, tgt_len, src_len
+            )
+            return attn_output, attn_output_weights.sum(dim=1) / num_heads
+        else:
+            return attn_output, None
+
+
+class ConvolutionModule(nn.Module):
+    """ConvolutionModule in Conformer model.
+    Modified from https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/conformer/convolution.py
+
+    Args:
+        channels (int): The number of channels of conv layers.
+        kernel_size (int): Kernerl size of conv layers.
+        bias (bool): Whether to use bias in conv layers (default=True).
+
+    """
+
+    def __init__(
+        self, channels: int, kernel_size: int, bias: bool = True
+    ) -> None:
+        """Construct an ConvolutionModule object."""
+        super(ConvolutionModule, self).__init__()
+        # kernerl_size should be a odd number for 'SAME' padding
+        assert (kernel_size - 1) % 2 == 0
+
+        self.pointwise_conv1 = nn.Conv1d(
+            channels,
+            2 * channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            bias=bias,
+        )
+        self.depthwise_conv = nn.Conv1d(
+            channels,
+            channels,
+            kernel_size,
+            stride=1,
+            padding=(kernel_size - 1) // 2,
+            groups=channels,
+            bias=bias,
+        )
+        self.norm = nn.BatchNorm1d(channels)
+        self.pointwise_conv2 = nn.Conv1d(
+            channels,
+            channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            bias=bias,
+        )
+        self.activation = Swish()
+
+    def forward(self, x: Tensor) -> Tensor:
+        """Compute convolution module.
+
+        Args:
+            x: Input tensor (#time, batch, channels).
+
+        Returns:
+            Tensor: Output tensor (#time, batch, channels).
+
+        """
+        # exchange the temporal dimension and the feature dimension
+        x = x.permute(1, 2, 0)  # (#batch, channels, time).
+
+        # GLU mechanism
+        x = self.pointwise_conv1(x)  # (batch, 2*channels, time)
+        x = nn.functional.glu(x, dim=1)  # (batch, channels, time)
+
+        # 1D Depthwise Conv
+        x = self.depthwise_conv(x)
+        x = self.activation(self.norm(x))
+
+        x = self.pointwise_conv2(x)  # (batch, channel, time)
+
+        return x.permute(2, 0, 1)
+
+
+class Swish(torch.nn.Module):
+    """Construct an Swish object."""
+
+    def forward(self, x: Tensor) -> Tensor:
+        """Return Swich activation function."""
+        return x * torch.sigmoid(x)
+
+
+def identity(x):
+    return x

egs/librispeech/ASR/transducer/decode.py

@@ -0,0 +1,463 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Usage:
+(1) greedy search
+./transducer/decode.py \
+        --epoch 14 \
+        --avg 7 \
+        --exp-dir ./transducer/exp \
+        --max-duration 100 \
+        --decoding-method greedy_search
+
+(2) beam search
+./transducer/decode.py \
+        --epoch 14 \
+        --avg 7 \
+        --exp-dir ./transducer/exp \
+        --max-duration 100 \
+        --decoding-method beam_search \
+        --beam-size 8
+"""
+
+
+import argparse
+import logging
+from collections import defaultdict
+from pathlib import Path
+from typing import Dict, List, Tuple
+
+import sentencepiece as spm
+import torch
+import torch.nn as nn
+from asr_datamodule import LibriSpeechAsrDataModule
+from beam_search import beam_search, greedy_search
+from conformer import Conformer
+from decoder import Decoder
+from joiner import Joiner
+from model import Transducer
+
+from icefall.checkpoint import average_checkpoints, load_checkpoint
+from icefall.env import get_env_info
+from icefall.utils import (
+    AttributeDict,
+    setup_logger,
+    store_transcripts,
+    write_error_stats,
+)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=26,
+        help="It specifies the checkpoint to use for decoding."
+        "Note: Epoch counts from 0.",
+    )
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=12,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="transducer/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="greedy_search",
+        help="""Possible values are:
+          - greedy_search
+          - beam_search
+        """,
+    )
+
+    parser.add_argument(
+        "--beam-size",
+        type=int,
+        default=5,
+        help="Used only when --decoding-method is beam_search",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            # parameters for conformer
+            "feature_dim": 80,
+            "encoder_out_dim": 512,
+            "subsampling_factor": 4,
+            "attention_dim": 512,
+            "nhead": 8,
+            "dim_feedforward": 2048,
+            "num_encoder_layers": 12,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            # decoder params
+            "decoder_embedding_dim": 1024,
+            "num_decoder_layers": 4,
+            "decoder_hidden_dim": 512,
+            "env_info": get_env_info(),
+        }
+    )
+    return params
+
+
+def get_encoder_model(params: AttributeDict):
+    # TODO: We can add an option to switch between Conformer and Transformer
+    encoder = Conformer(
+        num_features=params.feature_dim,
+        output_dim=params.encoder_out_dim,
+        subsampling_factor=params.subsampling_factor,
+        d_model=params.attention_dim,
+        nhead=params.nhead,
+        dim_feedforward=params.dim_feedforward,
+        num_encoder_layers=params.num_encoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+    return encoder
+
+
+def get_decoder_model(params: AttributeDict):
+    decoder = Decoder(
+        vocab_size=params.vocab_size,
+        embedding_dim=params.decoder_embedding_dim,
+        blank_id=params.blank_id,
+        sos_id=params.sos_id,
+        num_layers=params.num_decoder_layers,
+        hidden_dim=params.decoder_hidden_dim,
+        output_dim=params.encoder_out_dim,
+    )
+    return decoder
+
+
+def get_joiner_model(params: AttributeDict):
+    joiner = Joiner(
+        input_dim=params.encoder_out_dim,
+        output_dim=params.vocab_size,
+    )
+    return joiner
+
+
+def get_transducer_model(params: AttributeDict):
+    encoder = get_encoder_model(params)
+    decoder = get_decoder_model(params)
+    joiner = get_joiner_model(params)
+
+    model = Transducer(
+        encoder=encoder,
+        decoder=decoder,
+        joiner=joiner,
+    )
+    return model
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    batch: dict,
+) -> Dict[str, List[List[str]]]:
+    """Decode one batch and return the result in a dict. The dict has the
+    following format:
+
+        - key: It indicates the setting used for decoding. For example,
+               if greedy_search is used, it would be "greedy_search"
+               If beam search with a beam size of 7 is used, it would be
+               "beam_7"
+        - value: It contains the decoding result. `len(value)` equals to
+                 batch size. `value[i]` is the decoding result for the i-th
+                 utterance in the given batch.
+    Args:
+      params:
+        It's the return value of :func:`get_params`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+      batch:
+        It is the return value from iterating
+        `lhotse.dataset.K2SpeechRecognitionDataset`. See its documentation
+        for the format of the `batch`.
+    Returns:
+      Return the decoding result. See above description for the format of
+      the returned dict.
+    """
+    device = model.device
+    feature = batch["inputs"]
+    assert feature.ndim == 3
+
+    feature = feature.to(device)
+    # at entry, feature is (N, T, C)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    encoder_out, encoder_out_lens = model.encoder(
+        x=feature, x_lens=feature_lens
+    )
+    hyps = []
+    batch_size = encoder_out.size(0)
+
+    for i in range(batch_size):
+        # fmt: off
+        encoder_out_i = encoder_out[i:i+1, :encoder_out_lens[i]]
+        # fmt: on
+        if params.decoding_method == "greedy_search":
+            hyp = greedy_search(model=model, encoder_out=encoder_out_i)
+        elif params.decoding_method == "beam_search":
+            hyp = beam_search(
+                model=model, encoder_out=encoder_out_i, beam=params.beam_size
+            )
+        else:
+            raise ValueError(
+                f"Unsupported decoding method: {params.decoding_method}"
+            )
+        hyps.append(sp.decode(hyp).split())
+
+    if params.decoding_method == "greedy_search":
+        return {"greedy_search": hyps}
+    else:
+        return {f"beam_{params.beam_size}": hyps}
+
+
+def decode_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+) -> Dict[str, List[Tuple[List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      dl:
+        PyTorch's dataloader containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+    Returns:
+      Return a dict, whose key may be "greedy_search" if greedy search
+      is used, or it may be "beam_7" if beam size of 7 is used.
+      Its value is a list of tuples. Each tuple contains two elements:
+      The first is the reference transcript, and the second is the
+      predicted result.
+    """
+    num_cuts = 0
+
+    try:
+        num_batches = len(dl)
+    except TypeError:
+        num_batches = "?"
+
+    if params.decoding_method == "greedy_search":
+        log_interval = 100
+    else:
+        log_interval = 2
+
+    results = defaultdict(list)
+    for batch_idx, batch in enumerate(dl):
+        texts = batch["supervisions"]["text"]
+
+        hyps_dict = decode_one_batch(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+        )
+
+        for name, hyps in hyps_dict.items():
+            this_batch = []
+            assert len(hyps) == len(texts)
+            for hyp_words, ref_text in zip(hyps, texts):
+                ref_words = ref_text.split()
+                this_batch.append((ref_words, hyp_words))
+
+            results[name].extend(this_batch)
+
+        num_cuts += len(texts)
+
+        if batch_idx % log_interval == 0:
+            batch_str = f"{batch_idx}/{num_batches}"
+
+            logging.info(
+                f"batch {batch_str}, cuts processed until now is {num_cuts}"
+            )
+    return results
+
+
+def save_results(
+    params: AttributeDict,
+    test_set_name: str,
+    results_dict: Dict[str, List[Tuple[List[int], List[int]]]],
+):
+    test_set_wers = dict()
+    for key, results in results_dict.items():
+        recog_path = (
+            params.res_dir / f"recogs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        store_transcripts(filename=recog_path, texts=results)
+        logging.info(f"The transcripts are stored in {recog_path}")
+
+        # The following prints out WERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = (
+            params.res_dir / f"errs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        with open(errs_filename, "w") as f:
+            wer = write_error_stats(
+                f, f"{test_set_name}-{key}", results, enable_log=True
+            )
+            test_set_wers[key] = wer
+
+        logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_wers = sorted(test_set_wers.items(), key=lambda x: x[1])
+    errs_info = (
+        params.res_dir
+        / f"wer-summary-{test_set_name}-{key}-{params.suffix}.txt"
+    )
+    with open(errs_info, "w") as f:
+        print("settings\tWER", file=f)
+        for key, val in test_set_wers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, WER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_wers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    LibriSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    assert params.decoding_method in ("greedy_search", "beam_search")
+    params.res_dir = params.exp_dir / params.decoding_method
+
+    params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
+    if params.decoding_method == "beam_search":
+        params.suffix += f"-beam-{params.beam_size}"
+
+    setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
+    logging.info("Decoding started")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <sos/eos> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.sos_id = sp.piece_to_id("<sos/eos>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    if params.avg == 1:
+        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    else:
+        start = params.epoch - params.avg + 1
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if start >= 0:
+                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+        logging.info(f"averaging {filenames}")
+        model.to(device)
+        model.load_state_dict(average_checkpoints(filenames, device=device))
+
+    model.to(device)
+    model.eval()
+    model.device = device
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    librispeech = LibriSpeechAsrDataModule(args)
+
+    test_clean_cuts = librispeech.test_clean_cuts()
+    test_other_cuts = librispeech.test_other_cuts()
+
+    test_clean_dl = librispeech.test_dataloaders(test_clean_cuts)
+    test_other_dl = librispeech.test_dataloaders(test_other_cuts)
+
+    test_sets = ["test-clean", "test-other"]
+    test_dl = [test_clean_dl, test_other_dl]
+
+    for test_set, test_dl in zip(test_sets, test_dl):
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+            sp=sp,
+        )
+
+        save_results(
+            params=params,
+            test_set_name=test_set,
+            results_dict=results_dict,
+        )
+
+    logging.info("Done!")
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer/decoder.py

@@ -0,0 +1,101 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Optional, Tuple
+
+import torch
+import torch.nn as nn
+
+
+# TODO(fangjun): Support switching between LSTM and GRU
+class Decoder(nn.Module):
+    def __init__(
+        self,
+        vocab_size: int,
+        embedding_dim: int,
+        blank_id: int,
+        sos_id: int,
+        num_layers: int,
+        hidden_dim: int,
+        output_dim: int,
+        embedding_dropout: float = 0.0,
+        rnn_dropout: float = 0.0,
+    ):
+        """
+        Args:
+          vocab_size:
+            Number of tokens of the modeling unit including blank.
+          embedding_dim:
+            Dimension of the input embedding.
+          blank_id:
+            The ID of the blank symbol.
+          sos_id:
+            The ID of the SOS symbol.
+          num_layers:
+            Number of LSTM layers.
+          hidden_dim:
+            Hidden dimension of LSTM layers.
+          output_dim:
+            Output dimension of the decoder.
+          embedding_dropout:
+            Dropout rate for the embedding layer.
+          rnn_dropout:
+            Dropout for LSTM layers.
+        """
+        super().__init__()
+        self.embedding = nn.Embedding(
+            num_embeddings=vocab_size,
+            embedding_dim=embedding_dim,
+            padding_idx=blank_id,
+        )
+        self.embedding_dropout = nn.Dropout(embedding_dropout)
+        # TODO(fangjun): Use layer normalized LSTM
+        self.rnn = nn.LSTM(
+            input_size=embedding_dim,
+            hidden_size=hidden_dim,
+            num_layers=num_layers,
+            batch_first=True,
+            dropout=rnn_dropout,
+        )
+        self.blank_id = blank_id
+        self.sos_id = sos_id
+        self.output_linear = nn.Linear(hidden_dim, output_dim)
+
+    def forward(
+        self,
+        y: torch.Tensor,
+        states: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+    ) -> Tuple[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]:
+        """
+        Args:
+          y:
+            A 2-D tensor of shape (N, U) with BOS prepended.
+          states:
+            A tuple of two tensors containing the states information of
+            LSTM layers in this decoder.
+        Returns:
+          Return a tuple containing:
+
+            - rnn_output, a tensor of shape (N, U, C)
+            - (h, c), containing the state information for LSTM layers.
+              Both are of shape (num_layers, N, C)
+        """
+        embeding_out = self.embedding(y)
+        embeding_out = self.embedding_dropout(embeding_out)
+        rnn_out, (h, c) = self.rnn(embeding_out, states)
+        out = self.output_linear(rnn_out)
+
+        return out, (h, c)

egs/librispeech/ASR/transducer/encoder_interface.py

@@ -0,0 +1,43 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Tuple
+
+import torch
+import torch.nn as nn
+
+
+class EncoderInterface(nn.Module):
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            A tensor of shape (batch_size, input_seq_len, num_features)
+            containing the input features.
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames
+            in `x` before padding.
+        Returns:
+          Return a tuple containing two tensors:
+            - encoder_out, a tensor of (batch_size, out_seq_len, output_dim)
+              containing unnormalized probabilities, i.e., the output of a
+              linear layer.
+            - encoder_out_lens, a tensor of shape (batch_size,) containing
+              the number of frames in `encoder_out` before padding.
+        """
+        raise NotImplementedError("Please implement it in a subclass")

egs/librispeech/ASR/transducer/export.py

@@ -0,0 +1,250 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This script converts several saved checkpoints
+# to a single one using model averaging.
+"""
+Usage:
+./transducer/export.py \
+  --exp-dir ./transducer/exp \
+  --bpe-model data/lang_bpe_500/bpe.model \
+  --epoch 26 \
+  --avg 12
+
+It will generate a file exp_dir/pretrained.pt
+
+To use the generated file with `transducer/decode.py`, you can do:
+
+    cd /path/to/exp_dir
+    ln -s pretrained.pt epoch-9999.pt
+
+    cd /path/to/egs/librispeech/ASR
+    ./transducer/decode.py \
+        --exp-dir ./transducer/exp \
+        --epoch 9999 \
+        --avg 1 \
+        --max-duration 1 \
+        --bpe-model data/lang_bpe_500/bpe.model
+"""
+
+import argparse
+import logging
+from pathlib import Path
+
+import sentencepiece as spm
+import torch
+from conformer import Conformer
+from decoder import Decoder
+from joiner import Joiner
+from model import Transducer
+
+from icefall.checkpoint import average_checkpoints, load_checkpoint
+from icefall.env import get_env_info
+from icefall.utils import AttributeDict, str2bool
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=26,
+        help="It specifies the checkpoint to use for decoding."
+        "Note: Epoch counts from 0.",
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=12,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="transducer/exp",
+        help="""It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--jit",
+        type=str2bool,
+        default=False,
+        help="""True to save a model after applying torch.jit.script.
+        """,
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            # parameters for conformer
+            "feature_dim": 80,
+            "encoder_out_dim": 512,
+            "subsampling_factor": 4,
+            "attention_dim": 512,
+            "nhead": 8,
+            "dim_feedforward": 2048,
+            "num_encoder_layers": 12,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            # decoder params
+            "decoder_embedding_dim": 1024,
+            "num_decoder_layers": 4,
+            "decoder_hidden_dim": 512,
+            "env_info": get_env_info(),
+        }
+    )
+    return params
+
+
+def get_encoder_model(params: AttributeDict):
+    encoder = Conformer(
+        num_features=params.feature_dim,
+        output_dim=params.encoder_out_dim,
+        subsampling_factor=params.subsampling_factor,
+        d_model=params.attention_dim,
+        nhead=params.nhead,
+        dim_feedforward=params.dim_feedforward,
+        num_encoder_layers=params.num_encoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+    return encoder
+
+
+def get_decoder_model(params: AttributeDict):
+    decoder = Decoder(
+        vocab_size=params.vocab_size,
+        embedding_dim=params.decoder_embedding_dim,
+        blank_id=params.blank_id,
+        sos_id=params.sos_id,
+        num_layers=params.num_decoder_layers,
+        hidden_dim=params.decoder_hidden_dim,
+        output_dim=params.encoder_out_dim,
+    )
+    return decoder
+
+
+def get_joiner_model(params: AttributeDict):
+    joiner = Joiner(
+        input_dim=params.encoder_out_dim,
+        output_dim=params.vocab_size,
+    )
+    return joiner
+
+
+def get_transducer_model(params: AttributeDict):
+    encoder = get_encoder_model(params)
+    decoder = get_decoder_model(params)
+    joiner = get_joiner_model(params)
+
+    model = Transducer(
+        encoder=encoder,
+        decoder=decoder,
+        joiner=joiner,
+    )
+    return model
+
+
+def main():
+    args = get_parser().parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    assert args.jit is False, "Support torchscript will be added later"
+
+    params = get_params()
+    params.update(vars(args))
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <sos/eos> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.sos_id = sp.piece_to_id("<sos/eos>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    model.to(device)
+
+    if params.avg == 1:
+        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    else:
+        start = params.epoch - params.avg + 1
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if start >= 0:
+                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+        logging.info(f"averaging {filenames}")
+        model.to(device)
+        model.load_state_dict(average_checkpoints(filenames, device=device))
+
+    model.eval()
+
+    model.to("cpu")
+    model.eval()
+
+    if params.jit:
+        logging.info("Using torch.jit.script")
+        model = torch.jit.script(model)
+        filename = params.exp_dir / "cpu_jit.pt"
+        model.save(str(filename))
+        logging.info(f"Saved to {filename}")
+    else:
+        logging.info("Not using torch.jit.script")
+        # Save it using a format so that it can be loaded
+        # by :func:`load_checkpoint`
+        filename = params.exp_dir / "pretrained.pt"
+        torch.save({"model": model.state_dict()}, str(filename))
+        logging.info(f"Saved to {filename}")
+
+
+if __name__ == "__main__":
+    formatter = (
+        "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+    )
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/librispeech/ASR/transducer/joiner.py

@@ -0,0 +1,55 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class Joiner(nn.Module):
+    def __init__(self, input_dim: int, output_dim: int):
+        super().__init__()
+
+        self.output_linear = nn.Linear(input_dim, output_dim)
+
+    def forward(
+        self, encoder_out: torch.Tensor, decoder_out: torch.Tensor
+    ) -> torch.Tensor:
+        """
+        Args:
+          encoder_out:
+            Output from the encoder. Its shape is (N, T, C).
+          decoder_out:
+            Output from the decoder. Its shape is (N, U, C).
+        Returns:
+          Return a tensor of shape (N, T, U, C).
+        """
+        assert encoder_out.ndim == decoder_out.ndim == 3
+        assert encoder_out.size(0) == decoder_out.size(0)
+        assert encoder_out.size(2) == decoder_out.size(2)
+
+        encoder_out = encoder_out.unsqueeze(2)
+        # Now encoder_out is (N, T, 1, C)
+
+        decoder_out = decoder_out.unsqueeze(1)
+        # Now decoder_out is (N, 1, U, C)
+
+        logit = encoder_out + decoder_out
+        logit = F.relu(logit)
+
+        output = self.output_linear(logit)
+
+        return output

egs/librispeech/ASR/transducer/model.py

@@ -0,0 +1,127 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Note we use `rnnt_loss` from torchaudio, which exists only in
+torchaudio >= v0.10.0. It also means you have to use torch >= v1.10.0
+"""
+import k2
+import torch
+import torch.nn as nn
+import torchaudio
+import torchaudio.functional
+from encoder_interface import EncoderInterface
+
+from icefall.utils import add_sos
+
+assert hasattr(torchaudio.functional, "rnnt_loss"), (
+    f"Current torchaudio version: {torchaudio.__version__}\n"
+    "Please install a version >= 0.10.0"
+)
+
+
+class Transducer(nn.Module):
+    """It implements https://arxiv.org/pdf/1211.3711.pdf
+    "Sequence Transduction with Recurrent Neural Networks"
+    """
+
+    def __init__(
+        self,
+        encoder: EncoderInterface,
+        decoder: nn.Module,
+        joiner: nn.Module,
+    ):
+        """
+        Args:
+          encoder:
+            It is the transcription network in the paper. Its accepts
+            two inputs: `x` of (N, T, C) and `x_lens` of shape (N,).
+            It returns two tensors: `logits` of shape (N, T, C) and
+            `logit_lens` of shape (N,).
+          decoder:
+            It is the prediction network in the paper. Its input shape
+            is (N, U) and its output shape is (N, U, C). It should contain
+            two attributes: `blank_id` and `sos_id`.
+          joiner:
+            It has two inputs with shapes: (N, T, C) and (N, U, C). Its
+            output shape is (N, T, U, C). Note that its output contains
+            unnormalized probs, i.e., not processed by log-softmax.
+        """
+        super().__init__()
+        assert isinstance(encoder, EncoderInterface)
+        assert hasattr(decoder, "blank_id")
+        assert hasattr(decoder, "sos_id")
+
+        self.encoder = encoder
+        self.decoder = decoder
+        self.joiner = joiner
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        x_lens: torch.Tensor,
+        y: k2.RaggedTensor,
+    ) -> torch.Tensor:
+        """
+        Args:
+          x:
+            A 3-D tensor of shape (N, T, C).
+          x_lens:
+            A 1-D tensor of shape (N,). It contains the number of frames in `x`
+            before padding.
+          y:
+            A ragged tensor with 2 axes [utt][label]. It contains labels of each
+            utterance.
+        Returns:
+          Return the transducer loss.
+        """
+        assert x.ndim == 3, x.shape
+        assert x_lens.ndim == 1, x_lens.shape
+        assert y.num_axes == 2, y.num_axes
+
+        assert x.size(0) == x_lens.size(0) == y.dim0
+
+        encoder_out, x_lens = self.encoder(x, x_lens)
+        assert torch.all(x_lens > 0)
+
+        # Now for the decoder, i.e., the prediction network
+        row_splits = y.shape.row_splits(1)
+        y_lens = row_splits[1:] - row_splits[:-1]
+
+        blank_id = self.decoder.blank_id
+        sos_id = self.decoder.sos_id
+        sos_y = add_sos(y, sos_id=sos_id)
+
+        sos_y_padded = sos_y.pad(mode="constant", padding_value=blank_id)
+
+        decoder_out, _ = self.decoder(sos_y_padded)
+
+        logits = self.joiner(encoder_out, decoder_out)
+
+        # rnnt_loss requires 0 padded targets
+        # Note: y does not start with SOS
+        y_padded = y.pad(mode="constant", padding_value=0)
+
+        loss = torchaudio.functional.rnnt_loss(
+            logits=logits,
+            targets=y_padded,
+            logit_lengths=x_lens,
+            target_lengths=y_lens,
+            blank=blank_id,
+            reduction="sum",
+        )
+
+        return loss

