Whisper Fine-tuning Recipe on Aishell1 (#1466)

* add decode seamlessm4t

* add requirements

* add decoding with avg model

* add token files

* add custom tokenizer

* support deepspeed to finetune large model

* support large-v3

* add model saving

* using monkey patch to replace models

* add manifest dir option

egs/aishell/ASR/README.md

@@ -24,3 +24,10 @@ The following table lists the differences among them.
 The decoder in `transducer_stateless` is modified from the paper
 [Rnn-Transducer with Stateless Prediction Network](https://ieeexplore.ieee.org/document/9054419/).
 We place an additional Conv1d layer right after the input embedding layer.
+
+# Whisper
+
+Recipe to finetune large pretrained models
+|                                    | Encoder   | Decoder            | Comment                                                                           |
+|------------------------------------|-----------|--------------------|-----------------------------------------------------------------------------------|
+| `whisper`             | Transformer | Transformer | support fine-tuning using deepspeed

egs/aishell/ASR/RESULTS.md

@@ -1,5 +1,63 @@
 ## Results
 
+### Aishell training results (Fine-tuning Pretrained Models)
+#### Whisper
+[./whisper](./whisper)
+##### fine-tuning results on Aishell test set on whisper medium, large-v2, large-v3
+
+|                        | test (before fine-tuning) | test (after fine-tuning)  | comment                                 |
+|------------------------|------|------|-----------------------------------------|
+| medium         | 7.23 | 3.27 | --epoch 10 --avg 4,  ddp                         |
+| large-v2       | 6.56 | 2.47 | --epoch 10 --avg 6,  deepspeed zero stage1       |
+| large-v3       | 6.06 | 2.84 | --epoch 5 --avg 3,   deepspeed zero stage1       |
+
+Command for training is:
+```bash
+pip install -r whisper/requirements.txt
+
+./prepare.sh --stage 30 --stop_stage 30
+
+#fine-tuning with deepspeed zero stage 1
+torchrun --nproc-per-node 8 ./whisper/train.py \
+  --max-duration 200 \
+  --exp-dir whisper/exp_large_v2 \
+  --model-name large-v2 \
+  --deepspeed \
+  --deepspeed_config ./whisper/ds_config_zero1.json
+
+# fine-tuning with ddp
+torchrun --nproc-per-node 8 ./whisper/train.py \
+  --max-duration 200 \
+  --exp-dir whisper/exp_medium \
+  --base-lr 1e-5 \
+  --model-name medium
+```
+
+Command for decoding using fine-tuned models:
+```bash
+git lfs install
+git clone https://huggingface.co/yuekai/icefall_asr_aishell_whisper
+ln -s icefall_asr_aishell_whisper/exp_large_v2/epoch-10-avg6.pt whisper/exp_large_v2/epoch-999.pt
+
+python3 ./whisper/decode.py \
+  --exp-dir whisper/exp_large_v2 \
+  --model-name large-v2 \
+  --epoch 999 --avg 1 \
+  --beam-size 10 --max-duration 50
+```
+Command for decoding using pretrained models (before fine-tuning):
+```bash
+python3 ./whisper/decode.py \
+  --exp-dir whisper/exp_large_v2 \
+  --model-name large-v2 \
+  --epoch -1 --avg 1 \
+  --remove-whisper-encoder-input-length-restriction False \
+  --beam-size 10 --max-duration 50
+```
+Fine-tuned models, training logs, decoding logs, tensorboard and decoding results
+are available at
+<https://huggingface.co/yuekai/icefall_asr_aishell_whisper>
+
 ### Aishell training result (Stateless Transducer)
 