egs/librispeech/ASR/transducer/pretrained.py

@@ -0,0 +1,299 @@
+#!/usr/bin/env python3
+# Copyright      2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Usage:
+
+./transducer/pretrained.py \
+        --checkpoint ./transducer/exp/pretrained.pt \
+        --bpe-model ./data/lang_bpe_500/bpe.model \
+        --method greedy_search \
+        /path/to/foo.wav \
+        /path/to/bar.wav \
+
+You can also use `./transducer/exp/epoch-xx.pt`.
+
+Note: ./transducer/exp/pretrained.pt is generated by
+./transducer/export.py
+"""
+
+
+import argparse
+import logging
+import math
+from typing import List
+
+import kaldifeat
+import sentencepiece as spm
+import torch
+import torchaudio
+from beam_search import beam_search, greedy_search
+from conformer import Conformer
+from decoder import Decoder
+from joiner import Joiner
+from model import Transducer
+from torch.nn.utils.rnn import pad_sequence
+
+from icefall.env import get_env_info
+from icefall.utils import AttributeDict
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--checkpoint",
+        type=str,
+        required=True,
+        help="Path to the checkpoint. "
+        "The checkpoint is assumed to be saved by "
+        "icefall.checkpoint.save_checkpoint().",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        help="""Path to bpe.model.
+        Used only when method is ctc-decoding.
+        """,
+    )
+
+    parser.add_argument(
+        "--method",
+        type=str,
+        default="greedy_search",
+        help="""Possible values are:
+          - greedy_search
+          - beam_search
+        """,
+    )
+
+    parser.add_argument(
+        "sound_files",
+        type=str,
+        nargs="+",
+        help="The input sound file(s) to transcribe. "
+        "Supported formats are those supported by torchaudio.load(). "
+        "For example, wav and flac are supported. "
+        "The sample rate has to be 16kHz.",
+    )
+
+    parser.add_argument(
+        "--beam-size",
+        type=int,
+        default=5,
+        help="Used only when --method is beam_search",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            "sample_rate": 16000,
+            # parameters for conformer
+            "feature_dim": 80,
+            "encoder_out_dim": 512,
+            "subsampling_factor": 4,
+            "attention_dim": 512,
+            "nhead": 8,
+            "dim_feedforward": 2048,
+            "num_encoder_layers": 12,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            # decoder params
+            "decoder_embedding_dim": 1024,
+            "num_decoder_layers": 4,
+            "decoder_hidden_dim": 512,
+            "env_info": get_env_info(),
+        }
+    )
+    return params
+
+
+def get_encoder_model(params: AttributeDict):
+    encoder = Conformer(
+        num_features=params.feature_dim,
+        output_dim=params.encoder_out_dim,
+        subsampling_factor=params.subsampling_factor,
+        d_model=params.attention_dim,
+        nhead=params.nhead,
+        dim_feedforward=params.dim_feedforward,
+        num_encoder_layers=params.num_encoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+    return encoder
+
+
+def get_decoder_model(params: AttributeDict):
+    decoder = Decoder(
+        vocab_size=params.vocab_size,
+        embedding_dim=params.decoder_embedding_dim,
+        blank_id=params.blank_id,
+        sos_id=params.sos_id,
+        num_layers=params.num_decoder_layers,
+        hidden_dim=params.decoder_hidden_dim,
+        output_dim=params.encoder_out_dim,
+    )
+    return decoder
+
+
+def get_joiner_model(params: AttributeDict):
+    joiner = Joiner(
+        input_dim=params.encoder_out_dim,
+        output_dim=params.vocab_size,
+    )
+    return joiner
+
+
+def get_transducer_model(params: AttributeDict):
+    encoder = get_encoder_model(params)
+    decoder = get_decoder_model(params)
+    joiner = get_joiner_model(params)
+
+    model = Transducer(
+        encoder=encoder,
+        decoder=decoder,
+        joiner=joiner,
+    )
+    return model
+
+
+def read_sound_files(
+    filenames: List[str], expected_sample_rate: float
+) -> List[torch.Tensor]:
+    """Read a list of sound files into a list 1-D float32 torch tensors.
+    Args:
+      filenames:
+        A list of sound filenames.
+      expected_sample_rate:
+        The expected sample rate of the sound files.
+    Returns:
+      Return a list of 1-D float32 torch tensors.
+    """
+    ans = []
+    for f in filenames:
+        wave, sample_rate = torchaudio.load(f)
+        assert sample_rate == expected_sample_rate, (
+            f"expected sample rate: {expected_sample_rate}. "
+            f"Given: {sample_rate}"
+        )
+        # We use only the first channel
+        ans.append(wave[0])
+    return ans
+
+
+def main():
+    parser = get_parser()
+    args = parser.parse_args()
+
+    params = get_params()
+
+    params.update(vars(args))
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <sos/eos> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.sos_id = sp.piece_to_id("<sos/eos>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(f"{params}")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"device: {device}")
+
+    logging.info("Creating model")
+    model = get_transducer_model(params)
+
+    checkpoint = torch.load(args.checkpoint, map_location="cpu")
+    model.load_state_dict(checkpoint["model"], strict=False)
+    model.to(device)
+    model.eval()
+    model.device = device
+
+    logging.info("Constructing Fbank computer")
+    opts = kaldifeat.FbankOptions()
+    opts.device = device
+    opts.frame_opts.dither = 0
+    opts.frame_opts.snip_edges = False
+    opts.frame_opts.samp_freq = params.sample_rate
+    opts.mel_opts.num_bins = params.feature_dim
+
+    fbank = kaldifeat.Fbank(opts)
+
+    logging.info(f"Reading sound files: {params.sound_files}")
+    waves = read_sound_files(
+        filenames=params.sound_files, expected_sample_rate=params.sample_rate
+    )
+    waves = [w.to(device) for w in waves]
+
+    logging.info("Decoding started")
+    features = fbank(waves)
+    feature_lengths = [f.size(0) for f in features]
+
+    features = pad_sequence(
+        features, batch_first=True, padding_value=math.log(1e-10)
+    )
+
+    feature_lengths = torch.tensor(feature_lengths, device=device)
+
+    with torch.no_grad():
+        encoder_out, encoder_out_lens = model.encoder(
+            x=features, x_lens=feature_lengths
+        )
+
+    num_waves = encoder_out.size(0)
+    hyps = []
+    for i in range(num_waves):
+        # fmt: off
+        encoder_out_i = encoder_out[i:i+1, :encoder_out_lens[i]]
+        # fmt: on
+        if params.method == "greedy_search":
+            hyp = greedy_search(model=model, encoder_out=encoder_out_i)
+        elif params.method == "beam_search":
+            hyp = beam_search(
+                model=model, encoder_out=encoder_out_i, beam=params.beam_size
+            )
+        else:
+            raise ValueError(f"Unsupported method: {params.method}")
+
+        hyps.append(sp.decode(hyp).split())
+
+    s = "\n"
+    for filename, hyp in zip(params.sound_files, hyps):
+        words = " ".join(hyp)
+        s += f"{filename}:\n{words}\n\n"
+    logging.info(s)
+
+    logging.info("Decoding Done")
+
+
+if __name__ == "__main__":
+    formatter = (
+        "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+    )
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()

egs/librispeech/ASR/transducer/rnn.py

@@ -212,6 +212,24 @@ class LayerNormLSTMCell(nn.Module):
             if "layernorm" not in name:
                 nn.init.uniform_(weight, -stdv, stdv)
 
+            if "bias_ih" in name or "bias_hh" in name:
+                # See the paper
+                # An Empirical Exploration of Recurrent Network Architectures
+                # https://proceedings.mlr.press/v37/jozefowicz15.pdf
+                #
+                # It recommends initializing the bias of the forget gate to
+                # a large value, such as 1 or 2. In PyTorch, there are two
+                # biases for the forget gate, we set both of them to 1 here.
+                #
+                # See also https://arxiv.org/pdf/1804.04849.pdf
+                assert weight.ndim == 1
+                # Layout of the bias:
+                # | in_gate | forget_gate | cell_gate | output_gate |
+                start = weight.numel() // 4
+                end = weight.numel() // 2
+                with torch.no_grad():
+                    weight[start:end].fill_(1.0)
+
 
 class LayerNormLSTMLayer(nn.Module):
     """

egs/librispeech/ASR/transducer/subsampling.py

@@ -0,0 +1 @@
+../conformer_ctc/subsampling.py

egs/librispeech/ASR/transducer/test_conformer.py

@@ -0,0 +1,60 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+To run this file, do:
+
+    cd icefall/egs/librispeech/ASR
+    python ./transducer/test_conformer.py
+"""
+
+import torch
+from conformer import Conformer
+
+
+def test_conformer():
+    output_dim = 1024
+    conformer = Conformer(
+        num_features=80,
+        output_dim=output_dim,
+        subsampling_factor=4,
+        d_model=512,
+        nhead=8,
+        dim_feedforward=2048,
+        num_encoder_layers=12,
+        use_feat_batchnorm=True,
+    )
+    N = 3
+    T = 100
+    C = 80
+    x = torch.randn(N, T, C)
+    x_lens = torch.tensor([50, 100, 80])
+    logits, logit_lens = conformer(x, x_lens)
+
+    expected_T = ((T - 1) // 2 - 1) // 2
+    assert logits.shape == (N, expected_T, output_dim)
+    assert logit_lens.max().item() == expected_T
+    print(logits.shape)
+    print(logit_lens)
+
+
+def main():
+    test_conformer()
+
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer/test_decoder.py

@@ -0,0 +1,69 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+To run this file, do:
+
+    cd icefall/egs/librispeech/ASR
+    python ./transducer/test_decoder.py
+"""
+
+import torch
+from decoder import Decoder
+
+
+def test_decoder():
+    vocab_size = 3
+    blank_id = 0
+    sos_id = 2
+    embedding_dim = 128
+    num_layers = 2
+    hidden_dim = 6
+    output_dim = 8
+    N = 3
+    U = 5
+
+    decoder = Decoder(
+        vocab_size=vocab_size,
+        embedding_dim=embedding_dim,
+        blank_id=blank_id,
+        sos_id=sos_id,
+        num_layers=num_layers,
+        hidden_dim=hidden_dim,
+        output_dim=output_dim,
+        embedding_dropout=0.0,
+        rnn_dropout=0.0,
+    )
+    x = torch.randint(1, vocab_size, (N, U))
+    decoder_out, (h, c) = decoder(x)
+
+    assert decoder_out.shape == (N, U, output_dim)
+    assert h.shape == (num_layers, N, hidden_dim)
+    assert c.shape == (num_layers, N, hidden_dim)
+
+    decoder_out, (h, c) = decoder(x, (h, c))
+    assert decoder_out.shape == (N, U, output_dim)
+    assert h.shape == (num_layers, N, hidden_dim)
+    assert c.shape == (num_layers, N, hidden_dim)
+
+
+def main():
+    test_decoder()
+
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer/test_joiner.py

@@ -0,0 +1,50 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+To run this file, do:
+
+    cd icefall/egs/librispeech/ASR
+    python ./transducer/test_joiner.py
+"""
+
+
+import torch
+from joiner import Joiner
+
+
+def test_joiner():
+    N = 2
+    T = 3
+    C = 4
+    U = 5
+
+    joiner = Joiner(C, 10)
+
+    encoder_out = torch.rand(N, T, C)
+    decoder_out = torch.rand(N, U, C)
+
+    joint = joiner(encoder_out, decoder_out)
+    assert joint.shape == (N, T, U, 10)
+
+
+def main():
+    test_joiner()
+
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer/test_rnn.py

@@ -15,9 +15,15 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+"""
+To run this file, do:
+
+    cd icefall/egs/librispeech/ASR
+    python ./transducer/test_rnn.py
+"""
 import torch
 import torch.nn as nn
-from transducer.rnn import (
+from rnn import (
     LayerNormGRU,
     LayerNormGRUCell,
     LayerNormGRULayer,
@@ -499,6 +505,28 @@ def test_layernorm_lstm_with_projection_forward(device="cpu"):
     assert_allclose(x.grad, x_clone.grad)
 
 
+def test_lstm_forget_gate_bias(device="cpu"):
+    input_size = 2
+    hidden_size = 3
+    num_layers = 4
+    bias = True
+
+    lstm = LayerNormLSTM(
+        input_size=input_size,
+        hidden_size=hidden_size,
+        num_layers=num_layers,
+        bias=bias,
+        ln=nn.Identity,
+        device=device,
+    )
+    for name, weight in lstm.named_parameters():
+        if "bias_hh" in name or "bias_ih" in name:
+            start = weight.numel() // 4
+            end = weight.numel() // 2
+            expected = torch.ones(hidden_size).to(weight)
+            assert torch.all(torch.eq(weight[start:end], expected))
+
+
 def test_layernorm_gru_cell_jit(device="cpu"):
     input_size = 10
     hidden_size = 20
@@ -735,6 +763,8 @@ def _test_lstm(device):
     test_layernorm_lstm_with_projection_jit(device)
     test_layernorm_lstm_forward(device)
     test_layernorm_lstm_with_projection_forward(device)
+    #
+    test_lstm_forget_gate_bias(device)
 
 
 def _test_gru(device):

egs/librispeech/ASR/transducer/test_transducer.py

@@ -0,0 +1,89 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+To run this file, do:
+
+    cd icefall/egs/librispeech/ASR
+    python ./transducer/test_transducer.py
+"""
+
+
+import k2
+import torch
+from conformer import Conformer
+from decoder import Decoder
+from joiner import Joiner
+from model import Transducer
+
+
+def test_transducer():
+    # encoder params
+    input_dim = 10
+    output_dim = 20
+
+    # decoder params
+    vocab_size = 3
+    blank_id = 0
+    sos_id = 2
+    embedding_dim = 128
+    num_layers = 2
+
+    encoder = Conformer(
+        num_features=input_dim,
+        output_dim=output_dim,
+        subsampling_factor=4,
+        d_model=512,
+        nhead=8,
+        dim_feedforward=2048,
+        num_encoder_layers=12,
+        use_feat_batchnorm=True,
+    )
+
+    decoder = Decoder(
+        vocab_size=vocab_size,
+        embedding_dim=embedding_dim,
+        blank_id=blank_id,
+        sos_id=sos_id,
+        num_layers=num_layers,
+        hidden_dim=output_dim,
+        output_dim=output_dim,
+        embedding_dropout=0.0,
+        rnn_dropout=0.0,
+    )
+
+    joiner = Joiner(output_dim, vocab_size)
+    transducer = Transducer(encoder=encoder, decoder=decoder, joiner=joiner)
+
+    y = k2.RaggedTensor([[1, 2, 1], [1, 1, 1, 2, 1]])
+    N = y.dim0
+    T = 50
+
+    x = torch.rand(N, T, input_dim)
+    x_lens = torch.randint(low=30, high=T, size=(N,), dtype=torch.int32)
+    x_lens[0] = T
+
+    loss = transducer(x, x_lens, y)
+    print(loss)
+
+
+def main():
+    test_transducer()
+
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer/test_transformer.py

@@ -0,0 +1,60 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+To run this file, do:
+
+    cd icefall/egs/librispeech/ASR
+    python ./transducer/test_transformer.py
+"""
+
+import torch
+from transformer import Transformer
+
+
+def test_transformer():
+    output_dim = 1024
+    transformer = Transformer(
+        num_features=80,
+        output_dim=output_dim,
+        subsampling_factor=4,
+        d_model=512,
+        nhead=8,
+        dim_feedforward=2048,
+        num_encoder_layers=12,
+        use_feat_batchnorm=True,
+    )
+    N = 3
+    T = 100
+    C = 80
+    x = torch.randn(N, T, C)
+    x_lens = torch.tensor([50, 100, 80])
+    logits, logit_lens = transformer(x, x_lens)
+
+    expected_T = ((T - 1) // 2 - 1) // 2
+    assert logits.shape == (N, expected_T, output_dim)
+    assert logit_lens.max().item() == expected_T
+    print(logits.shape)
+    print(logit_lens)
+
+
+def main():
+    test_transformer()
+
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer/train.py

@@ -0,0 +1,743 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang,
+#                                                  Wei Kang
+#                                                  Mingshuang Luo)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Usage:
+
+export CUDA_VISIBLE_DEVICES="0,1,2,3"
+
+./transducer/train.py \
+  --world-size 4 \
+  --num-epochs 30 \
+  --start-epoch 0 \
+  --exp-dir transducer/exp \
+  --full-libri 1 \
+  --max-duration 250 \
+  --lr-factor 2.5
+"""
+
+
+import argparse
+import logging
+from pathlib import Path
+from shutil import copyfile
+from typing import Optional, Tuple
+
+import k2
+import sentencepiece as spm
+import torch
+import torch.multiprocessing as mp
+import torch.nn as nn
+from asr_datamodule import LibriSpeechAsrDataModule
+from conformer import Conformer
+from decoder import Decoder
+from joiner import Joiner
+from lhotse.cut import Cut
+from lhotse.utils import fix_random_seed
+from model import Transducer
+from torch import Tensor
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.nn.utils import clip_grad_norm_
+from torch.utils.tensorboard import SummaryWriter
+from transformer import Noam
+
+from icefall.checkpoint import load_checkpoint
+from icefall.checkpoint import save_checkpoint as save_checkpoint_impl
+from icefall.dist import cleanup_dist, setup_dist
+from icefall.env import get_env_info
+from icefall.utils import AttributeDict, MetricsTracker, setup_logger, str2bool
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--world-size",
+        type=int,
+        default=1,
+        help="Number of GPUs for DDP training.",
+    )
+
+    parser.add_argument(
+        "--master-port",
+        type=int,
+        default=12354,
+        help="Master port to use for DDP training.",
+    )
+
+    parser.add_argument(
+        "--tensorboard",
+        type=str2bool,
+        default=True,
+        help="Should various information be logged in tensorboard.",
+    )
+
+    parser.add_argument(
+        "--num-epochs",
+        type=int,
+        default=30,
+        help="Number of epochs to train.",
+    )
+
+    parser.add_argument(
+        "--start-epoch",
+        type=int,
+        default=0,
+        help="""Resume training from from this epoch.
+        If it is positive, it will load checkpoint from
+        transducer/exp/epoch-{start_epoch-1}.pt
+        """,
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="transducer/exp",
+        help="""The experiment dir.
+        It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--lr-factor",
+        type=float,
+        default=3.0,
+        help="The lr_factor for Noam optimizer",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    """Return a dict containing training parameters.
+
+    All training related parameters that are not passed from the commandline
+    are saved in the variable `params`.
+
+    Commandline options are merged into `params` after they are parsed, so
+    you can also access them via `params`.
+
+    Explanation of options saved in `params`:
+
+        - best_train_loss: Best training loss so far. It is used to select
+                           the model that has the lowest training loss. It is
+                           updated during the training.
+
+        - best_valid_loss: Best validation loss so far. It is used to select
+                           the model that has the lowest validation loss. It is
+                           updated during the training.
+
+        - best_train_epoch: It is the epoch that has the best training loss.
+
+        - best_valid_epoch: It is the epoch that has the best validation loss.
+
+        - batch_idx_train: Used to writing statistics to tensorboard. It
+                           contains number of batches trained so far across
+                           epochs.
+
+        - log_interval:  Print training loss if batch_idx % log_interval` is 0
+
+        - reset_interval: Reset statistics if batch_idx % reset_interval is 0
+
+        - valid_interval:  Run validation if batch_idx % valid_interval is 0
+
+        - feature_dim: The model input dim. It has to match the one used
+                       in computing features.
+
+        - subsampling_factor:  The subsampling factor for the model.
+
+        - use_feat_batchnorm: Whether to do batch normalization for the
+                              input features.
+
+        - attention_dim: Hidden dim for multi-head attention model.
+
+        - num_decoder_layers: Number of decoder layer of transformer decoder.
+
+        - weight_decay:  The weight_decay for the optimizer.
+
+        - warm_step: The warm_step for Noam optimizer.
+    """
+    params = AttributeDict(
+        {
+            "best_train_loss": float("inf"),
+            "best_valid_loss": float("inf"),
+            "best_train_epoch": -1,
+            "best_valid_epoch": -1,
+            "batch_idx_train": 0,
+            "log_interval": 50,
+            "reset_interval": 200,
+            "valid_interval": 3000,  # For the 100h subset, use 800
+            # parameters for conformer
+            "feature_dim": 80,
+            "encoder_out_dim": 512,
+            "subsampling_factor": 4,
+            "attention_dim": 512,
+            "nhead": 8,
+            "dim_feedforward": 2048,
+            "num_encoder_layers": 12,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            # decoder params
+            "decoder_embedding_dim": 1024,
+            "num_decoder_layers": 4,
+            "decoder_hidden_dim": 512,
+            # parameters for Noam
+            "weight_decay": 1e-6,
+            "warm_step": 80000,  # For the 100h subset, use 8k
+            "env_info": get_env_info(),
+        }
+    )
+
+    return params
+
+
+def get_encoder_model(params: AttributeDict):
+    # TODO: We can add an option to switch between Conformer and Transformer
+    encoder = Conformer(
+        num_features=params.feature_dim,
+        output_dim=params.encoder_out_dim,
+        subsampling_factor=params.subsampling_factor,
+        d_model=params.attention_dim,
+        nhead=params.nhead,
+        dim_feedforward=params.dim_feedforward,
+        num_encoder_layers=params.num_encoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+    return encoder
+
+
+def get_decoder_model(params: AttributeDict):
+    decoder = Decoder(
+        vocab_size=params.vocab_size,
+        embedding_dim=params.decoder_embedding_dim,
+        blank_id=params.blank_id,
+        sos_id=params.sos_id,
+        num_layers=params.num_decoder_layers,
+        hidden_dim=params.decoder_hidden_dim,
+        output_dim=params.encoder_out_dim,
+    )
+    return decoder
+
+
+def get_joiner_model(params: AttributeDict):
+    joiner = Joiner(
+        input_dim=params.encoder_out_dim,
+        output_dim=params.vocab_size,
+    )
+    return joiner
+
+
+def get_transducer_model(params: AttributeDict):
+    encoder = get_encoder_model(params)
+    decoder = get_decoder_model(params)
+    joiner = get_joiner_model(params)
+
+    model = Transducer(
+        encoder=encoder,
+        decoder=decoder,
+        joiner=joiner,
+    )
+    return model
+
+
+def load_checkpoint_if_available(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+) -> None:
+    """Load checkpoint from file.
+
+    If params.start_epoch is positive, it will load the checkpoint from
+    `params.start_epoch - 1`. Otherwise, this function does nothing.
+
+    Apart from loading state dict for `model`, `optimizer` and `scheduler`,
+    it also updates `best_train_epoch`, `best_train_loss`, `best_valid_epoch`,
+    and `best_valid_loss` in `params`.
+
+    Args:
+      params:
+        The return value of :func:`get_params`.
+      model:
+        The training model.
+      optimizer:
+        The optimizer that we are using.
+      scheduler:
+        The learning rate scheduler we are using.
+    Returns:
+      Return None.
+    """
+    if params.start_epoch <= 0:
+        return
+
+    filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
+    saved_params = load_checkpoint(
+        filename,
+        model=model,
+        optimizer=optimizer,
+        scheduler=scheduler,
+    )
+
+    keys = [
+        "best_train_epoch",
+        "best_valid_epoch",
+        "batch_idx_train",
+        "best_train_loss",
+        "best_valid_loss",
+    ]
+    for k in keys:
+        params[k] = saved_params[k]
+
+    return saved_params
+
+
+def save_checkpoint(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+    rank: int = 0,
+) -> None:
+    """Save model, optimizer, scheduler and training stats to file.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The training model.
+    """
+    if rank != 0:
+        return
+    filename = params.exp_dir / f"epoch-{params.cur_epoch}.pt"
+    save_checkpoint_impl(
+        filename=filename,
+        model=model,
+        params=params,
+        optimizer=optimizer,
+        scheduler=scheduler,
+        rank=rank,
+    )
+
+    if params.best_train_epoch == params.cur_epoch:
+        best_train_filename = params.exp_dir / "best-train-loss.pt"
+        copyfile(src=filename, dst=best_train_filename)
+
+    if params.best_valid_epoch == params.cur_epoch:
+        best_valid_filename = params.exp_dir / "best-valid-loss.pt"
+        copyfile(src=filename, dst=best_valid_filename)
+
+
+def compute_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    batch: dict,
+    is_training: bool,
+) -> Tuple[Tensor, MetricsTracker]:
+    """
+    Compute CTC loss given the model and its inputs.
+
+    Args:
+      params:
+        Parameters for training. See :func:`get_params`.
+      model:
+        The model for training. It is an instance of Conformer in our case.
+      batch:
+        A batch of data. See `lhotse.dataset.K2SpeechRecognitionDataset()`
+        for the content in it.
+      is_training:
+        True for training. False for validation. When it is True, this
+        function enables autograd during computation; when it is False, it
+        disables autograd.
+    """
+    device = model.device
+    feature = batch["inputs"]
+    # at entry, feature is (N, T, C)
+    assert feature.ndim == 3
+    feature = feature.to(device)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    texts = batch["supervisions"]["text"]
+    y = sp.encode(texts, out_type=int)
+    y = k2.RaggedTensor(y).to(device)
+
+    with torch.set_grad_enabled(is_training):
+        loss = model(x=feature, x_lens=feature_lens, y=y)
+
+    assert loss.requires_grad == is_training
+
+    info = MetricsTracker()
+    info["frames"] = (feature_lens // params.subsampling_factor).sum().item()
+
+    # Note: We use reduction=sum while computing the loss.
+    info["loss"] = loss.detach().cpu().item()
+
+    return loss, info
+
+
+def compute_validation_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    valid_dl: torch.utils.data.DataLoader,
+    world_size: int = 1,
+) -> MetricsTracker:
+    """Run the validation process."""
+    model.eval()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(valid_dl):
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+            is_training=False,
+        )
+        assert loss.requires_grad is False
+        tot_loss = tot_loss + loss_info
+
+    if world_size > 1:
+        tot_loss.reduce(loss.device)
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    if loss_value < params.best_valid_loss:
+        params.best_valid_epoch = params.cur_epoch
+        params.best_valid_loss = loss_value
+
+    return tot_loss
+
+
+def train_one_epoch(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: torch.optim.Optimizer,
+    sp: spm.SentencePieceProcessor,
+    train_dl: torch.utils.data.DataLoader,
+    valid_dl: torch.utils.data.DataLoader,
+    tb_writer: Optional[SummaryWriter] = None,
+    world_size: int = 1,
+) -> None:
+    """Train the model for one epoch.
+
+    The training loss from the mean of all frames is saved in
+    `params.train_loss`. It runs the validation process every
+    `params.valid_interval` batches.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The model for training.
+      optimizer:
+        The optimizer we are using.
+      train_dl:
+        Dataloader for the training dataset.
+      valid_dl:
+        Dataloader for the validation dataset.
+      tb_writer:
+        Writer to write log messages to tensorboard.
+      world_size:
+        Number of nodes in DDP training. If it is 1, DDP is disabled.
+    """
+    model.train()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(train_dl):
+        params.batch_idx_train += 1
+        batch_size = len(batch["supervisions"]["text"])
+
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+            is_training=True,
+        )
+        # summary stats
+        tot_loss = (tot_loss * (1 - 1 / params.reset_interval)) + loss_info
+
+        # NOTE: We use reduction==sum and loss is computed over utterances
+        # in the batch and there is no normalization to it so far.
+
+        optimizer.zero_grad()
+        loss.backward()
+        clip_grad_norm_(model.parameters(), 5.0, 2.0)
+        optimizer.step()
+
+        if batch_idx % params.log_interval == 0:
+            logging.info(
+                f"Epoch {params.cur_epoch}, "
+                f"batch {batch_idx}, loss[{loss_info}], "
+                f"tot_loss[{tot_loss}], batch size: {batch_size}"
+            )
+
+        if batch_idx % params.log_interval == 0:
+
+            if tb_writer is not None:
+                loss_info.write_summary(
+                    tb_writer, "train/current_", params.batch_idx_train
+                )
+                tot_loss.write_summary(
+                    tb_writer, "train/tot_", params.batch_idx_train
+                )
+
+        if batch_idx > 0 and batch_idx % params.valid_interval == 0:
+            logging.info("Computing validation loss")
+            valid_info = compute_validation_loss(
+                params=params,
+                model=model,
+                sp=sp,
+                valid_dl=valid_dl,
+                world_size=world_size,
+            )
+            model.train()
+            logging.info(f"Epoch {params.cur_epoch}, validation: {valid_info}")
+            if tb_writer is not None:
+                valid_info.write_summary(
+                    tb_writer, "train/valid_", params.batch_idx_train
+                )
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    params.train_loss = loss_value
+    if params.train_loss < params.best_train_loss:
+        params.best_train_epoch = params.cur_epoch
+        params.best_train_loss = params.train_loss
+
+
+def run(rank, world_size, args):
+    """
+    Args:
+      rank:
+        It is a value between 0 and `world_size-1`, which is
+        passed automatically by `mp.spawn()` in :func:`main`.
+        The node with rank 0 is responsible for saving checkpoint.
+      world_size:
+        Number of GPUs for DDP training.
+      args:
+        The return value of get_parser().parse_args()
+    """
+    params = get_params()
+    params.update(vars(args))
+    if params.full_libri is False:
+        params.valid_interval = 800
+        params.warm_step = 8000
+
+    fix_random_seed(42)
+    if world_size > 1:
+        setup_dist(rank, world_size, params.master_port)
+
+    setup_logger(f"{params.exp_dir}/log/log-train")
+    logging.info("Training started")
+
+    if args.tensorboard and rank == 0:
+        tb_writer = SummaryWriter(log_dir=f"{params.exp_dir}/tensorboard")
+    else:
+        tb_writer = None
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", rank)
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <sos/eos> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.sos_id = sp.piece_to_id("<sos/eos>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    checkpoints = load_checkpoint_if_available(params=params, model=model)
+
+    model.to(device)
+    if world_size > 1:
+        logging.info("Using DDP")
+        model = DDP(model, device_ids=[rank])
+    model.device = device
+
+    optimizer = Noam(
+        model.parameters(),
+        model_size=params.attention_dim,
+        factor=params.lr_factor,
+        warm_step=params.warm_step,
+        weight_decay=params.weight_decay,
+    )
+
+    if checkpoints and "optimizer" in checkpoints:
+        logging.info("Loading optimizer state dict")
+        optimizer.load_state_dict(checkpoints["optimizer"])
+
+    librispeech = LibriSpeechAsrDataModule(args)
+
+    train_cuts = librispeech.train_clean_100_cuts()
+    if params.full_libri:
+        train_cuts += librispeech.train_clean_360_cuts()
+        train_cuts += librispeech.train_other_500_cuts()
+
+    def remove_short_and_long_utt(c: Cut):
+        # Keep only utterances with duration between 1 second and 20 seconds
+        return 1.0 <= c.duration <= 20.0
+
+    num_in_total = len(train_cuts)
+
+    train_cuts = train_cuts.filter(remove_short_and_long_utt)
+
+    num_left = len(train_cuts)
+    num_removed = num_in_total - num_left
+    removed_percent = num_removed / num_in_total * 100
+
+    logging.info(f"Before removing short and long utterances: {num_in_total}")
+    logging.info(f"After removing short and long utterances: {num_left}")
+    logging.info(f"Removed {num_removed} utterances ({removed_percent:.5f}%)")
+
+    train_dl = librispeech.train_dataloaders(train_cuts)
+
+    valid_cuts = librispeech.dev_clean_cuts()
+    valid_cuts += librispeech.dev_other_cuts()
+    valid_dl = librispeech.valid_dataloaders(valid_cuts)
+
+    scan_pessimistic_batches_for_oom(
+        model=model,
+        train_dl=train_dl,
+        optimizer=optimizer,
+        sp=sp,
+        params=params,
+    )
+
+    for epoch in range(params.start_epoch, params.num_epochs):
+        train_dl.sampler.set_epoch(epoch)
+
+        cur_lr = optimizer._rate
+        if tb_writer is not None:
+            tb_writer.add_scalar(
+                "train/learning_rate", cur_lr, params.batch_idx_train
+            )
+            tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)
+
+        if rank == 0:
+            logging.info("epoch {}, learning rate {}".format(epoch, cur_lr))
+
+        params.cur_epoch = epoch
+
+        train_one_epoch(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            sp=sp,
+            train_dl=train_dl,
+            valid_dl=valid_dl,
+            tb_writer=tb_writer,
+            world_size=world_size,
+        )
+
+        save_checkpoint(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            rank=rank,
+        )
+
+    logging.info("Done!")
+
+    if world_size > 1:
+        torch.distributed.barrier()
+        cleanup_dist()
+
+
+def scan_pessimistic_batches_for_oom(
+    model: nn.Module,
+    train_dl: torch.utils.data.DataLoader,
+    optimizer: torch.optim.Optimizer,
+    sp: spm.SentencePieceProcessor,
+    params: AttributeDict,
+):
+    from lhotse.dataset import find_pessimistic_batches
+
+    logging.info(
+        "Sanity check -- see if any of the batches in epoch 0 would cause OOM."
+    )
+    batches, crit_values = find_pessimistic_batches(train_dl.sampler)
+    for criterion, cuts in batches.items():
+        batch = train_dl.dataset[cuts]
+        try:
+            optimizer.zero_grad()
+            loss, _ = compute_loss(
+                params=params,
+                model=model,
+                sp=sp,
+                batch=batch,
+                is_training=True,
+            )
+            loss.backward()
+            clip_grad_norm_(model.parameters(), 5.0, 2.0)
+            optimizer.step()
+        except RuntimeError as e:
+            if "CUDA out of memory" in str(e):
+                logging.error(
+                    "Your GPU ran out of memory with the current "
+                    "max_duration setting. We recommend decreasing "
+                    "max_duration and trying again.\n"
+                    f"Failing criterion: {criterion} "
+                    f"(={crit_values[criterion]}) ..."
+                )
+            raise
+
+
+def main():
+    parser = get_parser()
+    LibriSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    world_size = args.world_size
+    assert world_size >= 1
+    if world_size > 1:
+        mp.spawn(run, args=(world_size, args), nprocs=world_size, join=True)
+    else:
+        run(rank=0, world_size=1, args=args)
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer/transformer.py