 #### Zipformer (Byte-level BPE)
@@ -755,7 +813,6 @@ python3 ./transducer_stateless/decode.py \
        --max-sym-per-frame 3
 ```
 
-### Aishell training results (Transducer-stateless)
 #### 2022-02-18
 (Pingfeng Luo) : The tensorboard log for training is available at <https://tensorboard.dev/experiment/k3QL6QMhRbCwCKYKM9po9w/>
 And pretrained model is available at <https://huggingface.co/pfluo/icefall-aishell-transducer-stateless-char-2021-12-29>

egs/aishell/ASR/local/compute_fbank_aishell.py

@@ -29,7 +29,14 @@ import os
 from pathlib import Path
 
 import torch
-from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter
+from lhotse import (
+    CutSet,
+    Fbank,
+    FbankConfig,
+    LilcomChunkyWriter,
+    WhisperFbank,
+    WhisperFbankConfig,
+)
 from lhotse.recipes.utils import read_manifests_if_cached
 
 from icefall.utils import get_executor, str2bool
@@ -42,9 +49,14 @@ torch.set_num_threads(1)
 torch.set_num_interop_threads(1)
 
 
-def compute_fbank_aishell(num_mel_bins: int = 80, perturb_speed: bool = False):
+def compute_fbank_aishell(
+    num_mel_bins: int = 80,
+    perturb_speed: bool = False,
+    whisper_fbank: bool = False,
+    output_dir: str = "data/fbank",
+):
     src_dir = Path("data/manifests")
-    output_dir = Path("data/fbank")
+    output_dir = Path(output_dir)
     num_jobs = min(15, os.cpu_count())
 
     dataset_parts = (
@@ -68,8 +80,12 @@ def compute_fbank_aishell(num_mel_bins: int = 80, perturb_speed: bool = False):
         list(manifests.keys()),
         dataset_parts,
     )
-
+    if whisper_fbank:
-    extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
+        extractor = WhisperFbank(
+            WhisperFbankConfig(num_filters=num_mel_bins, device="cuda")
+        )
+    else:
+        extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
 
     with get_executor() as ex:  # Initialize the executor only once.
         for partition, m in manifests.items():
@@ -82,7 +98,7 @@ def compute_fbank_aishell(num_mel_bins: int = 80, perturb_speed: bool = False):
                 supervisions=m["supervisions"],
             )
             if "train" in partition and perturb_speed:
-                logging.info(f"Doing speed perturb")
+                logging.info("Doing speed perturb")
                 cut_set = (
                     cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
                 )
@@ -111,6 +127,18 @@ def get_args():
         default=False,
         help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
     )
+    parser.add_argument(
+        "--whisper-fbank",
+        type=str2bool,
+        default=False,
+        help="Use WhisperFbank instead of Fbank. Default: False.",
+    )
+    parser.add_argument(
+        "--output-dir",
+        type=str,
+        default="data/fbank",
+        help="Output directory. Default: data/fbank.",
+    )
     return parser.parse_args()
 
 
@@ -121,5 +149,8 @@ if __name__ == "__main__":
 
     args = get_args()
     compute_fbank_aishell(
-        num_mel_bins=args.num_mel_bins, perturb_speed=args.perturb_speed
+        num_mel_bins=args.num_mel_bins,
+        perturb_speed=args.perturb_speed,
+        whisper_fbank=args.whisper_fbank,
+        output_dir=args.output_dir,
     )

egs/aishell/ASR/prepare.sh

@@ -376,3 +376,16 @@ if [ $stage -le 12 ] && [ $stop_stage -ge 12 ]; then
     --vocab-size 4336 \
     --master-port 12345
 fi
+
+# whisper large-v3 using 128 mel bins, others using 80 mel bins
+whisper_mel_bins=80
+output_dir=data/fbank_whisper
+if [ $stage -le 30 ] && [ $stop_stage -ge 30 ]; then
+  log "Stage 30: Compute ${whisper_mel_bins} dim fbank for whisper model fine-tuning"
+  if [ ! -f $output_dir/.aishell.whisper.done ]; then
+    mkdir -p $output_dir
+    ./local/compute_fbank_aishell.py --perturb-speed ${perturb_speed} --num-mel-bins ${whisper_mel_bins} --whisper-fbank true --output-dir $output_dir
+    ./local/compute_fbank_musan.py --num-mel-bins ${whisper_mel_bins} --whisper-fbank true --output-dir $output_dir
+    touch $output_dir/.aishell.whisper.done
+  fi
+fi

egs/aishell/ASR/whisper/asr_datamodule.py

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

egs/aishell/ASR/whisper/decode.py

@@ -0,0 +1,503 @@
+#!/usr/bin/env python3
+# Copyright 2021 Xiaomi Corporation (Author: Liyong Guo,
+#                                            Fangjun Kuang,
+#                                            Wei Kang)
+#           2024 Yuekai Zhang
+#
+# 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:
+# Command for decoding using fine-tuned models:
+git lfs install
+git clone https://huggingface.co/yuekai/icefall_asr_aishell_whisper
+ln -s icefall_asr_aishell_whisper/exp_large_v2/epoch-10-avg6.pt whisper/exp_large_v2/epoch-999.pt
+
+python3 ./whisper/decode.py \
+  --exp-dir whisper/exp_large_v2 \
+  --model-name large-v2 \
+  --epoch 999 --avg 1 \
+  --manifest-dir data/fbank_whisper \
+  --beam-size 10 --max-duration 50
+
+# Command for decoding using pretrained models (before fine-tuning):
+
+python3 ./whisper/decode.py \
+  --exp-dir whisper/exp_large_v2 \
+  --model-name large-v2 \
+  --epoch -1 --avg 1 \
+  --manifest-dir data/fbank_whisper \
+  --remove-whisper-encoder-input-length-restriction False \
+  --beam-size 10 --max-duration 50
+
+"""
+
+import argparse
+import logging
+import re
+from collections import defaultdict
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple
+
+import k2
+import torch
+import torch.nn as nn
+import whisper
+from asr_datamodule import AishellAsrDataModule
+from tn.chinese.normalizer import Normalizer
+from whisper.normalizers import BasicTextNormalizer
+from whisper_encoder_forward_monkey_patch import replace_whisper_encoder_forward
+from zhconv import convert
+
+from icefall.checkpoint import average_checkpoints_with_averaged_model, load_checkpoint
+from icefall.env import get_env_info
+from icefall.utils import (
+    AttributeDict,
+    setup_logger,
+    store_transcripts,
+    str2bool,
+    write_error_stats,
+)
+
+
+def average_checkpoints(
+    filenames: List[Path], device: torch.device = torch.device("cpu")
+) -> dict:
+    """Average a list of checkpoints.
+    The function is mainly used for deepspeed converted checkpoint averaging, which only include model state_dict.
+
+    Args:
+      filenames:
+        Filenames of the checkpoints to be averaged. We assume all
+        checkpoints are saved by :func:`save_checkpoint`.
+      device:
+        Move checkpoints to this device before averaging.
+    Returns:
+      Return a dict (i.e., state_dict) which is the average of all
+      model state dicts contained in the checkpoints.
+    """
+    n = len(filenames)
+
+    if "model" in torch.load(filenames[0], map_location=device):
+        avg = torch.load(filenames[0], map_location=device)["model"]
+    else:
+        avg = torch.load(filenames[0], map_location=device)
+
+    # Identify shared parameters. Two parameters are said to be shared
+    # if they have the same data_ptr
+    uniqued: Dict[int, str] = dict()
+
+    for k, v in avg.items():
+        v_data_ptr = v.data_ptr()
+        if v_data_ptr in uniqued:
+            continue
+        uniqued[v_data_ptr] = k
+
+    uniqued_names = list(uniqued.values())
+
+    for i in range(1, n):
+        if "model" in torch.load(filenames[i], map_location=device):
+            state_dict = torch.load(filenames[i], map_location=device)["model"]
+        else:
+            state_dict = torch.load(filenames[i], map_location=device)
+        for k in uniqued_names:
+            avg[k] += state_dict[k]
+
+    for k in uniqued_names:
+        if avg[k].is_floating_point():
+            avg[k] /= n
+        else:
+            avg[k] //= n
+
+    return avg
+
+
+def remove_punctuation(text: str or List[str]):
+    """Modified from https://github.com/yeyupiaoling/Whisper-Finetune/blob/master/utils/data_utils.py
+
+    Args:
+        text: It can be a string or a list of strings.
+    Returns:
+        Return a string or a list of strings without any punctuation.
+    """
+    punctuation = "!,.;:?、！，。；：？《》 "
+    if isinstance(text, str):
+        text = re.sub(r"[{}]+".format(punctuation), "", text).strip()
+        return text
+    elif isinstance(text, list):
+        result_text = []
+        for t in text:
+            t = re.sub(r"[{}]+".format(punctuation), "", t).strip()
+            result_text.append(t)
+        return result_text
+    else:
+        raise Exception(f"Not support type {type(text)}")
+
+
+def to_simple(text: str or List[str]):
+    """Convert traditional Chinese to simplified Chinese.
+    Args:
+        text: It can be a string or a list of strings.
+    Returns:
+        Return a string or a list of strings converted to simplified Chinese.
+    """
+    if isinstance(text, str):
+        text = convert(text, "zh-cn")
+        return text
+    elif isinstance(text, list):
+        result_text = []
+        for t in text:
+            t = convert(t, "zh-cn")
+            result_text.append(t)
+        return result_text
+    else:
+        raise Exception(f"Not support type{type(text)}")
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=-1,
+        help="It specifies the checkpoint to use for decoding."
+        "Note: Epoch counts from 0.",
+    )
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=1,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch'. ",
+    )
+
+    parser.add_argument(
+        "--method",
+        type=str,
+        default="beam-search",
+        help="""Decoding method.
+        Supported values are:
+          - beam-search
+        """,
+    )
+
+    parser.add_argument(
+        "--beam-size",
+        type=int,
+        default=1,
+        help="beam size for beam search decoding",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="whisper/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--model-name",
+        type=str,
+        default="large-v2",
+        choices=["large-v2", "large-v3", "medium", "small", "tiny"],
+        help="""The model name to use.
+        """,
+    )
+
+    parser.add_argument(
+        "--remove-whisper-encoder-input-length-restriction",
+        type=str2bool,
+        default=True,
+        help="replace whisper encoder forward method to remove input length restriction",
+    )
+
+    return parser
+
+
+def get_params() -> AttributeDict:
+    params = AttributeDict(
+        {
+            "env_info": get_env_info(),
+        }
+    )
+    return params
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    batch: dict,
+) -> Dict[str, List[List[int]]]:
+    """Decode one batch and return the result in a dict. The dict has the
+    following format:
+
+        - key: "beam-search"
+        - value: A list of lists. Each sublist is a list of token IDs.
+    Args:
+        params:
+            It is returned by :func:`get_params`.
+        model:
+            The neural model.
+        batch:
+            It is returned by :meth:`torch.utils.data.DataLoader.__iter__`.
+    Returns:
+        Return a dict, whose key may be "beam-search".
+    """
+    dtype = torch.float16
+    device = torch.device("cuda")
+
+    feature = batch["inputs"]
+    assert feature.ndim == 3
+    feature = feature.to(device, dtype=dtype).transpose(1, 2)
+    if not params.remove_whisper_encoder_input_length_restriction:
+        T = 3000
+        if feature.shape[2] < T:
+            feature = torch.cat(
+                [
+                    feature,
+                    torch.zeros(
+                        feature.shape[0], feature.shape[1], T - feature.shape[2]
+                    ).to(device, dtype=dtype),
+                ],
+                2,
+            )
+
+    supervisions = batch["supervisions"]
+    feature_len = supervisions["num_frames"]
+    feature_len = feature_len.to(device, dtype=dtype)
+    results = model.decode(feature, params.decoding_options)
+    hyps = [result.text for result in results]
+
+    hyps = remove_punctuation(hyps)
+    hyps = to_simple(hyps)
+    hyps = [params.normalizer.normalize(hyp) for hyp in hyps]
+
+    return {"beam-search": hyps}
+
+
+def decode_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+) -> Dict[str, List[Tuple[str, List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+        dl:
+            The dataloader.
+        params:
+            It is returned by :func:`get_params`.
+        model:
+            The neural model.
+    Returns:
+        Return a dict, whose key may be "beam-search".
+    """
+    results = []
+
+    num_cuts = 0
+
+    try:
+        num_batches = len(dl)
+    except TypeError:
+        num_batches = "?"
+