@@ -0,0 +1,429 @@
+# Copyright    2021 University of Chinese Academy of Sciences (author: Han Zhu)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import math
+from typing import Optional, Tuple
+
+import torch
+import torch.nn as nn
+from encoder_interface import EncoderInterface
+from subsampling import Conv2dSubsampling, VggSubsampling
+
+from icefall.utils import make_pad_mask
+
+
+class Transformer(EncoderInterface):
+    def __init__(
+        self,
+        num_features: int,
+        output_dim: int,
+        subsampling_factor: int = 4,
+        d_model: int = 256,
+        nhead: int = 4,
+        dim_feedforward: int = 2048,
+        num_encoder_layers: int = 12,
+        dropout: float = 0.1,
+        normalize_before: bool = True,
+        vgg_frontend: bool = False,
+        use_feat_batchnorm: bool = False,
+    ) -> None:
+        """
+        Args:
+          num_features:
+            The input dimension of the model.
+          output_dim:
+            The output dimension of the model.
+          subsampling_factor:
+            Number of output frames is num_in_frames // subsampling_factor.
+            Currently, subsampling_factor MUST be 4.
+          d_model:
+            Attention dimension.
+          nhead:
+            Number of heads in multi-head attention.
+            Must satisfy d_model // nhead == 0.
+          dim_feedforward:
+            The output dimension of the feedforward layers in encoder.
+          num_encoder_layers:
+            Number of encoder layers.
+          dropout:
+            Dropout in encoder.
+          normalize_before:
+            If True, use pre-layer norm; False to use post-layer norm.
+          vgg_frontend:
+            True to use vgg style frontend for subsampling.
+          use_feat_batchnorm:
+            True to use batchnorm for the input layer.
+        """
+        super().__init__()
+        self.use_feat_batchnorm = use_feat_batchnorm
+        if use_feat_batchnorm:
+            self.feat_batchnorm = nn.BatchNorm1d(num_features)
+
+        self.num_features = num_features
+        self.output_dim = output_dim
+        self.subsampling_factor = subsampling_factor
+        if subsampling_factor != 4:
+            raise NotImplementedError("Support only 'subsampling_factor=4'.")
+
+        # self.encoder_embed converts the input of shape (N, T, num_features)
+        # to the shape (N, T//subsampling_factor, d_model).
+        # That is, it does two things simultaneously:
+        #   (1) subsampling: T -> T//subsampling_factor
+        #   (2) embedding: num_features -> d_model
+        if vgg_frontend:
+            self.encoder_embed = VggSubsampling(num_features, d_model)
+        else:
+            self.encoder_embed = Conv2dSubsampling(num_features, d_model)
+
+        self.encoder_pos = PositionalEncoding(d_model, dropout)
+
+        encoder_layer = TransformerEncoderLayer(
+            d_model=d_model,
+            nhead=nhead,
+            dim_feedforward=dim_feedforward,
+            dropout=dropout,
+            normalize_before=normalize_before,
+        )
+
+        if normalize_before:
+            encoder_norm = nn.LayerNorm(d_model)
+        else:
+            encoder_norm = None
+
+        self.encoder = nn.TransformerEncoder(
+            encoder_layer=encoder_layer,
+            num_layers=num_encoder_layers,
+            norm=encoder_norm,
+        )
+
+        # TODO(fangjun): remove dropout
+        self.encoder_output_layer = nn.Sequential(
+            nn.Dropout(p=dropout), nn.Linear(d_model, output_dim)
+        )
+
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            The input tensor. Its shape is (batch_size, seq_len, feature_dim).
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames in
+            `x` before padding.
+        Returns:
+          Return a tuple containing 2 tensors:
+            - logits, its shape is (batch_size, output_seq_len, output_dim)
+            - logit_lens, a tensor of shape (batch_size,) containing the number
+              of frames in `logits` before padding.
+        """
+        if self.use_feat_batchnorm:
+            x = x.permute(0, 2, 1)  # (N, T, C) -> (N, C, T)
+            x = self.feat_batchnorm(x)
+            x = x.permute(0, 2, 1)  # (N, C, T) -> (N, T, C)
+
+        x = self.encoder_embed(x)
+        x = self.encoder_pos(x)
+        x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+
+        # Caution: We assume the subsampling factor is 4!
+        lengths = ((x_lens - 1) // 2 - 1) // 2
+        assert x.size(0) == lengths.max().item()
+
+        mask = make_pad_mask(lengths)
+        x = self.encoder(x, src_key_padding_mask=mask)  # (T, N, C)
+
+        logits = self.encoder_output_layer(x)
+        logits = logits.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+
+        return logits, lengths
+
+
+class TransformerEncoderLayer(nn.Module):
+    """
+    Modified from torch.nn.TransformerEncoderLayer.
+    Add support of normalize_before,
+    i.e., use layer_norm before the first block.
+
+    Args:
+      d_model:
+        the number of expected features in the input (required).
+      nhead:
+        the number of heads in the multiheadattention models (required).
+      dim_feedforward:
+        the dimension of the feedforward network model (default=2048).
+      dropout:
+        the dropout value (default=0.1).
+      activation:
+        the activation function of intermediate layer, relu or
+        gelu (default=relu).
+      normalize_before:
+        whether to use layer_norm before the first block.
+
+    Examples::
+        >>> encoder_layer = TransformerEncoderLayer(d_model=512, nhead=8)
+        >>> src = torch.rand(10, 32, 512)
+        >>> out = encoder_layer(src)
+    """
+
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        activation: str = "relu",
+        normalize_before: bool = True,
+    ) -> None:
+        super(TransformerEncoderLayer, self).__init__()
+        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=0.0)
+        # Implementation of Feedforward model
+        self.linear1 = nn.Linear(d_model, dim_feedforward)
+        self.dropout = nn.Dropout(dropout)
+        self.linear2 = nn.Linear(dim_feedforward, d_model)
+
+        self.norm1 = nn.LayerNorm(d_model)
+        self.norm2 = nn.LayerNorm(d_model)
+        self.dropout1 = nn.Dropout(dropout)
+        self.dropout2 = nn.Dropout(dropout)
+
+        self.activation = _get_activation_fn(activation)
+
+        self.normalize_before = normalize_before
+
+    def __setstate__(self, state):
+        if "activation" not in state:
+            state["activation"] = nn.functional.relu
+        super(TransformerEncoderLayer, self).__setstate__(state)
+
+    def forward(
+        self,
+        src: torch.Tensor,
+        src_mask: Optional[torch.Tensor] = None,
+        src_key_padding_mask: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        """
+        Pass the input through the encoder layer.
+
+        Args:
+            src: the sequence to the encoder layer (required).
+            src_mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional)
+
+        Shape:
+            src: (S, N, E).
+            src_mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, T is the target sequence length,
+            N is the batch size, E is the feature number
+        """
+        residual = src
+        if self.normalize_before:
+            src = self.norm1(src)
+        src2 = self.self_attn(
+            src,
+            src,
+            src,
+            attn_mask=src_mask,
+            key_padding_mask=src_key_padding_mask,
+        )[0]
+        src = residual + self.dropout1(src2)
+        if not self.normalize_before:
+            src = self.norm1(src)
+
+        residual = src
+        if self.normalize_before:
+            src = self.norm2(src)
+        src2 = self.linear2(self.dropout(self.activation(self.linear1(src))))
+        src = residual + self.dropout2(src2)
+        if not self.normalize_before:
+            src = self.norm2(src)
+        return src
+
+
+def _get_activation_fn(activation: str):
+    if activation == "relu":
+        return nn.functional.relu
+    elif activation == "gelu":
+        return nn.functional.gelu
+
+    raise RuntimeError(
+        "activation should be relu/gelu, not {}".format(activation)
+    )
+
+
+class PositionalEncoding(nn.Module):
+    """This class implements the positional encoding
+    proposed in the following paper:
+
+    - Attention Is All You Need: https://arxiv.org/pdf/1706.03762.pdf
+
+        PE(pos, 2i) = sin(pos / (10000^(2i/d_modle))
+        PE(pos, 2i+1) = cos(pos / (10000^(2i/d_modle))
+
+    Note::
+
+      1 / (10000^(2i/d_model)) = exp(-log(10000^(2i/d_model)))
+                               = exp(-1* 2i / d_model * log(100000))
+                               = exp(2i * -(log(10000) / d_model))
+    """
+
+    def __init__(self, d_model: int, dropout: float = 0.1) -> None:
+        """
+        Args:
+          d_model:
+            Embedding dimension.
+          dropout:
+            Dropout probability to be applied to the output of this module.
+        """
+        super().__init__()
+        self.d_model = d_model
+        self.xscale = math.sqrt(self.d_model)
+        self.dropout = nn.Dropout(p=dropout)
+        # not doing: self.pe = None because of errors thrown by torchscript
+        self.pe = torch.zeros(1, 0, self.d_model, dtype=torch.float32)
+
+    def extend_pe(self, x: torch.Tensor) -> None:
+        """Extend the time t in the positional encoding if required.
+
+        The shape of `self.pe` is (1, T1, d_model). The shape of the input x
+        is (N, T, d_model). If T > T1, then we change the shape of self.pe
+        to (N, T, d_model). Otherwise, nothing is done.
+
+        Args:
+          x:
+            It is a tensor of shape (N, T, C).
+        Returns:
+          Return None.
+        """
+        if self.pe is not None:
+            if self.pe.size(1) >= x.size(1):
+                self.pe = self.pe.to(dtype=x.dtype, device=x.device)
+                return
+        pe = torch.zeros(x.size(1), self.d_model, dtype=torch.float32)
+        position = torch.arange(0, x.size(1), dtype=torch.float32).unsqueeze(1)
+        div_term = torch.exp(
+            torch.arange(0, self.d_model, 2, dtype=torch.float32)
+            * -(math.log(10000.0) / self.d_model)
+        )
+        pe[:, 0::2] = torch.sin(position * div_term)
+        pe[:, 1::2] = torch.cos(position * div_term)
+        pe = pe.unsqueeze(0)
+        # Now pe is of shape (1, T, d_model), where T is x.size(1)
+        self.pe = pe.to(device=x.device, dtype=x.dtype)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Add positional encoding.
+
+        Args:
+          x:
+            Its shape is (N, T, C)
+
+        Returns:
+          Return a tensor of shape (N, T, C)
+        """
+        self.extend_pe(x)
+        x = x * self.xscale + self.pe[:, : x.size(1), :]
+        return self.dropout(x)
+
+
+class Noam(object):
+    """
+    Implements Noam optimizer.
+
+    Proposed in
+    "Attention Is All You Need", https://arxiv.org/pdf/1706.03762.pdf
+
+    Modified from
+    https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/optimizer.py  # noqa
+
+    Args:
+      params:
+        iterable of parameters to optimize or dicts defining parameter groups
+      model_size:
+        attention dimension of the transformer model
+      factor:
+        learning rate factor
+      warm_step:
+        warmup steps
+    """
+
+    def __init__(
+        self,
+        params,
+        model_size: int = 256,
+        factor: float = 10.0,
+        warm_step: int = 25000,
+        weight_decay=0,
+    ) -> None:
+        """Construct an Noam object."""
+        self.optimizer = torch.optim.Adam(
+            params, lr=0, betas=(0.9, 0.98), eps=1e-9, weight_decay=weight_decay
+        )
+        self._step = 0
+        self.warmup = warm_step
+        self.factor = factor
+        self.model_size = model_size
+        self._rate = 0
+
+    @property
+    def param_groups(self):
+        """Return param_groups."""
+        return self.optimizer.param_groups
+
+    def step(self):
+        """Update parameters and rate."""
+        self._step += 1
+        rate = self.rate()
+        for p in self.optimizer.param_groups:
+            p["lr"] = rate
+        self._rate = rate
+        self.optimizer.step()
+
+    def rate(self, step=None):
+        """Implement `lrate` above."""
+        if step is None:
+            step = self._step
+        return (
+            self.factor
+            * self.model_size ** (-0.5)
+            * min(step ** (-0.5), step * self.warmup ** (-1.5))
+        )
+
+    def zero_grad(self):
+        """Reset gradient."""
+        self.optimizer.zero_grad()
+
+    def state_dict(self):
+        """Return state_dict."""
+        return {
+            "_step": self._step,
+            "warmup": self.warmup,
+            "factor": self.factor,
+            "model_size": self.model_size,
+            "_rate": self._rate,
+            "optimizer": self.optimizer.state_dict(),
+        }
+
+    def load_state_dict(self, state_dict):
+        """Load state_dict."""
+        for key, value in state_dict.items():
+            if key == "optimizer":
+                self.optimizer.load_state_dict(state_dict["optimizer"])
+            else:
+                setattr(self, key, value)

egs/librispeech/ASR/transducer_lstm/README.md

@@ -0,0 +1,19 @@
+## Introduction
+
+The encoder consists of LSTM layers in this folder. You can use the
+following command to start the training:
+
+```bash
+cd egs/librispeech/ASR
+
+export CUDA_VISIBLE_DEVICES="0,1,2"
+
+./transducer_lstm/train.py \
+  --world-size 3 \
+  --num-epochs 30 \
+  --start-epoch 0 \
+  --exp-dir transducer_lstm/exp \
+  --full-libri 1 \
+  --max-duration 300 \
+  --lr-factor 3
+```