+    results = defaultdict(list)
+    for batch_idx, batch in enumerate(dl):
+        texts = batch["supervisions"]["text"]
+        cut_ids = [cut.id for cut in batch["supervisions"]["cut"]]
+
+        hyps_dict = decode_one_batch(
+            params=params,
+            model=model,
+            batch=batch,
+        )
+
+        for lm_scale, hyps in hyps_dict.items():
+            this_batch = []
+            assert len(hyps) == len(texts)
+            for cut_id, hyp_words, ref_text in zip(cut_ids, hyps, texts):
+                ref_words = ref_text.split()
+                this_batch.append((cut_id, ref_words, hyp_words))
+
+            results[lm_scale].extend(this_batch)
+
+        num_cuts += len(batch["supervisions"]["text"])
+
+        if batch_idx % 100 == 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[str, List[str], List[str]]]],
+):
+
+    enable_log = True
+    test_set_wers = dict()
+    for key, results in results_dict.items():
+        recog_path = (
+            params.exp_dir / f"recogs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        results = sorted(results)
+        store_transcripts(filename=recog_path, texts=results)
+        if enable_log:
+            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.exp_dir / f"errs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        # we compute CER for aishell dataset.
+        results_char = []
+        for res in results:
+            results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
+        with open(errs_filename, "w") as f:
+            wer = write_error_stats(
+                f, f"{test_set_name}-{key}", results_char, enable_log=enable_log
+            )
+            test_set_wers[key] = wer
+
+        if enable_log:
+            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.exp_dir / f"cer-summary-{test_set_name}-{params.suffix}.txt"
+    with open(errs_info, "w") as f:
+        print("settings\tCER", file=f)
+        for key, val in test_set_wers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, CER 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()
+    AishellAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+    params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
+    setup_logger(
+        f"{params.exp_dir}/log-{params.method}-beam{params.beam_size}/log-decode-{params.suffix}"
+    )
+
+    options = whisper.DecodingOptions(
+        task="transcribe",
+        language="zh",
+        without_timestamps=True,
+        beam_size=params.beam_size,
+    )
+    params.decoding_options = options
+    params.cleaner = BasicTextNormalizer()
+    params.normalizer = Normalizer()
+
+    logging.info("Decoding started")
+    logging.info(params)
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda")
+
+    logging.info(f"device: {device}")
+
+    if params.remove_whisper_encoder_input_length_restriction:
+        replace_whisper_encoder_forward()
+    model = whisper.load_model(params.model_name, "cpu")
+    if params.epoch > 0:
+        if params.avg > 1:
+            start = params.epoch - params.avg
+            assert start >= 1, start
+            checkpoint = torch.load(
+                f"{params.exp_dir}/epoch-{params.epoch}.pt", map_location="cpu"
+            )
+            if "model" not in checkpoint:
+                # deepspeed converted checkpoint only contains model state_dict
+                filenames = [
+                    f"{params.exp_dir}/epoch-{epoch}.pt"
+                    for epoch in range(start, params.epoch + 1)
+                ]
+                model.load_state_dict(average_checkpoints(filenames))
+            else:
+                filename_start = f"{params.exp_dir}/epoch-{start}.pt"
+                filename_end = f"{params.exp_dir}/epoch-{params.epoch}.pt"
+                logging.info(
+                    f"Calculating the averaged model over epoch range from "
+                    f"{start} (excluded) to {params.epoch}"
+                )
+                model.to(device)
+                model.load_state_dict(
+                    average_checkpoints_with_averaged_model(
+                        filename_start=filename_start,
+                        filename_end=filename_end,
+                        device=device,
+                    )
+                )
+            # save checkpoints
+            filename = f"{params.exp_dir}/epoch-{params.epoch}-avg-{params.avg}.pt"
+            torch.save(model.state_dict(), filename)
+        else:
+            checkpoint = torch.load(
+                f"{params.exp_dir}/epoch-{params.epoch}.pt", map_location="cpu"
+            )
+            if "model" not in checkpoint:
+                model.load_state_dict(checkpoint, strict=True)
+            else:
+                load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+    model.to(device)
+    model.eval()
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    # we need cut ids to display recognition results.
+    args.return_cuts = True
+    aishell = AishellAsrDataModule(args)
+    valid_dl = aishell.valid_dataloaders(aishell.valid_cuts())
+    test_dl = aishell.test_dataloaders(aishell.test_cuts())
+    test_sets = ["valid", "test"]
+    test_dls = [valid_dl, test_dl]
+
+    for test_set, test_dl in zip(test_sets, test_dls):
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+        )
+
+        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/aishell/ASR/whisper/ds_config_zero1.json

@@ -0,0 +1,38 @@
+{
+    "fp16": {
+        "enabled": true,
+        "loss_scale": 0,
+        "loss_scale_window": 100,
+        "initial_scale_power": 16,
+        "hysteresis": 2,
+        "min_loss_scale": 0.01
+    },
+    "zero_optimization": {
+        "stage": 1,
+        "allgather_partitions": true,
+        "allgather_bucket_size": 2e8,
+        "overlap_comm": true,
+        "reduce_scatter": true,
+        "reduce_bucket_size": 2e8,
+        "contiguous_gradients": true
+    },
+    "optimizer": {
+        "type": "Adam",
+        "params": {
+            "lr": 1e-5
+        }
+    },
+    "scheduler": {
+        "type": "WarmupLR",
+        "params": {
+            "warmup_min_lr": 0,
+            "warmup_max_lr": 1e-5,
+            "warmup_num_steps": 100
+        }
+    },
+    "gradient_accumulation_steps": 1,
+    "gradient_clipping": 5,
+    "steps_per_print": 50,
+    "train_micro_batch_size_per_gpu": 1,
+    "wall_clock_breakdown": false
+}

egs/aishell/ASR/whisper/label_smoothing.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/conformer_ctc/label_smoothing.py

egs/aishell/ASR/whisper/optim.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/optim.py

egs/aishell/ASR/whisper/requirements.txt

@@ -0,0 +1,10 @@
+k2
+kaldialign
+git+https://github.com/lhotse-speech/lhotse
+sentencepiece
+tensorboard
+librosa
+git+https://github.com/yuekaizhang/whisper.git
+zhconv
+WeTextProcessing
+deepspeed

egs/aishell/ASR/whisper/train.py

@@ -0,0 +1,927 @@
+#!/usr/bin/env python3
+# Copyright    2023  Xiaomi Corp.        (authors: Xiaoyu Yang)
+#              2024  Yuekai Zhang
+#
+# 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:
+
+#fine-tuning with deepspeed zero stage 1
+torchrun --nproc-per-node 8 ./whisper/train.py \
+  --max-duration 200 \
+  --exp-dir whisper/exp_large_v2 \
+  --model-name large-v2 \
+  --manifest-dir data/fbank_whisper \
+  --deepspeed \
+  --deepspeed_config ./whisper/ds_config_zero1.json
+
+# fine-tuning with ddp
+torchrun --nproc-per-node 8 ./whisper/train.py \
+  --max-duration 200 \
+  --exp-dir whisper/exp_medium \
+  --manifest-dir data/fbank_whisper \
+  --base-lr 1e-5 \
+  --model-name medium
+"""
+
+
+import argparse
+import copy
+import logging
+import random
+import warnings
+from pathlib import Path
+from shutil import copyfile
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import deepspeed
+import k2
+import optim
+import torch
+import torch.multiprocessing as mp
+import torch.nn as nn
+import whisper
+from asr_datamodule import AishellAsrDataModule
+from deepspeed.utils.zero_to_fp32 import convert_zero_checkpoint_to_fp32_state_dict
+from label_smoothing import LabelSmoothingLoss
+from lhotse import CutSet, load_manifest
+from lhotse.cut import Cut
+from lhotse.dataset.sampling.base import CutSampler
+from lhotse.utils import fix_random_seed
+from optim import Eden, ScaledAdam
+from torch import Tensor
+from torch.cuda.amp import GradScaler
+from torch.nn.functional import pad as pad_tensor
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.utils.tensorboard import SummaryWriter
+from whisper_encoder_forward_monkey_patch import replace_whisper_encoder_forward
+
+from icefall import diagnostics
+from icefall.checkpoint import load_checkpoint, remove_checkpoints
+from icefall.checkpoint import save_checkpoint as save_checkpoint_impl
+from icefall.checkpoint import update_averaged_model
+from icefall.dist import cleanup_dist, get_rank, get_world_size, setup_dist
+from icefall.env import get_env_info
+from icefall.hooks import register_inf_check_hooks
+from icefall.utils import (
+    AttributeDict,
+    MetricsTracker,
+    filter_uneven_sized_batch,
+    setup_logger,
+    str2bool,
+)
+
+LRSchedulerType = Union[torch.optim.lr_scheduler._LRScheduler, optim.LRScheduler]
+
+
+def set_batch_count(model: Union[nn.Module, DDP], batch_count: float) -> None:
+    if isinstance(model, DDP):
+        # get underlying nn.Module
+        model = model.module
+    for module in model.modules():
+        if hasattr(module, "batch_count"):
+            module.batch_count = batch_count
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--tensorboard",
+        type=str2bool,
+        default=True,
+        help="Should various information be logged in tensorboard.",
+    )
+
+    parser.add_argument(
+        "--num-epochs",
+        type=int,
+        default=10,
+        help="Number of epochs to train.",
+    )
+
+    parser.add_argument(
+        "--start-epoch",
+        type=int,
+        default=1,
+        help="""Resume training from this epoch. It should be positive.
+        If larger than 1, it will load checkpoint from
+        exp-dir/epoch-{start_epoch-1}.pt
+        """,
+    )
+
+    parser.add_argument(
+        "--start-batch",
+        type=int,
+        default=0,
+        help="""If positive, --start-epoch is ignored and
+        it loads the checkpoint from exp-dir/checkpoint-{start_batch}.pt
+        """,
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="pruned_transducer_stateless7/exp",
+        help="""The experiment dir.
+        It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--model-name",
+        type=str,
+        default="large-v2",
+        choices=["large-v2", "large-v3", "medium", "small", "tiny"],
+        help="""The model name to use.
+        """,
+    )
+
+    parser.add_argument(
+        "--base-lr", type=float, default=1e-5, help="The base learning rate."
+    )
+
+    parser.add_argument(
+        "--lr-batches",
+        type=float,
+        default=5000,
+        help="""Number of steps that affects how rapidly the learning rate
+        decreases. We suggest not to change this.""",
+    )
+
+    parser.add_argument(
+        "--lr-epochs",
+        type=float,
+        default=6,
+        help="""Number of epochs that affects how rapidly the learning rate decreases.
+        """,
+    )
+
+    parser.add_argument(
+        "--seed",
+        type=int,
+        default=42,
+        help="The seed for random generators intended for reproducibility",
+    )
+
+    parser.add_argument(
+        "--print-diagnostics",
+        type=str2bool,
+        default=False,
+        help="Accumulate stats on activations, print them and exit.",
+    )
+
+    parser.add_argument(
+        "--inf-check",
+        type=str2bool,
+        default=False,
+        help="Add hooks to check for infinite module outputs and gradients.",
+    )
+
+    parser.add_argument(
+        "--keep-last-k",
+        type=int,
+        default=30,
+        help="""Only keep this number of checkpoints on disk.
+        For instance, if it is 3, there are only 3 checkpoints
+        in the exp-dir with filenames `checkpoint-xxx.pt`.
+        It does not affect checkpoints with name `epoch-xxx.pt`.
+        """,
+    )
+
+    parser.add_argument(
+        "--average-period",
+        type=int,
+        default=200,
+        help="""Update the averaged model, namely `model_avg`, after processing
+        this number of batches. `model_avg` is a separate version of model,
+        in which each floating-point parameter is the average of all the