egs/librispeech/ASR/transducer_lstm/asr_datamodule.py

@@ -0,0 +1 @@
+../transducer/asr_datamodule.py

egs/librispeech/ASR/transducer_lstm/beam_search.py

@@ -0,0 +1,212 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from dataclasses import dataclass
+from typing import Dict, List, Optional, Tuple
+
+import torch
+from model import Transducer
+
+
+def greedy_search(model: Transducer, encoder_out: torch.Tensor) -> List[int]:
+    """
+    Args:
+      model:
+        An instance of `Transducer`.
+      encoder_out:
+        A tensor of shape (N, T, C) from the encoder. Support only N==1 for now.
+    Returns:
+      Return the decoded result.
+    """
+    assert encoder_out.ndim == 3
+
+    # support only batch_size == 1 for now
+    assert encoder_out.size(0) == 1, encoder_out.size(0)
+    blank_id = model.decoder.blank_id
+    device = model.device
+
+    sos = torch.tensor([blank_id], device=device).reshape(1, 1)
+    decoder_out, (h, c) = model.decoder(sos)
+    T = encoder_out.size(1)
+    t = 0
+    hyp = []
+    max_u = 1000  # terminate after this number of steps
+    u = 0
+
+    while t < T and u < max_u:
+        # fmt: off
+        current_encoder_out = encoder_out[:, t:t+1, :]
+        # fmt: on
+        logits = model.joiner(current_encoder_out, decoder_out)
+        # logits is (1, 1, 1, vocab_size)
+
+        log_prob = logits.log_softmax(dim=-1)
+        # log_prob is (1, 1, 1, vocab_size)
+        # TODO: Use logits.argmax()
+        y = log_prob.argmax()
+        if y != blank_id:
+            hyp.append(y.item())
+            y = y.reshape(1, 1)
+            decoder_out, (h, c) = model.decoder(y, (h, c))
+            u += 1
+        else:
+            t += 1
+
+    return hyp
+
+
+@dataclass
+class Hypothesis:
+    ys: List[int]  # the predicted sequences so far
+    log_prob: float  # The log prob of ys
+
+    # Optional decoder state. We assume it is LSTM for now,
+    # so the state is a tuple (h, c)
+    decoder_state: Optional[Tuple[torch.Tensor, torch.Tensor]] = None
+
+
+def beam_search(
+    model: Transducer,
+    encoder_out: torch.Tensor,
+    beam: int = 5,
+) -> List[int]:
+    """
+    It implements Algorithm 1 in https://arxiv.org/pdf/1211.3711.pdf
+
+    espnet/nets/beam_search_transducer.py#L247 is used as a reference.
+
+    Args:
+      model:
+        An instance of `Transducer`.
+      encoder_out:
+        A tensor of shape (N, T, C) from the encoder. Support only N==1 for now.
+      beam:
+        Beam size.
+    Returns:
+      Return the decoded result.
+    """
+    assert encoder_out.ndim == 3
+
+    # support only batch_size == 1 for now
+    assert encoder_out.size(0) == 1, encoder_out.size(0)
+    blank_id = model.decoder.blank_id
+    sos_id = model.decoder.sos_id
+    device = model.device
+
+    sos = torch.tensor([blank_id], device=device).reshape(1, 1)
+    decoder_out, (h, c) = model.decoder(sos)
+    T = encoder_out.size(1)
+    t = 0
+    B = [Hypothesis(ys=[blank_id], log_prob=0.0, decoder_state=None)]
+    max_u = 20000  # terminate after this number of steps
+    u = 0
+
+    cache: Dict[
+        str, Tuple[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]
+    ] = {}
+
+    while t < T and u < max_u:
+        # fmt: off
+        current_encoder_out = encoder_out[:, t:t+1, :]
+        # fmt: on
+        A = B
+        B = []
+        #  for hyp in A:
+        #      for h in A:
+        #          if h.ys == hyp.ys[:-1]:
+        #              # update the score of hyp
+        #              decoder_input = torch.tensor(
+        #                  [h.ys[-1]], device=device
+        #              ).reshape(1, 1)
+        #              decoder_out, _ = model.decoder(
+        #                  decoder_input, h.decoder_state
+        #              )
+        #              logits = model.joiner(current_encoder_out, decoder_out)
+        #              log_prob = logits.log_softmax(dim=-1)
+        #              log_prob = log_prob.squeeze()
+        #              hyp.log_prob += h.log_prob + log_prob[hyp.ys[-1]].item()
+
+        while u < max_u:
+            y_star = max(A, key=lambda hyp: hyp.log_prob)
+            A.remove(y_star)
+
+            # Note: y_star.ys is unhashable, i.e., cannot be used
+            # as a key into a dict
+            cached_key = "_".join(map(str, y_star.ys))
+
+            if cached_key not in cache:
+                decoder_input = torch.tensor(
+                    [y_star.ys[-1]], device=device
+                ).reshape(1, 1)
+
+                decoder_out, decoder_state = model.decoder(
+                    decoder_input,
+                    y_star.decoder_state,
+                )
+                cache[cached_key] = (decoder_out, decoder_state)
+            else:
+                decoder_out, decoder_state = cache[cached_key]
+
+            logits = model.joiner(current_encoder_out, decoder_out)
+            log_prob = logits.log_softmax(dim=-1)
+            # log_prob is (1, 1, 1, vocab_size)
+            log_prob = log_prob.squeeze()
+            # Now log_prob is (vocab_size,)
+
+            # If we choose blank here, add the new hypothesis to B.
+            # Otherwise, add the new hypothesis to A
+
+            # First, choose blank
+            skip_log_prob = log_prob[blank_id]
+            new_y_star_log_prob = y_star.log_prob + skip_log_prob.item()
+
+            # ys[:] returns a copy of ys
+            new_y_star = Hypothesis(
+                ys=y_star.ys[:],
+                log_prob=new_y_star_log_prob,
+                # Caution: Use y_star.decoder_state here
+                decoder_state=y_star.decoder_state,
+            )
+            B.append(new_y_star)
+
+            # Second, choose other labels
+            for i, v in enumerate(log_prob.tolist()):
+                if i in (blank_id, sos_id):
+                    continue
+                new_ys = y_star.ys + [i]
+                new_log_prob = y_star.log_prob + v
+                new_hyp = Hypothesis(
+                    ys=new_ys,
+                    log_prob=new_log_prob,
+                    decoder_state=decoder_state,
+                )
+                A.append(new_hyp)
+            u += 1
+            # check whether B contains more than "beam" elements more probable
+            # than the most probable in A
+            A_most_probable = max(A, key=lambda hyp: hyp.log_prob)
+            B = sorted(
+                [hyp for hyp in B if hyp.log_prob > A_most_probable.log_prob],
+                key=lambda hyp: hyp.log_prob,
+                reverse=True,
+            )
+            if len(B) >= beam:
+                B = B[:beam]
+                break
+        t += 1
+    best_hyp = max(B, key=lambda hyp: hyp.log_prob / len(hyp.ys[1:]))
+    ys = best_hyp.ys[1:]  # [1:] to remove the blank
+    return ys

egs/librispeech/ASR/transducer_lstm/decode.py

@@ -0,0 +1,457 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021 Xiaomi Corporation (Author: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Usage:
+(1) greedy search
+./transducer_lstm/decode.py \
+        --epoch 14 \
+        --avg 7 \
+        --exp-dir ./transducer_lstm/exp \
+        --max-duration 100 \
+        --decoding-method greedy_search
+(2) beam search
+
+./transducer_lstm/decode.py \
+        --epoch 14 \
+        --avg 7 \
+        --exp-dir ./transducer_lstm/exp \
+        --max-duration 100 \
+        --decoding-method beam_search \
+        --beam-size 8
+"""
+
+
+import argparse
+import logging
+from collections import defaultdict
+from pathlib import Path
+from typing import Dict, List, Tuple
+
+import sentencepiece as spm
+import torch
+import torch.nn as nn
+from asr_datamodule import LibriSpeechAsrDataModule
+from beam_search import beam_search, greedy_search
+from decoder import Decoder
+from encoder import LstmEncoder
+from joiner import Joiner
+from model import Transducer
+
+from icefall.checkpoint import average_checkpoints, load_checkpoint
+from icefall.env import get_env_info
+from icefall.utils import (
+    AttributeDict,
+    setup_logger,
+    store_transcripts,
+    write_error_stats,
+)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=77,
+        help="It specifies the checkpoint to use for decoding."
+        "Note: Epoch counts from 0.",
+    )
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=55,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="transducer_lstm/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="greedy_search",
+        help="""Possible values are:
+          - greedy_search
+          - beam_search
+        """,
+    )
+
+    parser.add_argument(
+        "--beam-size",
+        type=int,
+        default=5,
+        help="Used only when --decoding-method is beam_search",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            # parameters for conformer
+            "feature_dim": 80,
+            "encoder_out_dim": 512,
+            "subsampling_factor": 4,
+            "encoder_hidden_size": 1024,
+            "num_encoder_layers": 4,
+            "proj_size": 512,
+            "vgg_frontend": False,
+            # decoder params
+            "decoder_embedding_dim": 1024,
+            "num_decoder_layers": 4,
+            "decoder_hidden_dim": 512,
+            "env_info": get_env_info(),
+        }
+    )
+    return params
+
+
+def get_encoder_model(params: AttributeDict):
+    encoder = LstmEncoder(
+        num_features=params.feature_dim,
+        hidden_size=params.encoder_hidden_size,
+        output_dim=params.encoder_out_dim,
+        subsampling_factor=params.subsampling_factor,
+        num_encoder_layers=params.num_encoder_layers,
+        vgg_frontend=params.vgg_frontend,
+    )
+    return encoder
+
+
+def get_decoder_model(params: AttributeDict):
+    decoder = Decoder(
+        vocab_size=params.vocab_size,
+        embedding_dim=params.decoder_embedding_dim,
+        blank_id=params.blank_id,
+        sos_id=params.sos_id,
+        num_layers=params.num_decoder_layers,
+        hidden_dim=params.decoder_hidden_dim,
+        output_dim=params.encoder_out_dim,
+    )
+    return decoder
+
+
+def get_joiner_model(params: AttributeDict):
+    joiner = Joiner(
+        input_dim=params.encoder_out_dim,
+        output_dim=params.vocab_size,
+    )
+    return joiner
+
+
+def get_transducer_model(params: AttributeDict):
+    encoder = get_encoder_model(params)
+    decoder = get_decoder_model(params)
+    joiner = get_joiner_model(params)
+
+    model = Transducer(
+        encoder=encoder,
+        decoder=decoder,
+        joiner=joiner,
+    )
+    return model
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    batch: dict,
+) -> Dict[str, List[List[str]]]:
+    """Decode one batch and return the result in a dict. The dict has the
+    following format:
+
+        - key: It indicates the setting used for decoding. For example,
+               if greedy_search is used, it would be "greedy_search"
+               If beam search with a beam size of 7 is used, it would be
+               "beam_7"
+        - value: It contains the decoding result. `len(value)` equals to
+                 batch size. `value[i]` is the decoding result for the i-th
+                 utterance in the given batch.
+    Args:
+      params:
+        It's the return value of :func:`get_params`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+      batch:
+        It is the return value from iterating
+        `lhotse.dataset.K2SpeechRecognitionDataset`. See its documentation
+        for the format of the `batch`.
+    Returns:
+      Return the decoding result. See above description for the format of
+      the returned dict.
+    """
+    device = model.device
+    feature = batch["inputs"]
+    assert feature.ndim == 3
+
+    feature = feature.to(device)
+    # at entry, feature is (N, T, C)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    encoder_out, encoder_out_lens = model.encoder(
+        x=feature, x_lens=feature_lens
+    )
+    hyps = []
+    batch_size = encoder_out.size(0)
+
+    for i in range(batch_size):
+        # fmt: off
+        encoder_out_i = encoder_out[i:i+1, :encoder_out_lens[i]]
+        # fmt: on
+        if params.decoding_method == "greedy_search":
+            hyp = greedy_search(model=model, encoder_out=encoder_out_i)
+        elif params.decoding_method == "beam_search":
+            hyp = beam_search(
+                model=model, encoder_out=encoder_out_i, beam=params.beam_size
+            )
+        else:
+            raise ValueError(
+                f"Unsupported decoding method: {params.decoding_method}"
+            )
+        hyps.append(sp.decode(hyp).split())
+
+    if params.decoding_method == "greedy_search":
+        return {"greedy_search": hyps}
+    else:
+        return {f"beam_{params.beam_size}": hyps}
+
+
+def decode_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+) -> Dict[str, List[Tuple[List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      dl:
+        PyTorch's dataloader containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+    Returns:
+      Return a dict, whose key may be "greedy_search" if greedy search
+      is used, or it may be "beam_7" if beam size of 7 is used.
+      Its value is a list of tuples. Each tuple contains two elements:
+      The first is the reference transcript, and the second is the
+      predicted result.
+    """
+    num_cuts = 0
+
+    try:
+        num_batches = len(dl)
+    except TypeError:
+        num_batches = "?"
+
+    if params.decoding_method == "greedy_search":
+        log_interval = 100
+    else:
+        log_interval = 2
+
+    results = defaultdict(list)
+    for batch_idx, batch in enumerate(dl):
+        texts = batch["supervisions"]["text"]
+
+        hyps_dict = decode_one_batch(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+        )
+
+        for name, hyps in hyps_dict.items():
+            this_batch = []
+            assert len(hyps) == len(texts)
+            for hyp_words, ref_text in zip(hyps, texts):
+                ref_words = ref_text.split()
+                this_batch.append((ref_words, hyp_words))
+
+            results[name].extend(this_batch)
+
+        num_cuts += len(texts)
+
+        if batch_idx % log_interval == 0:
+            batch_str = f"{batch_idx}/{num_batches}"
+
+            logging.info(
+                f"batch {batch_str}, cuts processed until now is {num_cuts}"
+            )
+    return results
+
+
+def save_results(
+    params: AttributeDict,
+    test_set_name: str,
+    results_dict: Dict[str, List[Tuple[List[int], List[int]]]],
+):
+    test_set_wers = dict()
+    for key, results in results_dict.items():
+        recog_path = (
+            params.res_dir / f"recogs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        store_transcripts(filename=recog_path, texts=results)
+        logging.info(f"The transcripts are stored in {recog_path}")
+
+        # The following prints out WERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = (
+            params.res_dir / f"errs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        with open(errs_filename, "w") as f:
+            wer = write_error_stats(
+                f, f"{test_set_name}-{key}", results, enable_log=True
+            )
+            test_set_wers[key] = wer
+
+        logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_wers = sorted(test_set_wers.items(), key=lambda x: x[1])
+    errs_info = (
+        params.res_dir
+        / f"wer-summary-{test_set_name}-{key}-{params.suffix}.txt"
+    )
+    with open(errs_info, "w") as f:
+        print("settings\tWER", file=f)
+        for key, val in test_set_wers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, WER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_wers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    LibriSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    assert params.decoding_method in ("greedy_search", "beam_search")
+    params.res_dir = params.exp_dir / params.decoding_method
+
+    params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
+    if params.decoding_method == "beam_search":
+        params.suffix += f"-beam-{params.beam_size}"
+
+    setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
+    logging.info("Decoding started")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <sos/eos> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.sos_id = sp.piece_to_id("<sos/eos>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    if params.avg == 1:
+        load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    else:
+        start = params.epoch - params.avg + 1
+        filenames = []
+        for i in range(start, params.epoch + 1):
+            if start >= 0:
+                filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+        logging.info(f"averaging {filenames}")
+        model.to(device)
+        model.load_state_dict(average_checkpoints(filenames, device=device))
+
+    model.to(device)
+    model.eval()
+    model.device = device
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    librispeech = LibriSpeechAsrDataModule(args)
+
+    test_clean_cuts = librispeech.test_clean_cuts()
+    test_other_cuts = librispeech.test_other_cuts()
+
+    test_clean_dl = librispeech.test_dataloaders(test_clean_cuts)
+    test_other_dl = librispeech.test_dataloaders(test_other_cuts)
+
+    test_sets = ["test-clean", "test-other"]
+    test_dl = [test_clean_dl, test_other_dl]
+
+    for test_set, test_dl in zip(test_sets, test_dl):
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+            sp=sp,
+        )
+
+        save_results(
+            params=params,
+            test_set_name=test_set,
+            results_dict=results_dict,
+        )
+
+    logging.info("Done!")
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer_lstm/decoder.py

@@ -0,0 +1,101 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Optional, Tuple
+
+import torch
+import torch.nn as nn
+
+
+# TODO(fangjun): Support switching between LSTM and GRU
+class Decoder(nn.Module):
+    def __init__(
+        self,
+        vocab_size: int,
+        embedding_dim: int,
+        blank_id: int,
+        sos_id: int,
+        num_layers: int,
+        hidden_dim: int,
+        output_dim: int,
+        embedding_dropout: float = 0.0,
+        rnn_dropout: float = 0.0,
+    ):
+        """
+        Args:
+          vocab_size:
+            Number of tokens of the modeling unit including blank.
+          embedding_dim:
+            Dimension of the input embedding.
+          blank_id:
+            The ID of the blank symbol.
+          sos_id:
+            The ID of the SOS symbol.
+          num_layers:
+            Number of LSTM layers.
+          hidden_dim:
+            Hidden dimension of LSTM layers.
+          output_dim:
+            Output dimension of the decoder.
+          embedding_dropout:
+            Dropout rate for the embedding layer.
+          rnn_dropout:
+            Dropout for LSTM layers.
+        """
+        super().__init__()
+        self.embedding = nn.Embedding(
+            num_embeddings=vocab_size,
+            embedding_dim=embedding_dim,
+            padding_idx=blank_id,
+        )
+        self.embedding_dropout = nn.Dropout(embedding_dropout)
+        # TODO(fangjun): Use layer normalized LSTM
+        self.rnn = nn.LSTM(
+            input_size=embedding_dim,
+            hidden_size=hidden_dim,
+            num_layers=num_layers,
+            batch_first=True,
+            dropout=rnn_dropout,
+        )
+        self.blank_id = blank_id
+        self.sos_id = sos_id
+        self.output_linear = nn.Linear(hidden_dim, output_dim)
+
+    def forward(
+        self,
+        y: torch.Tensor,
+        states: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+    ) -> Tuple[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]:
+        """
+        Args:
+          y:
+            A 2-D tensor of shape (N, U) with BOS prepended.
+          states:
+            A tuple of two tensors containing the states information of
+            LSTM layers in this decoder.
+        Returns:
+          Return a tuple containing:
+
+            - rnn_output, a tensor of shape (N, U, C)
+            - (h, c), containing the state information for LSTM layers.
+              Both are of shape (num_layers, N, C)
+        """
+        embeding_out = self.embedding(y)
+        embeding_out = self.embedding_dropout(embeding_out)
+        rnn_out, (h, c) = self.rnn(embeding_out, states)
+        out = self.output_linear(rnn_out)
+
+        return out, (h, c)

egs/librispeech/ASR/transducer_lstm/encoder.py

@@ -0,0 +1,115 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import Tuple
+
+import torch
+import torch.nn as nn
+from encoder_interface import EncoderInterface
+from subsampling import Conv2dSubsampling, VggSubsampling
+from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence
+
+
+class LstmEncoder(EncoderInterface):
+    def __init__(
+        self,
+        num_features: int,
+        hidden_size: int,
+        output_dim: int,
+        subsampling_factor: int = 4,
+        num_encoder_layers: int = 12,
+        dropout: float = 0.1,
+        vgg_frontend: bool = False,
+        proj_size: int = 0,
+    ):
+        super().__init__()
+        real_hidden_size = proj_size if proj_size > 0 else hidden_size
+        assert (
+            subsampling_factor == 4
+        ), "Only subsampling_factor==4 is supported at present"
+
+        # self.encoder_embed converts the input of shape (N, T, num_features)
+        # to the shape (N, T//subsampling_factor, d_model).
+        # That is, it does two things simultaneously:
+        #   (1) subsampling: T -> T//subsampling_factor
+        #   (2) embedding: num_features -> d_model
+        if vgg_frontend:
+            self.encoder_embed = VggSubsampling(num_features, real_hidden_size)
+        else:
+            self.encoder_embed = Conv2dSubsampling(
+                num_features, real_hidden_size
+            )
+
+        self.rnn = nn.LSTM(
+            input_size=hidden_size,
+            hidden_size=hidden_size,
+            num_layers=num_encoder_layers,
+            bias=True,
+            proj_size=proj_size,
+            batch_first=True,
+            dropout=dropout,
+            bidirectional=False,
+        )
+
+        self.encoder_output_layer = nn.Sequential(
+            nn.Dropout(p=dropout),
+            nn.Linear(real_hidden_size, output_dim),
+        )
+
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            The input tensor. Its shape is (batch_size, seq_len, feature_dim).
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames in
+            `x` before padding.
+        Returns:
+          Return a tuple containing 2 tensors:
+            - logits, its shape is (batch_size, output_seq_len, output_dim)
+            - logit_lens, a tensor of shape (batch_size,) containing the number
+              of frames in `logits` before padding.
+        """
+        x = self.encoder_embed(x)
+
+        # Caution: We assume the subsampling factor is 4!
+        lengths = ((x_lens - 1) // 2 - 1) // 2
+        assert x.size(1) == lengths.max().item(), (
+            x.size(1),
+            lengths.max(),
+        )
+
+        if False:
+            # It is commented out as DPP complains that not all parameters are
+            # used. Need more checks later for the reason.
+            #
+            # Caution: We assume the dataloader returns utterances with
+            # duration being sorted in decreasing order
+            packed_x = pack_padded_sequence(
+                input=x,
+                lengths=lengths.cpu(),
+                batch_first=True,
+                enforce_sorted=True,
+            )
+
+            packed_rnn_out, _ = self.rnn(packed_x)
+            rnn_out, _ = pad_packed_sequence(packed_x, batch_first=True)
+        else:
+            rnn_out, _ = self.rnn(x)
+
+        logits = self.encoder_output_layer(rnn_out)
+        return logits, lengths

egs/librispeech/ASR/transducer_lstm/encoder_interface.py

@@ -0,0 +1,43 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Tuple
+
+import torch
+import torch.nn as nn
+
+
+class EncoderInterface(nn.Module):
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            A tensor of shape (batch_size, input_seq_len, num_features)
+            containing the input features.
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames
+            in `x` before padding.
+        Returns:
+          Return a tuple containing two tensors:
+            - encoder_out, a tensor of (batch_size, out_seq_len, output_dim)
+              containing unnormalized probabilities, i.e., the output of a
+              linear layer.
+            - encoder_out_lens, a tensor of shape (batch_size,) containing
+              the number of frames in `encoder_out` before padding.
+        """
+        raise NotImplementedError("Please implement it in a subclass")

egs/librispeech/ASR/transducer_lstm/joiner.py

@@ -0,0 +1,55 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class Joiner(nn.Module):
+    def __init__(self, input_dim: int, output_dim: int):
+        super().__init__()
+
+        self.output_linear = nn.Linear(input_dim, output_dim)
+
+    def forward(
+        self, encoder_out: torch.Tensor, decoder_out: torch.Tensor
+    ) -> torch.Tensor:
+        """
+        Args:
+          encoder_out:
+            Output from the encoder. Its shape is (N, T, C).
+          decoder_out:
+            Output from the decoder. Its shape is (N, U, C).
+        Returns:
+          Return a tensor of shape (N, T, U, C).
+        """
+        assert encoder_out.ndim == decoder_out.ndim == 3
+        assert encoder_out.size(0) == decoder_out.size(0)
+        assert encoder_out.size(2) == decoder_out.size(2)
+
+        encoder_out = encoder_out.unsqueeze(2)
+        # Now encoder_out is (N, T, 1, C)
+
+        decoder_out = decoder_out.unsqueeze(1)
+        # Now decoder_out is (N, 1, U, C)
+
+        logit = encoder_out + decoder_out
+        logit = F.relu(logit)
+
+        output = self.output_linear(logit)
+
+        return output

egs/librispeech/ASR/transducer_lstm/model.py

@@ -0,0 +1,127 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Note we use `rnnt_loss` from torchaudio, which exists only in
+torchaudio >= v0.10.0. It also means you have to use torch >= v1.10.0
+"""
+import k2
+import torch
+import torch.nn as nn
+import torchaudio
+import torchaudio.functional
+from encoder_interface import EncoderInterface
+
+from icefall.utils import add_sos
+
+assert hasattr(torchaudio.functional, "rnnt_loss"), (
+    f"Current torchaudio version: {torchaudio.__version__}\n"
+    "Please install a version >= 0.10.0"
+)
+
+
+class Transducer(nn.Module):
+    """It implements https://arxiv.org/pdf/1211.3711.pdf
+    "Sequence Transduction with Recurrent Neural Networks"
+    """
+
+    def __init__(
+        self,
+        encoder: EncoderInterface,
+        decoder: nn.Module,
+        joiner: nn.Module,
+    ):
+        """
+        Args:
+          encoder:
+            It is the transcription network in the paper. Its accepts
+            two inputs: `x` of (N, T, C) and `x_lens` of shape (N,).
+            It returns two tensors: `logits` of shape (N, T, C) and
+            `logit_lens` of shape (N,).
+          decoder:
+            It is the prediction network in the paper. Its input shape
+            is (N, U) and its output shape is (N, U, C). It should contain
+            two attributes: `blank_id` and `sos_id`.
+          joiner:
+            It has two inputs with shapes: (N, T, C) and (N, U, C). Its
+            output shape is (N, T, U, C). Note that its output contains
+            unnormalized probs, i.e., not processed by log-softmax.
+        """
+        super().__init__()
+        assert isinstance(encoder, EncoderInterface)
+        assert hasattr(decoder, "blank_id")
+        assert hasattr(decoder, "sos_id")
+
+        self.encoder = encoder
+        self.decoder = decoder
+        self.joiner = joiner
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        x_lens: torch.Tensor,
+        y: k2.RaggedTensor,
+    ) -> torch.Tensor:
+        """
+        Args:
+          x:
+            A 3-D tensor of shape (N, T, C).
+          x_lens:
+            A 1-D tensor of shape (N,). It contains the number of frames in `x`
+            before padding.
+          y:
+            A ragged tensor with 2 axes [utt][label]. It contains labels of each
+            utterance.
+        Returns:
+          Return the transducer loss.
+        """
+        assert x.ndim == 3, x.shape
+        assert x_lens.ndim == 1, x_lens.shape
+        assert y.num_axes == 2, y.num_axes
+
+        assert x.size(0) == x_lens.size(0) == y.dim0
+
+        encoder_out, x_lens = self.encoder(x, x_lens)
+        assert torch.all(x_lens > 0)
+
+        # Now for the decoder, i.e., the prediction network
+        row_splits = y.shape.row_splits(1)
+        y_lens = row_splits[1:] - row_splits[:-1]
+
+        blank_id = self.decoder.blank_id
+        sos_id = self.decoder.sos_id
+        sos_y = add_sos(y, sos_id=sos_id)
+
+        sos_y_padded = sos_y.pad(mode="constant", padding_value=blank_id)
+
+        decoder_out, _ = self.decoder(sos_y_padded)
+
+        logits = self.joiner(encoder_out, decoder_out)
+
+        # rnnt_loss requires 0 padded targets
+        # Note: y does not start with SOS
+        y_padded = y.pad(mode="constant", padding_value=0)
+
+        loss = torchaudio.functional.rnnt_loss(
+            logits=logits,
+            targets=y_padded,
+            logit_lengths=x_lens,
+            target_lengths=y_lens,
+            blank=blank_id,
+            reduction="sum",
+        )
+
+        return loss

egs/librispeech/ASR/transducer_lstm/noam.py

@@ -0,0 +1,104 @@
+# Copyright    2021 University of Chinese Academy of Sciences (author: Han Zhu)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+
+
+class Noam(object):
+    """
+    Implements Noam optimizer.
+
+    Proposed in
+    "Attention Is All You Need", https://arxiv.org/pdf/1706.03762.pdf
+
+    Modified from
+    https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/optimizer.py  # noqa
+
+    Args:
+      params:
+        iterable of parameters to optimize or dicts defining parameter groups
+      model_size:
+        attention dimension of the transformer model
+      factor:
+        learning rate factor
+      warm_step:
+        warmup steps
+    """
+
+    def __init__(
+        self,
+        params,
+        model_size: int = 256,
+        factor: float = 10.0,
+        warm_step: int = 25000,
+        weight_decay=0,
+    ) -> None:
+        """Construct an Noam object."""
+        self.optimizer = torch.optim.Adam(
+            params, lr=0, betas=(0.9, 0.98), eps=1e-9, weight_decay=weight_decay
+        )
+        self._step = 0
+        self.warmup = warm_step
+        self.factor = factor
+        self.model_size = model_size
+        self._rate = 0
+
+    @property
+    def param_groups(self):
+        """Return param_groups."""
+        return self.optimizer.param_groups
+
+    def step(self):
+        """Update parameters and rate."""
+        self._step += 1
+        rate = self.rate()
+        for p in self.optimizer.param_groups:
+            p["lr"] = rate
+        self._rate = rate
+        self.optimizer.step()
+
+    def rate(self, step=None):
+        """Implement `lrate` above."""
+        if step is None:
+            step = self._step
+        return (
+            self.factor
+            * self.model_size ** (-0.5)
+            * min(step ** (-0.5), step * self.warmup ** (-1.5))
+        )
+
+    def zero_grad(self):
+        """Reset gradient."""
+        self.optimizer.zero_grad()
+
+    def state_dict(self):
+        """Return state_dict."""
+        return {
+            "_step": self._step,
+            "warmup": self.warmup,
+            "factor": self.factor,
+            "model_size": self.model_size,
+            "_rate": self._rate,
+            "optimizer": self.optimizer.state_dict(),
+        }
+
+    def load_state_dict(self, state_dict):
+        """Load state_dict."""
+        for key, value in state_dict.items():
+            if key == "optimizer":
+                self.optimizer.load_state_dict(state_dict["optimizer"])
+            else:
+                setattr(self, key, value)

egs/librispeech/ASR/transducer_lstm/subsampling.py

@@ -0,0 +1 @@
+../transducer/subsampling.py

egs/librispeech/ASR/transducer_lstm/test_encoder.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env python3
+#
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+To run this file, do:
+
+    cd icefall/egs/librispeech/ASR
+    python ./transducer_lstm/test_encoder.py
+"""
+
+from encoder import LstmEncoder
+
+
+def test_encoder():
+    encoder = LstmEncoder(
+        num_features=80,
+        hidden_size=1024,
+        proj_size=512,
+        output_dim=512,
+        subsampling_factor=4,
+        num_encoder_layers=12,
+    )
+    num_params = sum(p.numel() for p in encoder.parameters() if p.requires_grad)
+    print(num_params)
+    # 93979284
+    # 66427392
+
+
+def main():
+    test_encoder()
+
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer_lstm/train.py