+        parameters from the start of training. Each time we take the average,
+        we do: `model_avg = model * (average_period / batch_idx_train) +
+            model_avg * ((batch_idx_train - average_period) / batch_idx_train)`.
+        """,
+    )
+
+    parser.add_argument(
+        "--use-fp16",
+        type=str2bool,
+        default=True,
+        help="Whether to use half precision training.",
+    )
+
+    parser = deepspeed.add_config_arguments(parser)
+
+    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`:
+
+        - frame_shift_ms: The frame shift in milliseconds.
+        - allowed_excess_duration_ratio: The allowed excess duration ratio.
+        - best_train_loss: The best training loss so far.
+        - best_valid_loss: The best validation loss so far.
+        - best_train_epoch: The epoch where the best training loss is achieved.
+        - best_valid_epoch: The epoch where the best validation loss is achieved.
+        - batch_idx_train: The batch index of the current batch.
+        - log_interval: Log training stats every `log_interval` batches.
+        - reset_interval: Reset the stats every `reset_interval` batches.
+        - valid_interval: Run validation every `valid_interval` batches.
+        - env_info: The environment information.
+    """
+    params = AttributeDict(
+        {
+            "frame_shift_ms": 10.0,
+            "subsampling_factor": 2,
+            "allowed_excess_duration_ratio": 0.1,
+            "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": 5000,
+            "env_info": get_env_info(),
+        }
+    )
+
+    return params
+
+
+def load_checkpoint_if_available(
+    params: AttributeDict,
+    model: nn.Module,
+    model_avg: nn.Module = None,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[LRSchedulerType] = None,
+) -> Optional[Dict[str, Any]]:
+    """Load checkpoint from file.
+
+    If params.start_batch is positive, it will load the checkpoint from
+    `params.exp_dir/checkpoint-{params.start_batch}.pt`. Otherwise, if
+    params.start_epoch is larger than 1, it will load the checkpoint from
+    `params.start_epoch - 1`.
+
+    Apart from loading state dict for `model` and `optimizer` 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.
+      model_avg:
+        The stored model averaged from the start of training.
+      optimizer:
+        The optimizer that we are using.
+      scheduler:
+        The scheduler that we are using.
+    Returns:
+      Return a dict containing previously saved training info.
+    """
+    if params.start_batch > 0:
+        filename = params.exp_dir / f"checkpoint-{params.start_batch}.pt"
+    elif params.start_epoch > 1:
+        filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
+    else:
+        return None
+
+    assert filename.is_file(), f"{filename} does not exist!"
+
+    saved_params = load_checkpoint(
+        filename,
+        model=model,
+        model_avg=model_avg,
+        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]
+
+    if params.start_batch > 0:
+        if "cur_epoch" in saved_params:
+            params["start_epoch"] = saved_params["cur_epoch"]
+
+    return saved_params
+
+
+def save_checkpoint(
+    params: AttributeDict,
+    model: Union[nn.Module, DDP],
+    model_avg: Optional[nn.Module] = None,
+    optimizer: Optional[torch.optim.Optimizer] = None,
+    scheduler: Optional[LRSchedulerType] = None,
+    sampler: Optional[CutSampler] = None,
+    scaler: Optional[GradScaler] = 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.
+      model_avg:
+        The stored model averaged from the start of training.
+      optimizer:
+        The optimizer used in the training.
+      sampler:
+       The sampler for the training dataset.
+      scaler:
+        The scaler used for mix precision training.
+    """
+    if rank != 0:
+        return
+    filename = params.exp_dir / f"epoch-{params.cur_epoch}.pt"
+    save_checkpoint_impl(
+        filename=filename,
+        model=model,
+        model_avg=model_avg,
+        params=params,
+        optimizer=optimizer,
+        scheduler=scheduler,
+        sampler=sampler,
+        scaler=scaler,
+        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,
+    tokenizer: whisper.tokenizer.Tokenizer,
+    model: Union[nn.Module, DDP],
+    batch: dict,
+    is_training: bool,
+) -> Tuple[Tensor, MetricsTracker]:
+    """
+    Compute the loss for the given batch.
+    Args:
+        params:
+            It is returned by :func:`get_params`.
+        tokenizer:
+            The tokenizer used to encode the text.
+        model:
+            The model for training.
+        batch:
+            A batch of data. See `lhotse.dataset.K2SpeechRecognitionDataset()`
+            for the content in it.
+        is_training:
+            Whether it is training.
+    Returns:
+        Return a tuple of two elements. The first element is the loss tensor.
+    """
+    # For the uneven-sized batch, the total duration after padding would possibly
+    # cause OOM. Hence, for each batch, which is sorted descendingly by length,
+    # we simply drop the last few shortest samples, so that the retained total frames
+    # (after padding) would not exceed `allowed_max_frames`:
+    # `allowed_max_frames = int(max_frames * (1.0 + allowed_excess_duration_ratio))`,
+    # where `max_frames = max_duration * 1000 // frame_shift_ms`.
+    # We set allowed_excess_duration_ratio=0.1.
+    if isinstance(model, DDP):
+        # get underlying nn.Module
+        model = model.module
+
+    def _batch_tensors(tensors: List[Tensor], pad_value: Any) -> Tensor:
+        padding_size = max(tensor.shape[0] for tensor in tensors)
+        dims = len(tensors[0].shape)
+        padded_tensors = []
+        for tensor in tensors:
+            padding = [0] * 2 * dims
+            padding[-1] = padding_size - tensor.shape[0]
+            padded_tensors.append(pad_tensor(tensor, padding, "constant", pad_value))
+        return torch.stack([tensor for tensor in padded_tensors], dim=0)
+
+    max_frames = params.max_duration * 1000 // params.frame_shift_ms
+    allowed_max_frames = int(max_frames * (1.0 + params.allowed_excess_duration_ratio))
+    batch = filter_uneven_sized_batch(batch, allowed_max_frames)
+
+    device = model.device if isinstance(model, DDP) else next(model.parameters()).device