@@ -0,0 +1,738 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang,
+#                                                  Wei Kang
+#                                                  Mingshuang Luo)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Usage:
+
+export CUDA_VISIBLE_DEVICES="0,1,2"
+
+./transducer_lstm/train.py \
+  --world-size 3 \
+  --num-epochs 30 \
+  --start-epoch 0 \
+  --exp-dir transducer_lstm/exp \
+  --full-libri 1 \
+  --max-duration 400 \
+  --lr-factor 3
+"""
+
+
+import argparse
+import logging
+from pathlib import Path
+from shutil import copyfile
+from typing import Optional, Tuple
+
+import k2
+import sentencepiece as spm
+import torch
+import torch.multiprocessing as mp
+import torch.nn as nn
+from asr_datamodule import LibriSpeechAsrDataModule
+from decoder import Decoder
+from encoder import LstmEncoder
+from joiner import Joiner
+from lhotse.cut import Cut
+from lhotse.utils import fix_random_seed
+from model import Transducer
+from noam import Noam
+from torch import Tensor
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.nn.utils import clip_grad_norm_
+from torch.utils.tensorboard import SummaryWriter
+
+from icefall.checkpoint import load_checkpoint
+from icefall.checkpoint import save_checkpoint as save_checkpoint_impl
+from icefall.dist import cleanup_dist, setup_dist
+from icefall.env import get_env_info
+from icefall.utils import AttributeDict, MetricsTracker, setup_logger, str2bool
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--world-size",
+        type=int,
+        default=1,
+        help="Number of GPUs for DDP training.",
+    )
+
+    parser.add_argument(
+        "--master-port",
+        type=int,
+        default=12354,
+        help="Master port to use for DDP training.",
+    )
+
+    parser.add_argument(
+        "--tensorboard",
+        type=str2bool,
+        default=True,
+        help="Should various information be logged in tensorboard.",
+    )
+
+    parser.add_argument(
+        "--num-epochs",
+        type=int,
+        default=30,
+        help="Number of epochs to train.",
+    )
+
+    parser.add_argument(
+        "--start-epoch",
+        type=int,
+        default=0,
+        help="""Resume training from from this epoch.
+        If it is positive, it will load checkpoint from
+        transducer_lstm/exp/epoch-{start_epoch-1}.pt
+        """,
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="transducer_lstm/exp",
+        help="""The experiment dir.
+        It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--lr-factor",
+        type=float,
+        default=3.0,
+        help="The lr_factor for Noam optimizer",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    """Return a dict containing training parameters.
+
+    All training related parameters that are not passed from the commandline
+    are saved in the variable `params`.
+
+    Commandline options are merged into `params` after they are parsed, so
+    you can also access them via `params`.
+
+    Explanation of options saved in `params`:
+
+        - best_train_loss: Best training loss so far. It is used to select
+                           the model that has the lowest training loss. It is
+                           updated during the training.
+
+        - best_valid_loss: Best validation loss so far. It is used to select
+                           the model that has the lowest validation loss. It is
+                           updated during the training.
+
+        - best_train_epoch: It is the epoch that has the best training loss.
+
+        - best_valid_epoch: It is the epoch that has the best validation loss.
+
+        - batch_idx_train: Used to writing statistics to tensorboard. It
+                           contains number of batches trained so far across
+                           epochs.
+
+        - log_interval:  Print training loss if batch_idx % log_interval` is 0
+
+        - reset_interval: Reset statistics if batch_idx % reset_interval is 0
+
+        - valid_interval:  Run validation if batch_idx % valid_interval is 0
+
+        - feature_dim: The model input dim. It has to match the one used
+                       in computing features.
+
+        - subsampling_factor:  The subsampling factor for the model.
+
+        - use_feat_batchnorm: Whether to do batch normalization for the
+                              input features.
+
+        - attention_dim: Hidden dim for multi-head attention model.
+
+        - num_decoder_layers: Number of decoder layer of transformer decoder.
+
+        - weight_decay:  The weight_decay for the optimizer.
+
+        - warm_step: The warm_step for Noam optimizer.
+    """
+    params = AttributeDict(
+        {
+            "best_train_loss": float("inf"),
+            "best_valid_loss": float("inf"),
+            "best_train_epoch": -1,
+            "best_valid_epoch": -1,
+            "batch_idx_train": 0,
+            "log_interval": 50,
+            "reset_interval": 200,
+            "valid_interval": 3000,  # For the 100h subset, use 800
+            # parameters for conformer
+            "feature_dim": 80,
+            "encoder_out_dim": 512,
+            "subsampling_factor": 4,
+            "encoder_hidden_size": 1024,
+            "num_encoder_layers": 4,
+            "proj_size": 512,
+            "vgg_frontend": False,
+            # decoder params
+            "decoder_embedding_dim": 1024,
+            "num_decoder_layers": 4,
+            "decoder_hidden_dim": 512,
+            # parameters for Noam
+            "weight_decay": 1e-6,
+            "warm_step": 80000,  # For the 100h subset, use 8k
+            "env_info": get_env_info(),
+        }
+    )
+
+    return params
+
+
+def get_encoder_model(params: AttributeDict):
+    encoder = LstmEncoder(
+        num_features=params.feature_dim,
+        hidden_size=params.encoder_hidden_size,
+        output_dim=params.encoder_out_dim,
+        subsampling_factor=params.subsampling_factor,
+        num_encoder_layers=params.num_encoder_layers,
+        vgg_frontend=params.vgg_frontend,
+    )
+    return encoder
+
+
+def get_decoder_model(params: AttributeDict):
+    decoder = Decoder(
+        vocab_size=params.vocab_size,
+        embedding_dim=params.decoder_embedding_dim,
+        blank_id=params.blank_id,
+        sos_id=params.sos_id,
+        num_layers=params.num_decoder_layers,
+        hidden_dim=params.decoder_hidden_dim,
+        output_dim=params.encoder_out_dim,
+    )
+    return decoder
+
+
+def get_joiner_model(params: AttributeDict):
+    joiner = Joiner(
+        input_dim=params.encoder_out_dim,
+        output_dim=params.vocab_size,
+    )
+    return joiner
+
+
+def get_transducer_model(params: AttributeDict):
+    encoder = get_encoder_model(params)
+    decoder = get_decoder_model(params)
+    joiner = get_joiner_model(params)
+
+    model = Transducer(
+        encoder=encoder,
+        decoder=decoder,
+        joiner=joiner,
+    )
+    return model
+
+
+def load_checkpoint_if_available(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+) -> None:
+    """Load checkpoint from file.
+
+    If params.start_epoch is positive, it will load the checkpoint from
+    `params.start_epoch - 1`. Otherwise, this function does nothing.
+
+    Apart from loading state dict for `model`, `optimizer` and `scheduler`,
+    it also updates `best_train_epoch`, `best_train_loss`, `best_valid_epoch`,
+    and `best_valid_loss` in `params`.
+
+    Args:
+      params:
+        The return value of :func:`get_params`.
+      model:
+        The training model.
+      optimizer:
+        The optimizer that we are using.
+      scheduler:
+        The learning rate scheduler we are using.
+    Returns:
+      Return None.
+    """
+    if params.start_epoch <= 0:
+        return
+
+    filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
+    saved_params = load_checkpoint(
+        filename,
+        model=model,
+        optimizer=optimizer,
+        scheduler=scheduler,
+    )
+
+    keys = [
+        "best_train_epoch",
+        "best_valid_epoch",
+        "batch_idx_train",
+        "best_train_loss",
+        "best_valid_loss",
+    ]
+    for k in keys:
+        params[k] = saved_params[k]
+
+    return saved_params
+
+
+def save_checkpoint(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+    rank: int = 0,
+) -> None:
+    """Save model, optimizer, scheduler and training stats to file.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The training model.
+    """
+    if rank != 0:
+        return
+    filename = params.exp_dir / f"epoch-{params.cur_epoch}.pt"
+    save_checkpoint_impl(
+        filename=filename,
+        model=model,
+        params=params,
+        optimizer=optimizer,
+        scheduler=scheduler,
+        rank=rank,
+    )
+
+    if params.best_train_epoch == params.cur_epoch:
+        best_train_filename = params.exp_dir / "best-train-loss.pt"
+        copyfile(src=filename, dst=best_train_filename)
+
+    if params.best_valid_epoch == params.cur_epoch:
+        best_valid_filename = params.exp_dir / "best-valid-loss.pt"
+        copyfile(src=filename, dst=best_valid_filename)
+
+
+def compute_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    batch: dict,
+    is_training: bool,
+) -> Tuple[Tensor, MetricsTracker]:
+    """
+    Compute CTC loss given the model and its inputs.
+
+    Args:
+      params:
+        Parameters for training. See :func:`get_params`.
+      model:
+        The model for training. It is an instance of Conformer in our case.
+      batch:
+        A batch of data. See `lhotse.dataset.K2SpeechRecognitionDataset()`
+        for the content in it.
+      is_training:
+        True for training. False for validation. When it is True, this
+        function enables autograd during computation; when it is False, it
+        disables autograd.
+    """
+    device = model.device
+    feature = batch["inputs"]
+    # at entry, feature is (N, T, C)
+    assert feature.ndim == 3
+    feature = feature.to(device)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    texts = batch["supervisions"]["text"]
+    y = sp.encode(texts, out_type=int)
+    y = k2.RaggedTensor(y).to(device)
+
+    with torch.set_grad_enabled(is_training):
+        loss = model(x=feature, x_lens=feature_lens, y=y)
+
+    assert loss.requires_grad == is_training
+
+    info = MetricsTracker()
+    info["frames"] = (feature_lens // params.subsampling_factor).sum().item()
+
+    # Note: We use reduction=sum while computing the loss.
+    info["loss"] = loss.detach().cpu().item()
+
+    return loss, info
+
+
+def compute_validation_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    valid_dl: torch.utils.data.DataLoader,
+    world_size: int = 1,
+) -> MetricsTracker:
+    """Run the validation process."""
+    model.eval()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(valid_dl):
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+            is_training=False,
+        )
+        assert loss.requires_grad is False
+        tot_loss = tot_loss + loss_info
+
+    if world_size > 1:
+        tot_loss.reduce(loss.device)
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    if loss_value < params.best_valid_loss:
+        params.best_valid_epoch = params.cur_epoch
+        params.best_valid_loss = loss_value
+
+    return tot_loss
+
+
+def train_one_epoch(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: torch.optim.Optimizer,
+    sp: spm.SentencePieceProcessor,
+    train_dl: torch.utils.data.DataLoader,
+    valid_dl: torch.utils.data.DataLoader,
+    tb_writer: Optional[SummaryWriter] = None,
+    world_size: int = 1,
+) -> None:
+    """Train the model for one epoch.
+
+    The training loss from the mean of all frames is saved in
+    `params.train_loss`. It runs the validation process every
+    `params.valid_interval` batches.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The model for training.
+      optimizer:
+        The optimizer we are using.
+      train_dl:
+        Dataloader for the training dataset.
+      valid_dl:
+        Dataloader for the validation dataset.
+      tb_writer:
+        Writer to write log messages to tensorboard.
+      world_size:
+        Number of nodes in DDP training. If it is 1, DDP is disabled.
+    """
+    model.train()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(train_dl):
+        params.batch_idx_train += 1
+        batch_size = len(batch["supervisions"]["text"])
+
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+            is_training=True,
+        )
+        # summary stats
+        tot_loss = (tot_loss * (1 - 1 / params.reset_interval)) + loss_info
+
+        # NOTE: We use reduction==sum and loss is computed over utterances
+        # in the batch and there is no normalization to it so far.
+
+        optimizer.zero_grad()
+        loss.backward()
+        clip_grad_norm_(model.parameters(), 5.0, 2.0)
+        optimizer.step()
+
+        if batch_idx % params.log_interval == 0:
+            logging.info(
+                f"Epoch {params.cur_epoch}, "
+                f"batch {batch_idx}, loss[{loss_info}], "
+                f"tot_loss[{tot_loss}], batch size: {batch_size}"
+            )
+
+        if batch_idx % params.log_interval == 0:
+
+            if tb_writer is not None:
+                loss_info.write_summary(
+                    tb_writer, "train/current_", params.batch_idx_train
+                )
+                tot_loss.write_summary(
+                    tb_writer, "train/tot_", params.batch_idx_train
+                )
+
+        if batch_idx > 0 and batch_idx % params.valid_interval == 0:
+            logging.info("Computing validation loss")
+            valid_info = compute_validation_loss(
+                params=params,
+                model=model,
+                sp=sp,
+                valid_dl=valid_dl,
+                world_size=world_size,
+            )
+            model.train()
+            logging.info(f"Epoch {params.cur_epoch}, validation: {valid_info}")
+            if tb_writer is not None:
+                valid_info.write_summary(
+                    tb_writer, "train/valid_", params.batch_idx_train
+                )
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    params.train_loss = loss_value
+    if params.train_loss < params.best_train_loss:
+        params.best_train_epoch = params.cur_epoch
+        params.best_train_loss = params.train_loss
+
+
+def run(rank, world_size, args):
+    """
+    Args:
+      rank:
+        It is a value between 0 and `world_size-1`, which is
+        passed automatically by `mp.spawn()` in :func:`main`.
+        The node with rank 0 is responsible for saving checkpoint.
+      world_size:
+        Number of GPUs for DDP training.
+      args:
+        The return value of get_parser().parse_args()
+    """
+    params = get_params()
+    params.update(vars(args))
+    if params.full_libri is False:
+        params.valid_interval = 800
+        params.warm_step = 8000
+
+    fix_random_seed(42)
+    if world_size > 1:
+        setup_dist(rank, world_size, params.master_port)
+
+    setup_logger(f"{params.exp_dir}/log/log-train")
+    logging.info("Training started")
+
+    if args.tensorboard and rank == 0:
+        tb_writer = SummaryWriter(log_dir=f"{params.exp_dir}/tensorboard")
+    else:
+        tb_writer = None
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", rank)
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <sos/eos> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.sos_id = sp.piece_to_id("<sos/eos>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    checkpoints = load_checkpoint_if_available(params=params, model=model)
+
+    num_param = sum([p.numel() for p in model.parameters() if p.requires_grad])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    model.to(device)
+    if world_size > 1:
+        logging.info("Using DDP")
+        model = DDP(model, device_ids=[rank])
+    model.device = device
+
+    optimizer = Noam(
+        model.parameters(),
+        model_size=params.encoder_hidden_size,
+        factor=params.lr_factor,
+        warm_step=params.warm_step,
+        weight_decay=params.weight_decay,
+    )
+
+    if checkpoints and "optimizer" in checkpoints:
+        logging.info("Loading optimizer state dict")
+        optimizer.load_state_dict(checkpoints["optimizer"])
+
+    librispeech = LibriSpeechAsrDataModule(args)
+
+    train_cuts = librispeech.train_clean_100_cuts()
+    if params.full_libri:
+        train_cuts += librispeech.train_clean_360_cuts()
+        train_cuts += librispeech.train_other_500_cuts()
+
+    def remove_short_and_long_utt(c: Cut):
+        # Keep only utterances with duration between 1 second and 20 seconds
+        return 1.0 <= c.duration <= 20.0
+
+    num_in_total = len(train_cuts)
+
+    train_cuts = train_cuts.filter(remove_short_and_long_utt)
+
+    num_left = len(train_cuts)
+    num_removed = num_in_total - num_left
+    removed_percent = num_removed / num_in_total * 100
+
+    logging.info(f"Before removing short and long utterances: {num_in_total}")
+    logging.info(f"After removing short and long utterances: {num_left}")
+    logging.info(f"Removed {num_removed} utterances ({removed_percent:.5f}%)")
+
+    train_dl = librispeech.train_dataloaders(train_cuts)
+
+    valid_cuts = librispeech.dev_clean_cuts()
+    valid_cuts += librispeech.dev_other_cuts()
+    valid_dl = librispeech.valid_dataloaders(valid_cuts)
+
+    scan_pessimistic_batches_for_oom(
+        model=model,
+        train_dl=train_dl,
+        optimizer=optimizer,
+        sp=sp,
+        params=params,
+    )
+
+    for epoch in range(params.start_epoch, params.num_epochs):
+        train_dl.sampler.set_epoch(epoch)
+
+        cur_lr = optimizer._rate
+        if tb_writer is not None:
+            tb_writer.add_scalar(
+                "train/learning_rate", cur_lr, params.batch_idx_train
+            )
+            tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)
+
+        if rank == 0:
+            logging.info("epoch {}, learning rate {}".format(epoch, cur_lr))
+
+        params.cur_epoch = epoch
+
+        train_one_epoch(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            sp=sp,
+            train_dl=train_dl,
+            valid_dl=valid_dl,
+            tb_writer=tb_writer,
+            world_size=world_size,
+        )
+
+        save_checkpoint(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            rank=rank,
+        )
+
+    logging.info("Done!")
+
+    if world_size > 1:
+        torch.distributed.barrier()
+        cleanup_dist()
+
+
+def scan_pessimistic_batches_for_oom(
+    model: nn.Module,
+    train_dl: torch.utils.data.DataLoader,
+    optimizer: torch.optim.Optimizer,
+    sp: spm.SentencePieceProcessor,
+    params: AttributeDict,
+):
+    from lhotse.dataset import find_pessimistic_batches
+
+    logging.info(
+        "Sanity check -- see if any of the batches in epoch 0 would cause OOM."
+    )
+    batches, crit_values = find_pessimistic_batches(train_dl.sampler)
+    for criterion, cuts in batches.items():
+        batch = train_dl.dataset[cuts]
+        try:
+            optimizer.zero_grad()
+            loss, _ = compute_loss(
+                params=params,
+                model=model,
+                sp=sp,
+                batch=batch,
+                is_training=True,
+            )
+            loss.backward()
+            clip_grad_norm_(model.parameters(), 5.0, 2.0)
+            optimizer.step()
+        except RuntimeError as e:
+            if "CUDA out of memory" in str(e):
+                logging.error(
+                    "Your GPU ran out of memory with the current "
+                    "max_duration setting. We recommend decreasing "
+                    "max_duration and trying again.\n"
+                    f"Failing criterion: {criterion} "
+                    f"(={crit_values[criterion]}) ..."
+                )
+            raise
+
+
+def main():
+    parser = get_parser()
+    LibriSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    world_size = args.world_size
+    assert world_size >= 1
+    if world_size > 1:
+        mp.spawn(run, args=(world_size, args), nprocs=world_size, join=True)
+    else:
+        run(rank=0, world_size=1, args=args)
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer_stateless/README.md