+    feature = batch["inputs"]
+
+    assert feature.ndim == 3
+    feature = feature.to(device)
+    feature = feature.transpose(1, 2)  # (N, C, T)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    batch_idx_train = params.batch_idx_train
+
+    texts = batch["supervisions"]["text"]
+    # remove spaces in texts
+    texts = [text.replace(" ", "") for text in texts]
+
+    text_tokens_list = [
+        list(tokenizer.sot_sequence_including_notimestamps)
+        + tokenizer.encode(text)
+        + [tokenizer.eot]
+        for text in texts
+    ]
+    # convert it to torch tensor
+    text_tokens_list = [
+        torch.LongTensor(text_tokens) for text_tokens in text_tokens_list
+    ]
+
+    # 50256 is the index of <pad> for all whisper models
+    prev_outputs_tokens = _batch_tensors(
+        [tokens[:-1] for tokens in text_tokens_list], pad_value=50256
+    )
+    target_tokens = _batch_tensors(
+        [tokens[1:] for tokens in text_tokens_list], pad_value=50256
+    )
+    target_lengths = torch.LongTensor(
+        [tokens.shape[0] - 1 for tokens in text_tokens_list]
+    )
+
+    decoder_criterion = LabelSmoothingLoss(
+        ignore_index=50256, label_smoothing=0.1, reduction="sum"
+    )
+
+    # ignore the first 3 tokens, which are always <|lang_id|>, <|transcibe|>, <|notimestampes|>
+    ignore_prefix_size = 3
+    with torch.set_grad_enabled(is_training):
+        encoder_out = model.encoder(feature)
+        text_logits = model.decoder(prev_outputs_tokens.to(device), encoder_out)
+        text_logits = text_logits[:, ignore_prefix_size:, :]
+        target_tokens = target_tokens[:, ignore_prefix_size:]
+        loss = decoder_criterion(text_logits, target_tokens.to(device))
+
+    assert loss.requires_grad == is_training
+
+    info = MetricsTracker()
+    with warnings.catch_warnings():
+        warnings.simplefilter("ignore")
+        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,
+    tokenizer: whisper.tokenizer.Tokenizer,
+    model: Union[nn.Module, DDP],
+    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):
+        with torch.cuda.amp.autocast(enabled=params.use_fp16):
+            loss, loss_info = compute_loss(
+                params=params,
+                tokenizer=tokenizer,
+                model=model,
+                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,
+    tokenizer: whisper.tokenizer.Tokenizer,
+    model: Union[nn.Module, DDP],
+    optimizer: torch.optim.Optimizer,
+    scheduler: LRSchedulerType,
+    train_dl: torch.utils.data.DataLoader,
+    valid_dl: torch.utils.data.DataLoader,
+    scaler: GradScaler,
+    model_avg: Optional[nn.Module] = None,
+    tb_writer: Optional[SummaryWriter] = None,
+    world_size: int = 1,
+    rank: int = 0,
+) -> 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.
+      scheduler:
+        The learning rate scheduler, we call step() every step.
+      train_dl:
+        Dataloader for the training dataset.
+      valid_dl:
+        Dataloader for the validation dataset.
+      scaler:
+        The scaler used for mix precision training.
+      model_avg:
+        The stored model averaged from the start of training.
+      tb_writer:
+        Writer to write log messages to tensorboard.
+      world_size:
+        Number of nodes in DDP training. If it is 1, DDP is disabled.
+      rank:
+        The rank of the node in DDP training. If no DDP is used, it should
+        be set to 0.
+    """
+    model.train()
+
+    tot_loss = MetricsTracker()
+
+    for batch_idx, batch in enumerate(train_dl):
+        params.batch_idx_train += 1
+        batch_size = len(batch["supervisions"]["text"])
+        if batch_idx % params.valid_interval == 0 and not params.print_diagnostics:
+            logging.info("Computing validation loss")
+            valid_info = compute_validation_loss(
+                params=params,
+                tokenizer=tokenizer,
+                model=model,
+                valid_dl=valid_dl,
+                world_size=world_size,
+            )
+            model.train()
+            logging.info(f"Epoch {params.cur_epoch}, validation: {valid_info}")
+            logging.info(
+                f"Maximum memory allocated so far is {torch.cuda.max_memory_allocated()//1000000}MB"
+            )
+            if tb_writer is not None:
+                valid_info.write_summary(
+                    tb_writer, "train/valid_", params.batch_idx_train
+                )
+
+        try:
+            with torch.cuda.amp.autocast(enabled=params.use_fp16):
+                loss, loss_info = compute_loss(
+                    params=params,
+                    tokenizer=tokenizer,
+                    model=model,
+                    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.
+            if params.deepspeed:
+                # deepspeed's backward() is different from torch's backward()
+                # in that it does not accept a loss tensor as input.
+                # It computes the loss internally.
+                model.backward(loss)
+                model.step()
+            else:
+                scaler.scale(loss).backward()
+                set_batch_count(model, params.batch_idx_train)
+                scheduler.step_batch(params.batch_idx_train)
+
+                scaler.step(optimizer)
+                scaler.update()
+                optimizer.zero_grad()
+        except:  # noqa
+            display_and_save_batch(batch, params=params)
+            raise
+
+        if params.print_diagnostics and batch_idx == 5:
+            return
+
+        if (
+            rank == 0
+            and params.batch_idx_train > 0
+            and params.batch_idx_train % params.average_period == 0
+            and not params.deepspeed
+        ):
+            update_averaged_model(
+                params=params,
+                model_cur=model,
+                model_avg=model_avg,
+            )
+
+        if batch_idx % 100 == 0 and params.use_fp16 and not params.deepspeed:
+            # If the grad scale was less than 1, try increasing it.    The _growth_interval
+            # of the grad scaler is configurable, but we can't configure it to have different
+            # behavior depending on the current grad scale.