@@ -0,0 +1,22 @@
+## Introduction
+
+The decoder, i.e., the prediction network, is from
+https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9054419
+(Rnn-Transducer with Stateless Prediction Network)
+
+You can use the following command to start the training:
+
+```bash
+cd egs/librispeech/ASR
+
+export CUDA_VISIBLE_DEVICES="0,1,2,3"
+
+./transducer_stateless/train.py \
+  --world-size 4 \
+  --num-epochs 30 \
+  --start-epoch 0 \
+  --exp-dir transducer_stateless/exp \
+  --full-libri 1 \
+  --max-duration 250 \
+  --lr-factor 2.5
+```

egs/librispeech/ASR/transducer_stateless/asr_datamodule.py

@@ -0,0 +1 @@
+../transducer/asr_datamodule.py

egs/librispeech/ASR/transducer_stateless/beam_search.py

@@ -0,0 +1,212 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from dataclasses import dataclass
+from typing import Dict, List, Optional, Tuple
+
+import torch
+from model import Transducer
+
+
+def greedy_search(model: Transducer, encoder_out: torch.Tensor) -> List[int]:
+    """
+    Args:
+      model:
+        An instance of `Transducer`.
+      encoder_out:
+        A tensor of shape (N, T, C) from the encoder. Support only N==1 for now.
+    Returns:
+      Return the decoded result.
+    """
+    assert encoder_out.ndim == 3
+
+    # support only batch_size == 1 for now
+    assert encoder_out.size(0) == 1, encoder_out.size(0)
+    blank_id = model.decoder.blank_id
+    device = model.device
+
+    sos = torch.tensor([blank_id], device=device).reshape(1, 1)
+    decoder_out, (h, c) = model.decoder(sos)
+    T = encoder_out.size(1)
+    t = 0
+    hyp = []
+    max_u = 1000  # terminate after this number of steps
+    u = 0
+
+    while t < T and u < max_u:
+        # fmt: off
+        current_encoder_out = encoder_out[:, t:t+1, :]
+        # fmt: on
+        logits = model.joiner(current_encoder_out, decoder_out)
+        # logits is (1, 1, 1, vocab_size)
+
+        log_prob = logits.log_softmax(dim=-1)
+        # log_prob is (1, 1, 1, vocab_size)
+        # TODO: Use logits.argmax()
+        y = log_prob.argmax()
+        if y != blank_id:
+            hyp.append(y.item())
+            y = y.reshape(1, 1)
+            decoder_out, (h, c) = model.decoder(y, (h, c))
+            u += 1
+        else:
+            t += 1
+
+    return hyp
+
+
+@dataclass
+class Hypothesis:
+    ys: List[int]  # the predicted sequences so far
+    log_prob: float  # The log prob of ys
+
+    # Optional decoder state. We assume it is LSTM for now,
+    # so the state is a tuple (h, c)
+    decoder_state: Optional[Tuple[torch.Tensor, torch.Tensor]] = None
+
+
+def beam_search(
+    model: Transducer,
+    encoder_out: torch.Tensor,
+    beam: int = 5,
+) -> List[int]:
+    """
+    It implements Algorithm 1 in https://arxiv.org/pdf/1211.3711.pdf
+
+    espnet/nets/beam_search_transducer.py#L247 is used as a reference.
+
+    Args:
+      model:
+        An instance of `Transducer`.
+      encoder_out:
+        A tensor of shape (N, T, C) from the encoder. Support only N==1 for now.
+      beam:
+        Beam size.
+    Returns:
+      Return the decoded result.
+    """
+    assert encoder_out.ndim == 3
+
+    # support only batch_size == 1 for now
+    assert encoder_out.size(0) == 1, encoder_out.size(0)
+    blank_id = model.decoder.blank_id
+    sos_id = model.decoder.sos_id
+    device = model.device
+
+    sos = torch.tensor([blank_id], device=device).reshape(1, 1)
+    decoder_out, (h, c) = model.decoder(sos)
+    T = encoder_out.size(1)
+    t = 0
+    B = [Hypothesis(ys=[blank_id], log_prob=0.0, decoder_state=None)]
+    max_u = 20000  # terminate after this number of steps
+    u = 0
+
+    cache: Dict[
+        str, Tuple[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]
+    ] = {}
+
+    while t < T and u < max_u:
+        # fmt: off
+        current_encoder_out = encoder_out[:, t:t+1, :]
+        # fmt: on
+        A = B
+        B = []
+        #  for hyp in A:
+        #      for h in A:
+        #          if h.ys == hyp.ys[:-1]:
+        #              # update the score of hyp
+        #              decoder_input = torch.tensor(
+        #                  [h.ys[-1]], device=device
+        #              ).reshape(1, 1)
+        #              decoder_out, _ = model.decoder(
+        #                  decoder_input, h.decoder_state
+        #              )
+        #              logits = model.joiner(current_encoder_out, decoder_out)
+        #              log_prob = logits.log_softmax(dim=-1)
+        #              log_prob = log_prob.squeeze()
+        #              hyp.log_prob += h.log_prob + log_prob[hyp.ys[-1]].item()
+
+        while u < max_u:
+            y_star = max(A, key=lambda hyp: hyp.log_prob)
+            A.remove(y_star)
+
+            # Note: y_star.ys is unhashable, i.e., cannot be used
+            # as a key into a dict
+            cached_key = "_".join(map(str, y_star.ys))
+
+            if cached_key not in cache:
+                decoder_input = torch.tensor(
+                    [y_star.ys[-1]], device=device
+                ).reshape(1, 1)
+
+                decoder_out, decoder_state = model.decoder(
+                    decoder_input,
+                    y_star.decoder_state,
+                )
+                cache[cached_key] = (decoder_out, decoder_state)
+            else:
+                decoder_out, decoder_state = cache[cached_key]
+
+            logits = model.joiner(current_encoder_out, decoder_out)
+            log_prob = logits.log_softmax(dim=-1)
+            # log_prob is (1, 1, 1, vocab_size)
+            log_prob = log_prob.squeeze()
+            # Now log_prob is (vocab_size,)
+
+            # If we choose blank here, add the new hypothesis to B.
+            # Otherwise, add the new hypothesis to A
+
+            # First, choose blank
+            skip_log_prob = log_prob[blank_id]
+            new_y_star_log_prob = y_star.log_prob + skip_log_prob.item()
+
+            # ys[:] returns a copy of ys
+            new_y_star = Hypothesis(
+                ys=y_star.ys[:],
+                log_prob=new_y_star_log_prob,
+                # Caution: Use y_star.decoder_state here
+                decoder_state=y_star.decoder_state,
+            )
+            B.append(new_y_star)
+
+            # Second, choose other labels
+            for i, v in enumerate(log_prob.tolist()):
+                if i in (blank_id, sos_id):
+                    continue
+                new_ys = y_star.ys + [i]
+                new_log_prob = y_star.log_prob + v
+                new_hyp = Hypothesis(
+                    ys=new_ys,
+                    log_prob=new_log_prob,
+                    decoder_state=decoder_state,
+                )
+                A.append(new_hyp)
+            u += 1
+            # check whether B contains more than "beam" elements more probable
+            # than the most probable in A
+            A_most_probable = max(A, key=lambda hyp: hyp.log_prob)
+            B = sorted(
+                [hyp for hyp in B if hyp.log_prob > A_most_probable.log_prob],
+                key=lambda hyp: hyp.log_prob,
+                reverse=True,
+            )
+            if len(B) >= beam:
+                B = B[:beam]
+                break
+        t += 1
+    best_hyp = max(B, key=lambda hyp: hyp.log_prob / len(hyp.ys[1:]))
+    ys = best_hyp.ys[1:]  # [1:] to remove the blank
+    return ys

egs/librispeech/ASR/transducer_stateless/conformer.py

@@ -0,0 +1,922 @@
+#!/usr/bin/env python3
+# Copyright (c)  2021  University of Chinese Academy of Sciences (author: Han Zhu)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import math
+import warnings
+from typing import Optional, Tuple
+
+import torch
+from torch import Tensor, nn
+from transducer.transformer import Transformer
+
+from icefall.utils import make_pad_mask
+
+
+class Conformer(Transformer):
+    """
+    Args:
+        num_features (int): Number of input features
+        output_dim (int): Number of output dimension
+        subsampling_factor (int): subsampling factor of encoder (the convolution layers before transformers)
+        d_model (int): attention dimension
+        nhead (int): number of head
+        dim_feedforward (int): feedforward dimention
+        num_encoder_layers (int): number of encoder layers
+        dropout (float): dropout rate
+        cnn_module_kernel (int): Kernel size of convolution module
+        normalize_before (bool): whether to use layer_norm before the first block.
+        vgg_frontend (bool): whether to use vgg frontend.
+    """
+
+    def __init__(
+        self,
+        num_features: int,
+        output_dim: int,
+        subsampling_factor: int = 4,
+        d_model: int = 256,
+        nhead: int = 4,
+        dim_feedforward: int = 2048,
+        num_encoder_layers: int = 12,
+        dropout: float = 0.1,
+        cnn_module_kernel: int = 31,
+        normalize_before: bool = True,
+        vgg_frontend: bool = False,
+        use_feat_batchnorm: bool = False,
+    ) -> None:
+        super(Conformer, self).__init__(
+            num_features=num_features,
+            output_dim=output_dim,
+            subsampling_factor=subsampling_factor,
+            d_model=d_model,
+            nhead=nhead,
+            dim_feedforward=dim_feedforward,
+            num_encoder_layers=num_encoder_layers,
+            dropout=dropout,
+            normalize_before=normalize_before,
+            vgg_frontend=vgg_frontend,
+            use_feat_batchnorm=use_feat_batchnorm,
+        )
+
+        self.encoder_pos = RelPositionalEncoding(d_model, dropout)
+
+        encoder_layer = ConformerEncoderLayer(
+            d_model,
+            nhead,
+            dim_feedforward,
+            dropout,
+            cnn_module_kernel,
+            normalize_before,
+        )
+        self.encoder = ConformerEncoder(encoder_layer, num_encoder_layers)
+        self.normalize_before = normalize_before
+        if self.normalize_before:
+            self.after_norm = nn.LayerNorm(d_model)
+        else:
+            # Note: TorchScript detects that self.after_norm could be used inside forward()
+            #       and throws an error without this change.
+            self.after_norm = identity
+
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            The input tensor. Its shape is (batch_size, seq_len, feature_dim).
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames in
+            `x` before padding.
+        Returns:
+          Return a tuple containing 2 tensors:
+            - logits, its shape is (batch_size, output_seq_len, output_dim)
+            - logit_lens, a tensor of shape (batch_size,) containing the number
+              of frames in `logits` before padding.
+        """
+        if self.use_feat_batchnorm:
+            x = x.permute(0, 2, 1)  # (N, T, C) -> (N, C, T)
+            x = self.feat_batchnorm(x)
+            x = x.permute(0, 2, 1)  # (N, C, T) -> (N, T, C)
+
+        x = self.encoder_embed(x)
+        x, pos_emb = self.encoder_pos(x)
+        x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+
+        # Caution: We assume the subsampling factor is 4!
+        lengths = ((x_lens - 1) // 2 - 1) // 2
+        assert x.size(0) == lengths.max().item()
+        mask = make_pad_mask(lengths)
+
+        x = self.encoder(x, pos_emb, src_key_padding_mask=mask)  # (T, N, C)
+
+        if self.normalize_before:
+            x = self.after_norm(x)
+
+        logits = self.encoder_output_layer(x)
+        logits = logits.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+
+        return logits, lengths
+
+
+class ConformerEncoderLayer(nn.Module):
+    """
+    ConformerEncoderLayer is made up of self-attn, feedforward and convolution networks.
+    See: "Conformer: Convolution-augmented Transformer for Speech Recognition"
+
+    Args:
+        d_model: the number of expected features in the input (required).
+        nhead: the number of heads in the multiheadattention models (required).
+        dim_feedforward: the dimension of the feedforward network model (default=2048).
+        dropout: the dropout value (default=0.1).
+        cnn_module_kernel (int): Kernel size of convolution module.
+        normalize_before: whether to use layer_norm before the first block.
+
+    Examples::
+        >>> encoder_layer = ConformerEncoderLayer(d_model=512, nhead=8)
+        >>> src = torch.rand(10, 32, 512)
+        >>> pos_emb = torch.rand(32, 19, 512)
+        >>> out = encoder_layer(src, pos_emb)
+    """
+
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        cnn_module_kernel: int = 31,
+        normalize_before: bool = True,
+    ) -> None:
+        super(ConformerEncoderLayer, self).__init__()
+        self.self_attn = RelPositionMultiheadAttention(
+            d_model, nhead, dropout=0.0
+        )
+
+        self.feed_forward = nn.Sequential(
+            nn.Linear(d_model, dim_feedforward),
+            Swish(),
+            nn.Dropout(dropout),
+            nn.Linear(dim_feedforward, d_model),
+        )
+
+        self.feed_forward_macaron = nn.Sequential(
+            nn.Linear(d_model, dim_feedforward),
+            Swish(),
+            nn.Dropout(dropout),
+            nn.Linear(dim_feedforward, d_model),
+        )
+
+        self.conv_module = ConvolutionModule(d_model, cnn_module_kernel)
+
+        self.norm_ff_macaron = nn.LayerNorm(
+            d_model
+        )  # for the macaron style FNN module
+        self.norm_ff = nn.LayerNorm(d_model)  # for the FNN module
+        self.norm_mha = nn.LayerNorm(d_model)  # for the MHA module
+
+        self.ff_scale = 0.5
+
+        self.norm_conv = nn.LayerNorm(d_model)  # for the CNN module
+        self.norm_final = nn.LayerNorm(
+            d_model
+        )  # for the final output of the block
+
+        self.dropout = nn.Dropout(dropout)
+
+        self.normalize_before = normalize_before
+
+    def forward(
+        self,
+        src: Tensor,
+        pos_emb: Tensor,
+        src_mask: Optional[Tensor] = None,
+        src_key_padding_mask: Optional[Tensor] = None,
+    ) -> Tensor:
+        """
+        Pass the input through the encoder layer.
+
+        Args:
+            src: the sequence to the encoder layer (required).
+            pos_emb: Positional embedding tensor (required).
+            src_mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+
+        Shape:
+            src: (S, N, E).
+            pos_emb: (N, 2*S-1, E)
+            src_mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, N is the batch size, E is the feature number
+        """
+
+        # macaron style feed forward module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_ff_macaron(src)
+        src = residual + self.ff_scale * self.dropout(
+            self.feed_forward_macaron(src)
+        )
+        if not self.normalize_before:
+            src = self.norm_ff_macaron(src)
+
+        # multi-headed self-attention module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_mha(src)
+        src_att = self.self_attn(
+            src,
+            src,
+            src,
+            pos_emb=pos_emb,
+            attn_mask=src_mask,
+            key_padding_mask=src_key_padding_mask,
+        )[0]
+        src = residual + self.dropout(src_att)
+        if not self.normalize_before:
+            src = self.norm_mha(src)
+
+        # convolution module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_conv(src)
+        src = residual + self.dropout(self.conv_module(src))
+        if not self.normalize_before:
+            src = self.norm_conv(src)
+
+        # feed forward module
+        residual = src
+        if self.normalize_before:
+            src = self.norm_ff(src)
+        src = residual + self.ff_scale * self.dropout(self.feed_forward(src))
+        if not self.normalize_before:
+            src = self.norm_ff(src)
+
+        if self.normalize_before:
+            src = self.norm_final(src)
+
+        return src
+
+
+class ConformerEncoder(nn.TransformerEncoder):
+    r"""ConformerEncoder is a stack of N encoder layers
+
+    Args:
+        encoder_layer: an instance of the ConformerEncoderLayer() class (required).
+        num_layers: the number of sub-encoder-layers in the encoder (required).
+        norm: the layer normalization component (optional).
+
+    Examples::
+        >>> encoder_layer = ConformerEncoderLayer(d_model=512, nhead=8)
+        >>> conformer_encoder = ConformerEncoder(encoder_layer, num_layers=6)
+        >>> src = torch.rand(10, 32, 512)
+        >>> pos_emb = torch.rand(32, 19, 512)
+        >>> out = conformer_encoder(src, pos_emb)
+    """
+
+    def __init__(
+        self, encoder_layer: nn.Module, num_layers: int, norm: nn.Module = None
+    ) -> None:
+        super(ConformerEncoder, self).__init__(
+            encoder_layer=encoder_layer, num_layers=num_layers, norm=norm
+        )
+
+    def forward(
+        self,
+        src: Tensor,
+        pos_emb: Tensor,
+        mask: Optional[Tensor] = None,
+        src_key_padding_mask: Optional[Tensor] = None,
+    ) -> Tensor:
+        r"""Pass the input through the encoder layers in turn.
+
+        Args:
+            src: the sequence to the encoder (required).
+            pos_emb: Positional embedding tensor (required).
+            mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+
+        Shape:
+            src: (S, N, E).
+            pos_emb: (N, 2*S-1, E)
+            mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, T is the target sequence length, N is the batch size, E is the feature number
+
+        """
+        output = src
+
+        for mod in self.layers:
+            output = mod(
+                output,
+                pos_emb,
+                src_mask=mask,
+                src_key_padding_mask=src_key_padding_mask,
+            )
+
+        if self.norm is not None:
+            output = self.norm(output)
+
+        return output
+
+
+class RelPositionalEncoding(torch.nn.Module):
+    """Relative positional encoding module.
+
+    See : Appendix B in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+    Modified from https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/embedding.py
+
+    Args:
+        d_model: Embedding dimension.
+        dropout_rate: Dropout rate.
+        max_len: Maximum input length.
+
+    """
+
+    def __init__(
+        self, d_model: int, dropout_rate: float, max_len: int = 5000
+    ) -> None:
+        """Construct an PositionalEncoding object."""
+        super(RelPositionalEncoding, self).__init__()
+        self.d_model = d_model
+        self.xscale = math.sqrt(self.d_model)
+        self.dropout = torch.nn.Dropout(p=dropout_rate)
+        self.pe = None
+        self.extend_pe(torch.tensor(0.0).expand(1, max_len))
+
+    def extend_pe(self, x: Tensor) -> None:
+        """Reset the positional encodings."""
+        if self.pe is not None:
+            # self.pe contains both positive and negative parts
+            # the length of self.pe is 2 * input_len - 1
+            if self.pe.size(1) >= x.size(1) * 2 - 1:
+                # Note: TorchScript doesn't implement operator== for torch.Device
+                if self.pe.dtype != x.dtype or str(self.pe.device) != str(
+                    x.device
+                ):
+                    self.pe = self.pe.to(dtype=x.dtype, device=x.device)
+                return
+        # Suppose `i` means to the position of query vecotr and `j` means the
+        # position of key vector. We use position relative positions when keys
+        # are to the left (i>j) and negative relative positions otherwise (i<j).
+        pe_positive = torch.zeros(x.size(1), self.d_model)
+        pe_negative = torch.zeros(x.size(1), self.d_model)
+        position = torch.arange(0, x.size(1), dtype=torch.float32).unsqueeze(1)
+        div_term = torch.exp(
+            torch.arange(0, self.d_model, 2, dtype=torch.float32)
+            * -(math.log(10000.0) / self.d_model)
+        )
+        pe_positive[:, 0::2] = torch.sin(position * div_term)
+        pe_positive[:, 1::2] = torch.cos(position * div_term)
+        pe_negative[:, 0::2] = torch.sin(-1 * position * div_term)
+        pe_negative[:, 1::2] = torch.cos(-1 * position * div_term)
+
+        # Reserve the order of positive indices and concat both positive and
+        # negative indices. This is used to support the shifting trick
+        # as in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+        pe_positive = torch.flip(pe_positive, [0]).unsqueeze(0)
+        pe_negative = pe_negative[1:].unsqueeze(0)
+        pe = torch.cat([pe_positive, pe_negative], dim=1)
+        self.pe = pe.to(device=x.device, dtype=x.dtype)
+
+    def forward(self, x: torch.Tensor) -> Tuple[Tensor, Tensor]:
+        """Add positional encoding.
+
+        Args:
+            x (torch.Tensor): Input tensor (batch, time, `*`).
+
+        Returns:
+            torch.Tensor: Encoded tensor (batch, time, `*`).
+            torch.Tensor: Encoded tensor (batch, 2*time-1, `*`).
+
+        """
+        self.extend_pe(x)
+        x = x * self.xscale
+        pos_emb = self.pe[
+            :,
+            self.pe.size(1) // 2
+            - x.size(1)
+            + 1 : self.pe.size(1) // 2  # noqa E203
+            + x.size(1),
+        ]
+        return self.dropout(x), self.dropout(pos_emb)
+
+
+class RelPositionMultiheadAttention(nn.Module):
+    r"""Multi-Head Attention layer with relative position encoding
+
+    See reference: "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context"
+
+    Args:
+        embed_dim: total dimension of the model.
+        num_heads: parallel attention heads.
+        dropout: a Dropout layer on attn_output_weights. Default: 0.0.
+
+    Examples::
+
+        >>> rel_pos_multihead_attn = RelPositionMultiheadAttention(embed_dim, num_heads)
+        >>> attn_output, attn_output_weights = multihead_attn(query, key, value, pos_emb)
+    """
+
+    def __init__(
+        self,
+        embed_dim: int,
+        num_heads: int,
+        dropout: float = 0.0,
+    ) -> None:
+        super(RelPositionMultiheadAttention, self).__init__()
+        self.embed_dim = embed_dim
+        self.num_heads = num_heads
+        self.dropout = dropout
+        self.head_dim = embed_dim // num_heads
+        assert (
+            self.head_dim * num_heads == self.embed_dim
+        ), "embed_dim must be divisible by num_heads"
+
+        self.in_proj = nn.Linear(embed_dim, 3 * embed_dim, bias=True)
+        self.out_proj = nn.Linear(embed_dim, embed_dim, bias=True)
+
+        # linear transformation for positional encoding.
+        self.linear_pos = nn.Linear(embed_dim, embed_dim, bias=False)
+        # these two learnable bias are used in matrix c and matrix d
+        # as described in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context" Section 3.3
+        self.pos_bias_u = nn.Parameter(torch.Tensor(num_heads, self.head_dim))
+        self.pos_bias_v = nn.Parameter(torch.Tensor(num_heads, self.head_dim))
+
+        self._reset_parameters()
+
+    def _reset_parameters(self) -> None:
+        nn.init.xavier_uniform_(self.in_proj.weight)
+        nn.init.constant_(self.in_proj.bias, 0.0)
+        nn.init.constant_(self.out_proj.bias, 0.0)
+
+        nn.init.xavier_uniform_(self.pos_bias_u)
+        nn.init.xavier_uniform_(self.pos_bias_v)
+
+    def forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        pos_emb: Tensor,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        r"""
+        Args:
+            query, key, value: map a query and a set of key-value pairs to an output.
+            pos_emb: Positional embedding tensor
+            key_padding_mask: if provided, specified padding elements in the key will
+                be ignored by the attention. When given a binary mask and a value is True,
+                the corresponding value on the attention layer will be ignored. When given
+                a byte mask and a value is non-zero, the corresponding value on the attention
+                layer will be ignored
+            need_weights: output attn_output_weights.
+            attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all
+                the batches while a 3D mask allows to specify a different mask for the entries of each batch.
+
+        Shape:
+            - Inputs:
+            - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - pos_emb: :math:`(N, 2*L-1, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.
+            If a ByteTensor is provided, the non-zero positions will be ignored while the position
+            with the zero positions will be unchanged. If a BoolTensor is provided, the positions with the
+            value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.
+            - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.
+            3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,
+            S is the source sequence length. attn_mask ensure that position i is allowed to attend the unmasked
+            positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend
+            while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``
+            is not allowed to attend while ``False`` values will be unchanged. If a FloatTensor
+            is provided, it will be added to the attention weight.
+
+            - Outputs:
+            - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,
+            E is the embedding dimension.
+            - attn_output_weights: :math:`(N, L, S)` where N is the batch size,
+            L is the target sequence length, S is the source sequence length.
+        """
+        return self.multi_head_attention_forward(
+            query,
+            key,
+            value,
+            pos_emb,
+            self.embed_dim,
+            self.num_heads,
+            self.in_proj.weight,
+            self.in_proj.bias,
+            self.dropout,
+            self.out_proj.weight,
+            self.out_proj.bias,
+            training=self.training,
+            key_padding_mask=key_padding_mask,
+            need_weights=need_weights,
+            attn_mask=attn_mask,
+        )
+
+    def rel_shift(self, x: Tensor) -> Tensor:
+        """Compute relative positional encoding.
+
+        Args:
+            x: Input tensor (batch, head, time1, 2*time1-1).
+                time1 means the length of query vector.
+
+        Returns:
+            Tensor: tensor of shape (batch, head, time1, time2)
+          (note: time2 has the same value as time1, but it is for
+          the key, while time1 is for the query).
+        """
+        (batch_size, num_heads, time1, n) = x.shape
+        assert n == 2 * time1 - 1
+        # Note: TorchScript requires explicit arg for stride()
+        batch_stride = x.stride(0)
+        head_stride = x.stride(1)
+        time1_stride = x.stride(2)
+        n_stride = x.stride(3)
+        return x.as_strided(
+            (batch_size, num_heads, time1, time1),
+            (batch_stride, head_stride, time1_stride - n_stride, n_stride),
+            storage_offset=n_stride * (time1 - 1),
+        )
+
+    def multi_head_attention_forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        pos_emb: Tensor,
+        embed_dim_to_check: int,
+        num_heads: int,
+        in_proj_weight: Tensor,
+        in_proj_bias: Tensor,
+        dropout_p: float,
+        out_proj_weight: Tensor,
+        out_proj_bias: Tensor,
+        training: bool = True,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        r"""
+        Args:
+            query, key, value: map a query and a set of key-value pairs to an output.
+            pos_emb: Positional embedding tensor
+            embed_dim_to_check: total dimension of the model.
+            num_heads: parallel attention heads.
+            in_proj_weight, in_proj_bias: input projection weight and bias.
+            dropout_p: probability of an element to be zeroed.
+            out_proj_weight, out_proj_bias: the output projection weight and bias.
+            training: apply dropout if is ``True``.
+            key_padding_mask: if provided, specified padding elements in the key will
+                be ignored by the attention. This is an binary mask. When the value is True,
+                the corresponding value on the attention layer will be filled with -inf.
+            need_weights: output attn_output_weights.
+            attn_mask: 2D or 3D mask that prevents attention to certain positions. A 2D mask will be broadcasted for all
+                the batches while a 3D mask allows to specify a different mask for the entries of each batch.
+
+        Shape:
+            Inputs:
+            - query: :math:`(L, N, E)` where L is the target sequence length, N is the batch size, E is
+            the embedding dimension.
+            - key: :math:`(S, N, E)`, where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - value: :math:`(S, N, E)` where S is the source sequence length, N is the batch size, E is
+            the embedding dimension.
+            - pos_emb: :math:`(N, 2*L-1, E)` or :math:`(1, 2*L-1, E)` where L is the target sequence
+            length, N is the batch size, E is the embedding dimension.
+            - key_padding_mask: :math:`(N, S)` where N is the batch size, S is the source sequence length.
+            If a ByteTensor is provided, the non-zero positions will be ignored while the zero positions
+            will be unchanged. If a BoolTensor is provided, the positions with the
+            value of ``True`` will be ignored while the position with the value of ``False`` will be unchanged.
+            - attn_mask: 2D mask :math:`(L, S)` where L is the target sequence length, S is the source sequence length.
+            3D mask :math:`(N*num_heads, L, S)` where N is the batch size, L is the target sequence length,
+            S is the source sequence length. attn_mask ensures that position i is allowed to attend the unmasked
+            positions. If a ByteTensor is provided, the non-zero positions are not allowed to attend
+            while the zero positions will be unchanged. If a BoolTensor is provided, positions with ``True``
+            are not allowed to attend while ``False`` values will be unchanged. If a FloatTensor
+            is provided, it will be added to the attention weight.
+
+            Outputs:
+            - attn_output: :math:`(L, N, E)` where L is the target sequence length, N is the batch size,
+            E is the embedding dimension.
+            - attn_output_weights: :math:`(N, L, S)` where N is the batch size,
+            L is the target sequence length, S is the source sequence length.
+        """
+
+        tgt_len, bsz, embed_dim = query.size()
+        assert embed_dim == embed_dim_to_check
+        assert key.size(0) == value.size(0) and key.size(1) == value.size(1)
+
+        head_dim = embed_dim // num_heads
+        assert (
+            head_dim * num_heads == embed_dim
+        ), "embed_dim must be divisible by num_heads"
+        scaling = float(head_dim) ** -0.5
+
+        if torch.equal(query, key) and torch.equal(key, value):
+            # self-attention
+            q, k, v = nn.functional.linear(
+                query, in_proj_weight, in_proj_bias
+            ).chunk(3, dim=-1)
+
+        elif torch.equal(key, value):
+            # encoder-decoder attention
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = 0
+            _end = embed_dim
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            q = nn.functional.linear(query, _w, _b)
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim
+            _end = None
+            _w = in_proj_weight[_start:, :]
+            if _b is not None:
+                _b = _b[_start:]
+            k, v = nn.functional.linear(key, _w, _b).chunk(2, dim=-1)
+
+        else:
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = 0
+            _end = embed_dim
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            q = nn.functional.linear(query, _w, _b)
+
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim
+            _end = embed_dim * 2
+            _w = in_proj_weight[_start:_end, :]
+            if _b is not None:
+                _b = _b[_start:_end]
+            k = nn.functional.linear(key, _w, _b)
+
+            # This is inline in_proj function with in_proj_weight and in_proj_bias
+            _b = in_proj_bias
+            _start = embed_dim * 2
+            _end = None
+            _w = in_proj_weight[_start:, :]
+            if _b is not None:
+                _b = _b[_start:]
+            v = nn.functional.linear(value, _w, _b)
+
+        if attn_mask is not None:
+            assert (
+                attn_mask.dtype == torch.float32
+                or attn_mask.dtype == torch.float64
+                or attn_mask.dtype == torch.float16
+                or attn_mask.dtype == torch.uint8
+                or attn_mask.dtype == torch.bool
+            ), "Only float, byte, and bool types are supported for attn_mask, not {}".format(
+                attn_mask.dtype
+            )
+            if attn_mask.dtype == torch.uint8:
+                warnings.warn(
+                    "Byte tensor for attn_mask is deprecated. Use bool tensor instead."
+                )
+                attn_mask = attn_mask.to(torch.bool)
+
+            if attn_mask.dim() == 2:
+                attn_mask = attn_mask.unsqueeze(0)
+                if list(attn_mask.size()) != [1, query.size(0), key.size(0)]:
+                    raise RuntimeError(
+                        "The size of the 2D attn_mask is not correct."
+                    )
+            elif attn_mask.dim() == 3:
+                if list(attn_mask.size()) != [
+                    bsz * num_heads,
+                    query.size(0),
+                    key.size(0),
+                ]:
+                    raise RuntimeError(
+                        "The size of the 3D attn_mask is not correct."
+                    )
+            else:
+                raise RuntimeError(
+                    "attn_mask's dimension {} is not supported".format(
+                        attn_mask.dim()
+                    )
+                )
+            # attn_mask's dim is 3 now.
+
+        # convert ByteTensor key_padding_mask to bool
+        if (
+            key_padding_mask is not None
+            and key_padding_mask.dtype == torch.uint8
+        ):
+            warnings.warn(
+                "Byte tensor for key_padding_mask is deprecated. Use bool tensor instead."
+            )
+            key_padding_mask = key_padding_mask.to(torch.bool)
+
+        q = q.contiguous().view(tgt_len, bsz, num_heads, head_dim)
+        k = k.contiguous().view(-1, bsz, num_heads, head_dim)
+        v = v.contiguous().view(-1, bsz * num_heads, head_dim).transpose(0, 1)
+
+        src_len = k.size(0)
+
+        if key_padding_mask is not None:
+            assert key_padding_mask.size(0) == bsz, "{} == {}".format(
+                key_padding_mask.size(0), bsz
+            )
+            assert key_padding_mask.size(1) == src_len, "{} == {}".format(
+                key_padding_mask.size(1), src_len
+            )
+
+        q = q.transpose(0, 1)  # (batch, time1, head, d_k)
+
+        pos_emb_bsz = pos_emb.size(0)
+        assert pos_emb_bsz in (1, bsz)  # actually it is 1
+        p = self.linear_pos(pos_emb).view(pos_emb_bsz, -1, num_heads, head_dim)
+        p = p.transpose(1, 2)  # (batch, head, 2*time1-1, d_k)
+
+        q_with_bias_u = (q + self.pos_bias_u).transpose(
+            1, 2
+        )  # (batch, head, time1, d_k)
+
+        q_with_bias_v = (q + self.pos_bias_v).transpose(
+            1, 2
+        )  # (batch, head, time1, d_k)
+
+        # compute attention score
+        # first compute matrix a and matrix c
+        # as described in "Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context" Section 3.3
+        k = k.permute(1, 2, 3, 0)  # (batch, head, d_k, time2)
+        matrix_ac = torch.matmul(
+            q_with_bias_u, k
+        )  # (batch, head, time1, time2)
+
+        # compute matrix b and matrix d
+        matrix_bd = torch.matmul(
+            q_with_bias_v, p.transpose(-2, -1)
+        )  # (batch, head, time1, 2*time1-1)
+        matrix_bd = self.rel_shift(matrix_bd)
+
+        attn_output_weights = (
+            matrix_ac + matrix_bd
+        ) * scaling  # (batch, head, time1, time2)
+
+        attn_output_weights = attn_output_weights.view(
+            bsz * num_heads, tgt_len, -1
+        )
+
+        assert list(attn_output_weights.size()) == [
+            bsz * num_heads,
+            tgt_len,
+            src_len,
+        ]
+
+        if attn_mask is not None:
+            if attn_mask.dtype == torch.bool:
+                attn_output_weights.masked_fill_(attn_mask, float("-inf"))
+            else:
+                attn_output_weights += attn_mask
+
+        if key_padding_mask is not None:
+            attn_output_weights = attn_output_weights.view(
+                bsz, num_heads, tgt_len, src_len
+            )
+            attn_output_weights = attn_output_weights.masked_fill(
+                key_padding_mask.unsqueeze(1).unsqueeze(2),
+                float("-inf"),
+            )
+            attn_output_weights = attn_output_weights.view(
+                bsz * num_heads, tgt_len, src_len
+            )
+
+        attn_output_weights = nn.functional.softmax(attn_output_weights, dim=-1)
+        attn_output_weights = nn.functional.dropout(
+            attn_output_weights, p=dropout_p, training=training
+        )
+
+        attn_output = torch.bmm(attn_output_weights, v)
+        assert list(attn_output.size()) == [bsz * num_heads, tgt_len, head_dim]
+        attn_output = (
+            attn_output.transpose(0, 1)
+            .contiguous()
+            .view(tgt_len, bsz, embed_dim)
+        )
+        attn_output = nn.functional.linear(
+            attn_output, out_proj_weight, out_proj_bias
+        )
+
+        if need_weights:
+            # average attention weights over heads
+            attn_output_weights = attn_output_weights.view(
+                bsz, num_heads, tgt_len, src_len
+            )
+            return attn_output, attn_output_weights.sum(dim=1) / num_heads
+        else:
+            return attn_output, None
+
+
+class ConvolutionModule(nn.Module):
+    """ConvolutionModule in Conformer model.
+    Modified from https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/conformer/convolution.py
+
+    Args:
+        channels (int): The number of channels of conv layers.
+        kernel_size (int): Kernerl size of conv layers.
+        bias (bool): Whether to use bias in conv layers (default=True).
+
+    """
+
+    def __init__(
+        self, channels: int, kernel_size: int, bias: bool = True
+    ) -> None:
+        """Construct an ConvolutionModule object."""
+        super(ConvolutionModule, self).__init__()
+        # kernerl_size should be a odd number for 'SAME' padding
+        assert (kernel_size - 1) % 2 == 0
+
+        self.pointwise_conv1 = nn.Conv1d(
+            channels,
+            2 * channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            bias=bias,
+        )
+        self.depthwise_conv = nn.Conv1d(
+            channels,
+            channels,
+            kernel_size,
+            stride=1,
+            padding=(kernel_size - 1) // 2,
+            groups=channels,
+            bias=bias,
+        )
+        self.norm = nn.BatchNorm1d(channels)
+        self.pointwise_conv2 = nn.Conv1d(
+            channels,
+            channels,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            bias=bias,
+        )
+        self.activation = Swish()
+
+    def forward(self, x: Tensor) -> Tensor:
+        """Compute convolution module.
+
+        Args:
+            x: Input tensor (#time, batch, channels).
+
+        Returns:
+            Tensor: Output tensor (#time, batch, channels).
+
+        """
+        # exchange the temporal dimension and the feature dimension
+        x = x.permute(1, 2, 0)  # (#batch, channels, time).
+
+        # GLU mechanism
+        x = self.pointwise_conv1(x)  # (batch, 2*channels, time)
+        x = nn.functional.glu(x, dim=1)  # (batch, channels, time)
+
+        # 1D Depthwise Conv
+        x = self.depthwise_conv(x)
+        x = self.activation(self.norm(x))
+
+        x = self.pointwise_conv2(x)  # (batch, channel, time)
+
+        return x.permute(2, 0, 1)
+
+
+class Swish(torch.nn.Module):
+    """Construct an Swish object."""
+
+    def forward(self, x: Tensor) -> Tensor:
+        """Return Swich activation function."""
+        return x * torch.sigmoid(x)
+
+
+def identity(x):
+    return x

egs/librispeech/ASR/transducer_stateless/decoder.py

@@ -0,0 +1,65 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+import torch.nn as nn
+
+
+class Decoder(nn.Module):
+    """This class implements the stateless decoder from the following paper:
+
+        RNN-transducer with stateless prediction network
+        https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9054419
+
+    It removes the recurrent connection from the decoder, i.e., the prediction
+    network.
+
+    TODO: Implement https://arxiv.org/pdf/2109.07513.pdf
+    """
+
+    def __init__(
+        self,
+        vocab_size: int,
+        embedding_dim: int,
+        blank_id: int,
+    ):
+        """
+        Args:
+          vocab_size:
+            Number of tokens of the modeling unit including blank.
+          embedding_dim:
+            Dimension of the input embedding.
+          blank_id:
+            The ID of the blank symbol.
+        """
+        super().__init__()
+        self.embedding = nn.Embedding(
+            num_embeddings=vocab_size,
+            embedding_dim=embedding_dim,
+            padding_idx=blank_id,
+        )
+        self.blank_id = blank_id
+
+    def forward(self, y: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+          y:
+            A 2-D tensor of shape (N, U) with blank prepended.
+        Returns:
+          Return a tensor of shape (N, U, embedding_dim).
+        """
+        embeding_out = self.embedding(y)
+        return embeding_out

egs/librispeech/ASR/transducer_stateless/encoder_interface.py

@@ -0,0 +1,43 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Tuple
+
+import torch
+import torch.nn as nn
+
+
+class EncoderInterface(nn.Module):
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            A tensor of shape (batch_size, input_seq_len, num_features)
+            containing the input features.
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames
+            in `x` before padding.
+        Returns:
+          Return a tuple containing two tensors:
+            - encoder_out, a tensor of (batch_size, out_seq_len, output_dim)
+              containing unnormalized probabilities, i.e., the output of a
+              linear layer.
+            - encoder_out_lens, a tensor of shape (batch_size,) containing
+              the number of frames in `encoder_out` before padding.
+        """
+        raise NotImplementedError("Please implement it in a subclass")

egs/librispeech/ASR/transducer_stateless/joiner.py

@@ -0,0 +1,55 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class Joiner(nn.Module):
+    def __init__(self, input_dim: int, output_dim: int):
+        super().__init__()
+
+        self.output_linear = nn.Linear(input_dim, output_dim)
+
+    def forward(
+        self, encoder_out: torch.Tensor, decoder_out: torch.Tensor
+    ) -> torch.Tensor:
+        """
+        Args:
+          encoder_out:
+            Output from the encoder. Its shape is (N, T, C).
+          decoder_out:
+            Output from the decoder. Its shape is (N, U, C).
+        Returns:
+          Return a tensor of shape (N, T, U, C).
+        """
+        assert encoder_out.ndim == decoder_out.ndim == 3
+        assert encoder_out.size(0) == decoder_out.size(0)
+        assert encoder_out.size(2) == decoder_out.size(2)
+
+        encoder_out = encoder_out.unsqueeze(2)
+        # Now encoder_out is (N, T, 1, C)
+
+        decoder_out = decoder_out.unsqueeze(1)
+        # Now decoder_out is (N, 1, U, C)
+
+        logit = encoder_out + decoder_out
+        logit = F.relu(logit)
+
+        output = self.output_linear(logit)
+
+        return output

egs/librispeech/ASR/transducer_stateless/model.py

@@ -0,0 +1,125 @@
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Note we use `rnnt_loss` from torchaudio, which exists only in
+torchaudio >= v0.10.0. It also means you have to use torch >= v1.10.0
+"""
+import k2
+import torch
+import torch.nn as nn
+import torchaudio
+import torchaudio.functional
+from encoder_interface import EncoderInterface
+
+from icefall.utils import add_sos
+
+assert hasattr(torchaudio.functional, "rnnt_loss"), (
+    f"Current torchaudio version: {torchaudio.__version__}\n"
+    "Please install a version >= 0.10.0"
+)
+
+
+class Transducer(nn.Module):
+    """It implements https://arxiv.org/pdf/1211.3711.pdf
+    "Sequence Transduction with Recurrent Neural Networks"
+    """
+
+    def __init__(
+        self,
+        encoder: EncoderInterface,
+        decoder: nn.Module,
+        joiner: nn.Module,
+    ):
+        """
+        Args:
+          encoder:
+            It is the transcription network in the paper. Its accepts
+            two inputs: `x` of (N, T, C) and `x_lens` of shape (N,).
+            It returns two tensors: `logits` of shape (N, T, C) and
+            `logit_lens` of shape (N,).
+          decoder:
+            It is the prediction network in the paper. Its input shape
+            is (N, U) and its output shape is (N, U, C). It should contain
+            one attribute: `blank_id`.
+          joiner:
+            It has two inputs with shapes: (N, T, C) and (N, U, C). Its
+            output shape is (N, T, U, C). Note that its output contains
+            unnormalized probs, i.e., not processed by log-softmax.
+        """
+        super().__init__()
+        assert isinstance(encoder, EncoderInterface)
+        assert hasattr(decoder, "blank_id")
+
+        self.encoder = encoder
+        self.decoder = decoder
+        self.joiner = joiner
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        x_lens: torch.Tensor,
+        y: k2.RaggedTensor,
+    ) -> torch.Tensor:
+        """
+        Args:
+          x:
+            A 3-D tensor of shape (N, T, C).
+          x_lens:
+            A 1-D tensor of shape (N,). It contains the number of frames in `x`
+            before padding.
+          y:
+            A ragged tensor with 2 axes [utt][label]. It contains labels of each
+            utterance.
+        Returns:
+          Return the transducer loss.
+        """
+        assert x.ndim == 3, x.shape
+        assert x_lens.ndim == 1, x_lens.shape
+        assert y.num_axes == 2, y.num_axes
+
+        assert x.size(0) == x_lens.size(0) == y.dim0
+
+        encoder_out, x_lens = self.encoder(x, x_lens)
+        assert torch.all(x_lens > 0)
+
+        # Now for the decoder, i.e., the prediction network
+        row_splits = y.shape.row_splits(1)
+        y_lens = row_splits[1:] - row_splits[:-1]
+
+        blank_id = self.decoder.blank_id
+        sos_y = add_sos(y, sos_id=blank_id)
+
+        sos_y_padded = sos_y.pad(mode="constant", padding_value=blank_id)
+
+        decoder_out = self.decoder(sos_y_padded)
+
+        logits = self.joiner(encoder_out, decoder_out)
+
+        # rnnt_loss requires 0 padded targets
+        # Note: y does not start with SOS
+        y_padded = y.pad(mode="constant", padding_value=0)
+
+        loss = torchaudio.functional.rnnt_loss(
+            logits=logits,
+            targets=y_padded,
+            logit_lengths=x_lens,
+            target_lengths=y_lens,
+            blank=blank_id,
+            reduction="sum",
+        )
+
+        return loss