+            cur_grad_scale = scaler._scale.item()
+            if cur_grad_scale < 1.0 or (cur_grad_scale < 8.0 and batch_idx % 400 == 0):
+                scaler.update(cur_grad_scale * 2.0)
+            if cur_grad_scale < 0.01:
+                logging.warning(f"Grad scale is small: {cur_grad_scale}")
+            if cur_grad_scale < 1.0e-05:
+                raise RuntimeError(
+                    f"grad_scale is too small, exiting: {cur_grad_scale}"
+                )
+        if batch_idx % params.log_interval == 0:
+            try:
+                cur_lr = scheduler.get_last_lr()[0]
+            except:  # noqa
+                cur_lr = 0.0
+            cur_grad_scale = (
+                scaler._scale.item()
+                if (params.use_fp16 and not params.deepspeed)
+                else 1.0
+            )
+
+            logging.info(
+                f"Epoch {params.cur_epoch}, "
+                f"batch {batch_idx}, loss[{loss_info}], "
+                f"tot_loss[{tot_loss}], batch size: {batch_size}, "
+                f"lr: {cur_lr:.2e}, "
+                + (
+                    f"grad_scale: {scaler._scale.item()}"
+                    if (params.use_fp16 and not params.deepspeed)
+                    else ""
+                )
+            )
+
+            if tb_writer is not None:
+                tb_writer.add_scalar(
+                    "train/learning_rate", cur_lr, params.batch_idx_train
+                )
+
+                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 params.use_fp16:
+                    tb_writer.add_scalar(
+                        "train/grad_scale",
+                        cur_grad_scale,
+                        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))
+
+    fix_random_seed(params.seed)
+
+    setup_logger(f"{params.exp_dir}/log/log-train")
+    logging.info(params)
+
+    logging.info("About to create model")
+
+    replace_whisper_encoder_forward()
+    model = whisper.load_model(params.model_name, "cpu")
+    del model.alignment_heads
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    tokenizer = whisper.tokenizer.get_tokenizer(
+        model.is_multilingual,
+        num_languages=model.num_languages,
+        language="zh",
+        task="transcribe",
+    )
+
+    model_avg: Optional[nn.Module] = None
+    if rank == 0:
+        # model_avg is only used with rank 0
+        model_avg = copy.deepcopy(model).to(torch.float64)
+
+    assert params.start_epoch > 0, params.start_epoch
+    checkpoints = load_checkpoint_if_available(
+        params=params, model=model, model_avg=model_avg
+    )
+
+    if torch.cuda.is_available():
+        device = torch.device("cuda", rank)
+    else:
+        device = torch.device("cpu")
+    logging.info(f"Device: {device}")
+    model.to(device)
+
+    optimizer = torch.optim.AdamW(model.parameters(), lr=params.base_lr)
+    scheduler = Eden(optimizer, params.lr_batches, params.lr_epochs)
+
+    if checkpoints and "optimizer" in checkpoints:
+        logging.info("Loading optimizer state dict")
+        optimizer.load_state_dict(checkpoints["optimizer"])
+
+    if (
+        checkpoints
+        and "scheduler" in checkpoints
+        and checkpoints["scheduler"] is not None
+    ):
+        logging.info("Loading scheduler state dict")
+        scheduler.load_state_dict(checkpoints["scheduler"])
+
+    if world_size > 1:
+        if params.deepspeed:
+            logging.info("Using DeepSpeed")
+            model, optimizer, _, scheduler = deepspeed.initialize(
+                args=params, model=model, model_parameters=model.parameters()
+            )
+        else:
+            logging.info("Using DDP")
+            setup_dist(use_ddp_launch=True)
+            model = DDP(model, device_ids=[rank], find_unused_parameters=True)
+
+    if params.print_diagnostics:
+        opts = diagnostics.TensorDiagnosticOptions(
+            2**22
+        )  # allow 4 megabytes per sub-module
+        diagnostic = diagnostics.attach_diagnostics(model, opts)
+
+    if params.inf_check:
+        register_inf_check_hooks(model)
+
+    aishell = AishellAsrDataModule(args)
+
+    if params.start_batch > 0 and checkpoints and "sampler" in checkpoints:
+        # We only load the sampler's state dict when it loads a checkpoint
+        # saved in the middle of an epoch
+        sampler_state_dict = checkpoints["sampler"]
+    else:
+        sampler_state_dict = None
+
+    train_dl = aishell.train_dataloaders(aishell.train_cuts())
+    valid_dl = aishell.valid_dataloaders(aishell.valid_cuts())
+
+    scaler = GradScaler(enabled=params.use_fp16, init_scale=1.0)
+    if checkpoints and "grad_scaler" in checkpoints:
+        logging.info("Loading grad scaler state dict")
+        scaler.load_state_dict(checkpoints["grad_scaler"])
+
+    if args.tensorboard and rank == 0:
+        tb_writer = SummaryWriter(log_dir=f"{params.exp_dir}/tensorboard")
+    else:
+        tb_writer = None
+
+    logging.info(f"start training from epoch {params.start_epoch}")
+    for epoch in range(params.start_epoch, params.num_epochs + 1):
+        if not params.deepspeed:
+            scheduler.step_epoch(epoch - 1)
+        fix_random_seed(params.seed + epoch - 1)
+        train_dl.sampler.set_epoch(epoch - 1)
+
+        if tb_writer is not None:
+            tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)
+
+        params.cur_epoch = epoch
+
+        train_one_epoch(
+            params=params,
+            tokenizer=tokenizer,
+            model=model,
+            model_avg=model_avg,
+            optimizer=optimizer,
+            scheduler=scheduler,
+            train_dl=train_dl,
+            valid_dl=valid_dl,
+            scaler=scaler,
+            tb_writer=tb_writer,
+            world_size=world_size,
+            rank=rank,
+        )
+
+        if params.print_diagnostics:
+            diagnostic.print_diagnostics()
+            break
+
+        if params.deepspeed:
+            model.save_checkpoint(
+                save_dir=params.exp_dir,
+                tag=f"epoch-{params.cur_epoch}",
+                client_state={},
+            )
+            if rank == 0:
+                convert_zero_checkpoint_to_fp32_state_dict(
+                    params.exp_dir,
+                    f"{params.exp_dir}/epoch-{params