egs/librispeech/ASR/transducer_stateless/subsampling.py

@@ -0,0 +1 @@
+../transducer/subsampling.py

egs/librispeech/ASR/transducer_stateless/train.py

@@ -0,0 +1,734 @@
+#!/usr/bin/env python3
+# Copyright    2021  Xiaomi Corp.        (authors: Fangjun Kuang,
+#                                                  Wei Kang
+#                                                  Mingshuang Luo)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Usage:
+
+export CUDA_VISIBLE_DEVICES="0,1,2,3"
+
+./transducer_stateless/train.py \
+  --world-size 4 \
+  --num-epochs 30 \
+  --start-epoch 0 \
+  --exp-dir transducer_stateless/exp \
+  --full-libri 1 \
+  --max-duration 250 \
+  --lr-factor 2.5
+"""
+
+
+import argparse
+import logging
+from pathlib import Path
+from shutil import copyfile
+from typing import Optional, Tuple
+
+import k2
+import sentencepiece as spm
+import torch
+import torch.multiprocessing as mp
+import torch.nn as nn
+from asr_datamodule import LibriSpeechAsrDataModule
+from conformer import Conformer
+from decoder import Decoder
+from joiner import Joiner
+from lhotse.cut import Cut
+from lhotse.utils import fix_random_seed
+from model import Transducer
+from torch import Tensor
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.nn.utils import clip_grad_norm_
+from torch.utils.tensorboard import SummaryWriter
+from transformer import Noam
+
+from icefall.checkpoint import load_checkpoint
+from icefall.checkpoint import save_checkpoint as save_checkpoint_impl
+from icefall.dist import cleanup_dist, setup_dist
+from icefall.env import get_env_info
+from icefall.utils import AttributeDict, MetricsTracker, setup_logger, str2bool
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--world-size",
+        type=int,
+        default=1,
+        help="Number of GPUs for DDP training.",
+    )
+
+    parser.add_argument(
+        "--master-port",
+        type=int,
+        default=12354,
+        help="Master port to use for DDP training.",
+    )
+
+    parser.add_argument(
+        "--tensorboard",
+        type=str2bool,
+        default=True,
+        help="Should various information be logged in tensorboard.",
+    )
+
+    parser.add_argument(
+        "--num-epochs",
+        type=int,
+        default=78,
+        help="Number of epochs to train.",
+    )
+
+    parser.add_argument(
+        "--start-epoch",
+        type=int,
+        default=0,
+        help="""Resume training from from this epoch.
+        If it is positive, it will load checkpoint from
+        transducer_stateless/exp/epoch-{start_epoch-1}.pt
+        """,
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="transducer_stateless/exp",
+        help="""The experiment dir.
+        It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--lr-factor",
+        type=float,
+        default=5.0,
+        help="The lr_factor for Noam optimizer",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    """Return a dict containing training parameters.
+
+    All training related parameters that are not passed from the commandline
+    are saved in the variable `params`.
+
+    Commandline options are merged into `params` after they are parsed, so
+    you can also access them via `params`.
+
+    Explanation of options saved in `params`:
+
+        - best_train_loss: Best training loss so far. It is used to select
+                           the model that has the lowest training loss. It is
+                           updated during the training.
+
+        - best_valid_loss: Best validation loss so far. It is used to select
+                           the model that has the lowest validation loss. It is
+                           updated during the training.
+
+        - best_train_epoch: It is the epoch that has the best training loss.
+
+        - best_valid_epoch: It is the epoch that has the best validation loss.
+
+        - batch_idx_train: Used to writing statistics to tensorboard. It
+                           contains number of batches trained so far across
+                           epochs.
+
+        - log_interval:  Print training loss if batch_idx % log_interval` is 0
+
+        - reset_interval: Reset statistics if batch_idx % reset_interval is 0
+
+        - valid_interval:  Run validation if batch_idx % valid_interval is 0
+
+        - feature_dim: The model input dim. It has to match the one used
+                       in computing features.
+
+        - subsampling_factor:  The subsampling factor for the model.
+
+        - use_feat_batchnorm: Whether to do batch normalization for the
+                              input features.
+
+        - attention_dim: Hidden dim for multi-head attention model.
+
+        - num_decoder_layers: Number of decoder layer of transformer decoder.
+
+        - weight_decay:  The weight_decay for the optimizer.
+
+        - warm_step: The warm_step for Noam optimizer.
+    """
+    params = AttributeDict(
+        {
+            "best_train_loss": float("inf"),
+            "best_valid_loss": float("inf"),
+            "best_train_epoch": -1,
+            "best_valid_epoch": -1,
+            "batch_idx_train": 0,
+            "log_interval": 50,
+            "reset_interval": 200,
+            "valid_interval": 3000,  # For the 100h subset, use 800
+            # parameters for conformer
+            "feature_dim": 80,
+            "encoder_out_dim": 512,
+            "subsampling_factor": 4,
+            "attention_dim": 512,
+            "nhead": 8,
+            "dim_feedforward": 2048,
+            "num_encoder_layers": 12,
+            "vgg_frontend": False,
+            "use_feat_batchnorm": True,
+            # parameters for Noam
+            "weight_decay": 1e-6,
+            "warm_step": 80000,  # For the 100h subset, use 8k
+            "env_info": get_env_info(),
+        }
+    )
+
+    return params
+
+
+def get_encoder_model(params: AttributeDict):
+    # TODO: We can add an option to switch between Conformer and Transformer
+    encoder = Conformer(
+        num_features=params.feature_dim,
+        output_dim=params.encoder_out_dim,
+        subsampling_factor=params.subsampling_factor,
+        d_model=params.attention_dim,
+        nhead=params.nhead,
+        dim_feedforward=params.dim_feedforward,
+        num_encoder_layers=params.num_encoder_layers,
+        vgg_frontend=params.vgg_frontend,
+        use_feat_batchnorm=params.use_feat_batchnorm,
+    )
+    return encoder
+
+
+def get_decoder_model(params: AttributeDict):
+    decoder = Decoder(
+        vocab_size=params.vocab_size,
+        embedding_dim=params.encoder_out_dim,
+        blank_id=params.blank_id,
+    )
+    return decoder
+
+
+def get_joiner_model(params: AttributeDict):
+    joiner = Joiner(
+        input_dim=params.encoder_out_dim,
+        output_dim=params.vocab_size,
+    )
+    return joiner
+
+
+def get_transducer_model(params: AttributeDict):
+    encoder = get_encoder_model(params)
+    decoder = get_decoder_model(params)
+    joiner = get_joiner_model(params)
+
+    model = Transducer(
+        encoder=encoder,
+        decoder=decoder,
+        joiner=joiner,
+    )
+    return model
+
+
+def load_checkpoint_if_available(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+) -> None:
+    """Load checkpoint from file.
+
+    If params.start_epoch is positive, it will load the checkpoint from
+    `params.start_epoch - 1`. Otherwise, this function does nothing.
+
+    Apart from loading state dict for `model`, `optimizer` and `scheduler`,
+    it also updates `best_train_epoch`, `best_train_loss`, `best_valid_epoch`,
+    and `best_valid_loss` in `params`.
+
+    Args:
+      params:
+        The return value of :func:`get_params`.
+      model:
+        The training model.
+      optimizer:
+        The optimizer that we are using.
+      scheduler:
+        The learning rate scheduler we are using.
+    Returns:
+      Return None.
+    """
+    if params.start_epoch <= 0:
+        return
+
+    filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
+    saved_params = load_checkpoint(
+        filename,
+        model=model,
+        optimizer=optimizer,
+        scheduler=scheduler,
+    )
+
+    keys = [
+        "best_train_epoch",
+        "best_valid_epoch",
+        "batch_idx_train",
+        "best_train_loss",
+        "best_valid_loss",
+    ]
+    for k in keys:
+        params[k] = saved_params[k]
+
+    return saved_params
+
+
+def save_checkpoint(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[torch.optim.lr_scheduler._LRScheduler] = None,
+    rank: int = 0,
+) -> None:
+    """Save model, optimizer, scheduler and training stats to file.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The training model.
+    """
+    if rank != 0:
+        return
+    filename = params.exp_dir / f"epoch-{params.cur_epoch}.pt"
+    save_checkpoint_impl(
+        filename=filename,
+        model=model,
+        params=params,
+        optimizer=optimizer,
+        scheduler=scheduler,
+        rank=rank,
+    )
+
+    if params.best_train_epoch == params.cur_epoch:
+        best_train_filename = params.exp_dir / "best-train-loss.pt"
+        copyfile(src=filename, dst=best_train_filename)
+
+    if params.best_valid_epoch == params.cur_epoch:
+        best_valid_filename = params.exp_dir / "best-valid-loss.pt"
+        copyfile(src=filename, dst=best_valid_filename)
+
+
+def compute_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    batch: dict,
+    is_training: bool,
+) -> Tuple[Tensor, MetricsTracker]:
+    """
+    Compute CTC loss given the model and its inputs.
+
+    Args:
+      params:
+        Parameters for training. See :func:`get_params`.
+      model:
+        The model for training. It is an instance of Conformer in our case.
+      batch:
+        A batch of data. See `lhotse.dataset.K2SpeechRecognitionDataset()`
+        for the content in it.
+      is_training:
+        True for training. False for validation. When it is True, this
+        function enables autograd during computation; when it is False, it
+        disables autograd.
+    """
+    device = model.device
+    feature = batch["inputs"]
+    # at entry, feature is (N, T, C)
+    assert feature.ndim == 3
+    feature = feature.to(device)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    texts = batch["supervisions"]["text"]
+    y = sp.encode(texts, out_type=int)
+    y = k2.RaggedTensor(y).to(device)
+
+    with torch.set_grad_enabled(is_training):
+        loss = model(x=feature, x_lens=feature_lens, y=y)
+
+    assert loss.requires_grad == is_training
+
+    info = MetricsTracker()
+    info["frames"] = (feature_lens // params.subsampling_factor).sum().item()
+
+    # Note: We use reduction=sum while computing the loss.
+    info["loss"] = loss.detach().cpu().item()
+
+    return loss, info
+
+
+def compute_validation_loss(
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    valid_dl: torch.utils.data.DataLoader,
+    world_size: int = 1,
+) -> MetricsTracker:
+    """Run the validation process."""
+    model.eval()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(valid_dl):
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+            is_training=False,
+        )
+        assert loss.requires_grad is False
+        tot_loss = tot_loss + loss_info
+
+    if world_size > 1:
+        tot_loss.reduce(loss.device)
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    if loss_value < params.best_valid_loss:
+        params.best_valid_epoch = params.cur_epoch
+        params.best_valid_loss = loss_value
+
+    return tot_loss
+
+
+def train_one_epoch(
+    params: AttributeDict,
+    model: nn.Module,
+    optimizer: torch.optim.Optimizer,
+    sp: spm.SentencePieceProcessor,
+    train_dl: torch.utils.data.DataLoader,
+    valid_dl: torch.utils.data.DataLoader,
+    tb_writer: Optional[SummaryWriter] = None,
+    world_size: int = 1,
+) -> None:
+    """Train the model for one epoch.
+
+    The training loss from the mean of all frames is saved in
+    `params.train_loss`. It runs the validation process every
+    `params.valid_interval` batches.
+
+    Args:
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The model for training.
+      optimizer:
+        The optimizer we are using.
+      train_dl:
+        Dataloader for the training dataset.
+      valid_dl:
+        Dataloader for the validation dataset.
+      tb_writer:
+        Writer to write log messages to tensorboard.
+      world_size:
+        Number of nodes in DDP training. If it is 1, DDP is disabled.
+    """
+    model.train()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(train_dl):
+        params.batch_idx_train += 1
+        batch_size = len(batch["supervisions"]["text"])
+
+        loss, loss_info = compute_loss(
+            params=params,
+            model=model,
+            sp=sp,
+            batch=batch,
+            is_training=True,
+        )
+        # summary stats
+        tot_loss = (tot_loss * (1 - 1 / params.reset_interval)) + loss_info
+
+        # NOTE: We use reduction==sum and loss is computed over utterances
+        # in the batch and there is no normalization to it so far.
+
+        optimizer.zero_grad()
+        loss.backward()
+        clip_grad_norm_(model.parameters(), 5.0, 2.0)
+        optimizer.step()
+
+        if batch_idx % params.log_interval == 0:
+            logging.info(
+                f"Epoch {params.cur_epoch}, "
+                f"batch {batch_idx}, loss[{loss_info}], "
+                f"tot_loss[{tot_loss}], batch size: {batch_size}"
+            )
+
+        if batch_idx % params.log_interval == 0:
+
+            if tb_writer is not None:
+                loss_info.write_summary(
+                    tb_writer, "train/current_", params.batch_idx_train
+                )
+                tot_loss.write_summary(
+                    tb_writer, "train/tot_", params.batch_idx_train
+                )
+
+        if batch_idx > 0 and batch_idx % params.valid_interval == 0:
+            logging.info("Computing validation loss")
+            valid_info = compute_validation_loss(
+                params=params,
+                model=model,
+                sp=sp,
+                valid_dl=valid_dl,
+                world_size=world_size,
+            )
+            model.train()
+            logging.info(f"Epoch {params.cur_epoch}, validation: {valid_info}")
+            if tb_writer is not None:
+                valid_info.write_summary(
+                    tb_writer, "train/valid_", params.batch_idx_train
+                )
+
+    loss_value = tot_loss["loss"] / tot_loss["frames"]
+    params.train_loss = loss_value
+    if params.train_loss < params.best_train_loss:
+        params.best_train_epoch = params.cur_epoch
+        params.best_train_loss = params.train_loss
+
+
+def run(rank, world_size, args):
+    """
+    Args:
+      rank:
+        It is a value between 0 and `world_size-1`, which is
+        passed automatically by `mp.spawn()` in :func:`main`.
+        The node with rank 0 is responsible for saving checkpoint.
+      world_size:
+        Number of GPUs for DDP training.
+      args:
+        The return value of get_parser().parse_args()
+    """
+    params = get_params()
+    params.update(vars(args))
+    if params.full_libri is False:
+        params.valid_interval = 800
+        params.warm_step = 8000
+
+    fix_random_seed(42)
+    if world_size > 1:
+        setup_dist(rank, world_size, params.master_port)
+
+    setup_logger(f"{params.exp_dir}/log/log-train")
+    logging.info("Training started")
+
+    if args.tensorboard and rank == 0:
+        tb_writer = SummaryWriter(log_dir=f"{params.exp_dir}/tensorboard")
+    else:
+        tb_writer = None
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", rank)
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <sos/eos> are defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_transducer_model(params)
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    checkpoints = load_checkpoint_if_available(params=params, model=model)
+
+    model.to(device)
+    if world_size > 1:
+        logging.info("Using DDP")
+        model = DDP(model, device_ids=[rank])
+    model.device = device
+
+    optimizer = Noam(
+        model.parameters(),
+        model_size=params.attention_dim,
+        factor=params.lr_factor,
+        warm_step=params.warm_step,
+        weight_decay=params.weight_decay,
+    )
+
+    if checkpoints and "optimizer" in checkpoints:
+        logging.info("Loading optimizer state dict")
+        optimizer.load_state_dict(checkpoints["optimizer"])
+
+    librispeech = LibriSpeechAsrDataModule(args)
+
+    train_cuts = librispeech.train_clean_100_cuts()
+    if params.full_libri:
+        train_cuts += librispeech.train_clean_360_cuts()
+        train_cuts += librispeech.train_other_500_cuts()
+
+    def remove_short_and_long_utt(c: Cut):
+        # Keep only utterances with duration between 1 second and 20 seconds
+        return 1.0 <= c.duration <= 20.0
+
+    num_in_total = len(train_cuts)
+
+    train_cuts = train_cuts.filter(remove_short_and_long_utt)
+
+    num_left = len(train_cuts)
+    num_removed = num_in_total - num_left
+    removed_percent = num_removed / num_in_total * 100
+
+    logging.info(f"Before removing short and long utterances: {num_in_total}")
+    logging.info(f"After removing short and long utterances: {num_left}")
+    logging.info(f"Removed {num_removed} utterances ({removed_percent:.5f}%)")
+
+    train_dl = librispeech.train_dataloaders(train_cuts)
+
+    valid_cuts = librispeech.dev_clean_cuts()
+    valid_cuts += librispeech.dev_other_cuts()
+    valid_dl = librispeech.valid_dataloaders(valid_cuts)
+
+    scan_pessimistic_batches_for_oom(
+        model=model,
+        train_dl=train_dl,
+        optimizer=optimizer,
+        sp=sp,
+        params=params,
+    )
+
+    for epoch in range(params.start_epoch, params.num_epochs):
+        train_dl.sampler.set_epoch(epoch)
+
+        cur_lr = optimizer._rate
+        if tb_writer is not None:
+            tb_writer.add_scalar(
+                "train/learning_rate", cur_lr, params.batch_idx_train
+            )
+            tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)
+
+        if rank == 0:
+            logging.info("epoch {}, learning rate {}".format(epoch, cur_lr))
+
+        params.cur_epoch = epoch
+
+        train_one_epoch(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            sp=sp,
+            train_dl=train_dl,
+            valid_dl=valid_dl,
+            tb_writer=tb_writer,
+            world_size=world_size,
+        )
+
+        save_checkpoint(
+            params=params,
+            model=model,
+            optimizer=optimizer,
+            rank=rank,
+        )
+
+    logging.info("Done!")
+
+    if world_size > 1:
+        torch.distributed.barrier()
+        cleanup_dist()
+
+
+def scan_pessimistic_batches_for_oom(
+    model: nn.Module,
+    train_dl: torch.utils.data.DataLoader,
+    optimizer: torch.optim.Optimizer,
+    sp: spm.SentencePieceProcessor,
+    params: AttributeDict,
+):
+    from lhotse.dataset import find_pessimistic_batches
+
+    logging.info(
+        "Sanity check -- see if any of the batches in epoch 0 would cause OOM."
+    )
+    batches, crit_values = find_pessimistic_batches(train_dl.sampler)
+    for criterion, cuts in batches.items():
+        batch = train_dl.dataset[cuts]
+        try:
+            optimizer.zero_grad()
+            loss, _ = compute_loss(
+                params=params,
+                model=model,
+                sp=sp,
+                batch=batch,
+                is_training=True,
+            )
+            loss.backward()
+            clip_grad_norm_(model.parameters(), 5.0, 2.0)
+            optimizer.step()
+        except RuntimeError as e:
+            if "CUDA out of memory" in str(e):
+                logging.error(
+                    "Your GPU ran out of memory with the current "
+                    "max_duration setting. We recommend decreasing "
+                    "max_duration and trying again.\n"
+                    f"Failing criterion: {criterion} "
+                    f"(={crit_values[criterion]}) ..."
+                )
+            raise
+
+
+def main():
+    parser = get_parser()
+    LibriSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    world_size = args.world_size
+    assert world_size >= 1
+    if world_size > 1:
+        mp.spawn(run, args=(world_size, args), nprocs=world_size, join=True)
+    else:
+        run(rank=0, world_size=1, args=args)
+
+
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
+
+if __name__ == "__main__":
+    main()

egs/librispeech/ASR/transducer_stateless/transformer.py

@@ -0,0 +1,429 @@
+# Copyright    2021 University of Chinese Academy of Sciences (author: Han Zhu)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import math
+from typing import Optional, Tuple
+
+import torch
+import torch.nn as nn
+from transducer.encoder_interface import EncoderInterface
+from transducer.subsampling import Conv2dSubsampling, VggSubsampling
+
+from icefall.utils import make_pad_mask
+
+
+class Transformer(EncoderInterface):
+    def __init__(
+        self,
+        num_features: int,
+        output_dim: int,
+        subsampling_factor: int = 4,
+        d_model: int = 256,
+        nhead: int = 4,
+        dim_feedforward: int = 2048,
+        num_encoder_layers: int = 12,
+        dropout: float = 0.1,
+        normalize_before: bool = True,
+        vgg_frontend: bool = False,
+        use_feat_batchnorm: bool = False,
+    ) -> None:
+        """
+        Args:
+          num_features:
+            The input dimension of the model.
+          output_dim:
+            The output dimension of the model.
+          subsampling_factor:
+            Number of output frames is num_in_frames // subsampling_factor.
+            Currently, subsampling_factor MUST be 4.
+          d_model:
+            Attention dimension.
+          nhead:
+            Number of heads in multi-head attention.
+            Must satisfy d_model // nhead == 0.
+          dim_feedforward:
+            The output dimension of the feedforward layers in encoder.
+          num_encoder_layers:
+            Number of encoder layers.
+          dropout:
+            Dropout in encoder.
+          normalize_before:
+            If True, use pre-layer norm; False to use post-layer norm.
+          vgg_frontend:
+            True to use vgg style frontend for subsampling.
+          use_feat_batchnorm:
+            True to use batchnorm for the input layer.
+        """
+        super().__init__()
+        self.use_feat_batchnorm = use_feat_batchnorm
+        if use_feat_batchnorm:
+            self.feat_batchnorm = nn.BatchNorm1d(num_features)
+
+        self.num_features = num_features
+        self.output_dim = output_dim
+        self.subsampling_factor = subsampling_factor
+        if subsampling_factor != 4:
+            raise NotImplementedError("Support only 'subsampling_factor=4'.")
+
+        # self.encoder_embed converts the input of shape (N, T, num_features)
+        # to the shape (N, T//subsampling_factor, d_model).
+        # That is, it does two things simultaneously:
+        #   (1) subsampling: T -> T//subsampling_factor
+        #   (2) embedding: num_features -> d_model
+        if vgg_frontend:
+            self.encoder_embed = VggSubsampling(num_features, d_model)
+        else:
+            self.encoder_embed = Conv2dSubsampling(num_features, d_model)
+
+        self.encoder_pos = PositionalEncoding(d_model, dropout)
+
+        encoder_layer = TransformerEncoderLayer(
+            d_model=d_model,
+            nhead=nhead,
+            dim_feedforward=dim_feedforward,
+            dropout=dropout,
+            normalize_before=normalize_before,
+        )
+
+        if normalize_before:
+            encoder_norm = nn.LayerNorm(d_model)
+        else:
+            encoder_norm = None
+
+        self.encoder = nn.TransformerEncoder(
+            encoder_layer=encoder_layer,
+            num_layers=num_encoder_layers,
+            norm=encoder_norm,
+        )
+
+        # TODO(fangjun): remove dropout
+        self.encoder_output_layer = nn.Sequential(
+            nn.Dropout(p=dropout), nn.Linear(d_model, output_dim)
+        )
+
+    def forward(
+        self, x: torch.Tensor, x_lens: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Args:
+          x:
+            The input tensor. Its shape is (batch_size, seq_len, feature_dim).
+          x_lens:
+            A tensor of shape (batch_size,) containing the number of frames in
+            `x` before padding.
+        Returns:
+          Return a tuple containing 2 tensors:
+            - logits, its shape is (batch_size, output_seq_len, output_dim)
+            - logit_lens, a tensor of shape (batch_size,) containing the number
+              of frames in `logits` before padding.
+        """
+        if self.use_feat_batchnorm:
+            x = x.permute(0, 2, 1)  # (N, T, C) -> (N, C, T)
+            x = self.feat_batchnorm(x)
+            x = x.permute(0, 2, 1)  # (N, C, T) -> (N, T, C)
+
+        x = self.encoder_embed(x)
+        x = self.encoder_pos(x)
+        x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+
+        # Caution: We assume the subsampling factor is 4!
+        lengths = ((x_lens - 1) // 2 - 1) // 2
+        assert x.size(0) == lengths.max().item()
+
+        mask = make_pad_mask(lengths)
+        x = self.encoder(x, src_key_padding_mask=mask)  # (T, N, C)
+
+        logits = self.encoder_output_layer(x)
+        logits = logits.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+
+        return logits, lengths
+
+
+class TransformerEncoderLayer(nn.Module):
+    """
+    Modified from torch.nn.TransformerEncoderLayer.
+    Add support of normalize_before,
+    i.e., use layer_norm before the first block.
+
+    Args:
+      d_model:
+        the number of expected features in the input (required).
+      nhead:
+        the number of heads in the multiheadattention models (required).
+      dim_feedforward:
+        the dimension of the feedforward network model (default=2048).
+      dropout:
+        the dropout value (default=0.1).
+      activation:
+        the activation function of intermediate layer, relu or
+        gelu (default=relu).
+      normalize_before:
+        whether to use layer_norm before the first block.
+
+    Examples::
+        >>> encoder_layer = TransformerEncoderLayer(d_model=512, nhead=8)
+        >>> src = torch.rand(10, 32, 512)
+        >>> out = encoder_layer(src)
+    """
+
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        activation: str = "relu",
+        normalize_before: bool = True,
+    ) -> None:
+        super(TransformerEncoderLayer, self).__init__()
+        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=0.0)
+        # Implementation of Feedforward model
+        self.linear1 = nn.Linear(d_model, dim_feedforward)
+        self.dropout = nn.Dropout(dropout)
+        self.linear2 = nn.Linear(dim_feedforward, d_model)
+
+        self.norm1 = nn.LayerNorm(d_model)
+        self.norm2 = nn.LayerNorm(d_model)
+        self.dropout1 = nn.Dropout(dropout)
+        self.dropout2 = nn.Dropout(dropout)
+
+        self.activation = _get_activation_fn(activation)
+
+        self.normalize_before = normalize_before
+
+    def __setstate__(self, state):
+        if "activation" not in state:
+            state["activation"] = nn.functional.relu
+        super(TransformerEncoderLayer, self).__setstate__(state)
+
+    def forward(
+        self,
+        src: torch.Tensor,
+        src_mask: Optional[torch.Tensor] = None,
+        src_key_padding_mask: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        """
+        Pass the input through the encoder layer.
+
+        Args:
+            src: the sequence to the encoder layer (required).
+            src_mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional)
+
+        Shape:
+            src: (S, N, E).
+            src_mask: (S, S).
+            src_key_padding_mask: (N, S).
+            S is the source sequence length, T is the target sequence length,
+            N is the batch size, E is the feature number
+        """
+        residual = src
+        if self.normalize_before:
+            src = self.norm1(src)
+        src2 = self.self_attn(
+            src,
+            src,
+            src,
+            attn_mask=src_mask,
+            key_padding_mask=src_key_padding_mask,
+        )[0]
+        src = residual + self.dropout1(src2)
+        if not self.normalize_before:
+            src = self.norm1(src)
+
+        residual = src
+        if self.normalize_before:
+            src = self.norm2(src)
+        src2 = self.linear2(self.dropout(self.activation(self.linear1(src))))
+        src = residual + self.dropout2(src2)
+        if not self.normalize_before:
+            src = self.norm2(src)
+        return src
+
+
+def _get_activation_fn(activation: str):
+    if activation == "relu":
+        return nn.functional.relu
+    elif activation == "gelu":
+        return nn.functional.gelu
+
+    raise RuntimeError(
+        "activation should be relu/gelu, not {}".format(activation)
+    )
+
+
+class PositionalEncoding(nn.Module):
+    """This class implements the positional encoding
+    proposed in the following paper:
+
+    - Attention Is All You Need: https://arxiv.org/pdf/1706.03762.pdf
+
+        PE(pos, 2i) = sin(pos / (10000^(2i/d_modle))
+        PE(pos, 2i+1) = cos(pos / (10000^(2i/d_modle))
+
+    Note::
+
+      1 / (10000^(2i/d_model)) = exp(-log(10000^(2i/d_model)))
+                               = exp(-1* 2i / d_model * log(100000))
+                               = exp(2i * -(log(10000) / d_model))
+    """
+
+    def __init__(self, d_model: int, dropout: float = 0.1) -> None:
+        """
+        Args:
+          d_model:
+            Embedding dimension.
+          dropout:
+            Dropout probability to be applied to the output of this module.
+        """
+        super().__init__()
+        self.d_model = d_model
+        self.xscale = math.sqrt(self.d_model)
+        self.dropout = nn.Dropout(p=dropout)
+        # not doing: self.pe = None because of errors thrown by torchscript
+        self.pe = torch.zeros(1, 0, self.d_model, dtype=torch.float32)
+
+    def extend_pe(self, x: torch.Tensor) -> None:
+        """Extend the time t in the positional encoding if required.
+
+        The shape of `self.pe` is (1, T1, d_model). The shape of the input x
+        is (N, T, d_model). If T > T1, then we change the shape of self.pe
+        to (N, T, d_model). Otherwise, nothing is done.
+
+        Args:
+          x:
+            It is a tensor of shape (N, T, C).
+        Returns:
+          Return None.
+        """
+        if self.pe is not None:
+            if self.pe.size(1) >= x.size(1):
+                self.pe = self.pe.to(dtype=x.dtype, device=x.device)
+                return
+        pe = torch.zeros(x.size(1), self.d_model, dtype=torch.float32)
+        position = torch.arange(0, x.size(1), dtype=torch.float32).unsqueeze(1)
+        div_term = torch.exp(
+            torch.arange(0, self.d_model, 2, dtype=torch.float32)
+            * -(math.log(10000.0) / self.d_model)
+        )
+        pe[:, 0::2] = torch.sin(position * div_term)
+        pe[:, 1::2] = torch.cos(position * div_term)
+        pe = pe.unsqueeze(0)
+        # Now pe is of shape (1, T, d_model), where T is x.size(1)
+        self.pe = pe.to(device=x.device, dtype=x.dtype)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Add positional encoding.
+
+        Args:
+          x:
+            Its shape is (N, T, C)
+
+        Returns:
+          Return a tensor of shape (N, T, C)
+        """
+        self.extend_pe(x)
+        x = x * self.xscale + self.pe[:, : x.size(1), :]
+        return self.dropout(x)
+
+
+class Noam(object):
+    """
+    Implements Noam optimizer.
+
+    Proposed in
+    "Attention Is All You Need", https://arxiv.org/pdf/1706.03762.pdf
+
+    Modified from
+    https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/transformer/optimizer.py  # noqa
+
+    Args:
+      params:
+        iterable of parameters to optimize or dicts defining parameter groups
+      model_size:
+        attention dimension of the transformer model
+      factor:
+        learning rate factor
+      warm_step:
+        warmup steps
+    """
+
+    def __init__(
+        self,
+        params,
+        model_size: int = 256,
+        factor: float = 10.0,
+        warm_step: int = 25000,
+        weight_decay=0,
+    ) -> None:
+        """Construct an Noam object."""
+        self.optimizer = torch.optim.Adam(
+            params, lr=0, betas=(0.9, 0.98), eps=1e-9, weight_decay=weight_decay
+        )
+        self._step = 0
+        self.warmup = warm_step
+        self.factor = factor
+        self.model_size = model_size
+        self._rate = 0
+
+    @property
+    def param_groups(self):
+        """Return param_groups."""
+        return self.optimizer.param_groups
+
+    def step(self):
+        """Update parameters and rate."""
+        self._step += 1
+        rate = self.rate()
+        for p in self.optimizer.param_groups:
+            p["lr"] = rate
+        self._rate = rate
+        self.optimizer.step()
+
+    def rate(self, step=None):
+        """Implement `lrate` above."""
+        if step is None:
+            step = self._step
+        return (
+            self.factor
+            * self.model_size ** (-0.5)
+            * min(step ** (-0.5), step * self.warmup ** (-1.5))
+        )
+
+    def zero_grad(self):
+        """Reset gradient."""
+        self.optimizer.zero_grad()
+
+    def state_dict(self):
+        """Return state_dict."""
+        return {
+            "_step": self._step,
+            "warmup": self.warmup,
+            "factor": self.factor,
+            "model_size": self.model_size,
+            "_rate": self._rate,
+            "optimizer": self.optimizer.state_dict(),
+        }
+
+    def load_state_dict(self, state_dict):
+        """Load state_dict."""
+        for key, value in state_dict.items():
+            if key == "optimizer":
+                self.optimizer.load_state_dict(state_dict["optimizer"])
+            else:
+                setattr(self, key, value)

egs/yesno/ASR/transducer/beam_search.py

@@ -20,7 +20,7 @@ import torch
 from transducer.model import Transducer
 
 
-def greedy_search(model: Transducer, encoder_out: torch.Tensor) -> List[int]:
+def greedy_search(model: Transducer, encoder_out: torch.Tensor) -> List[str]:
     """
     Args:
       model:
@@ -42,7 +42,7 @@ def greedy_search(model: Transducer, encoder_out: torch.Tensor) -> List[int]:
     T = encoder_out.size(1)
     t = 0
     hyp = []
-    max_u = 1000  # terminte after this number of steps
+    max_u = 1000  # terminate after this number of steps
     u = 0
 
     while t < T and u < max_u: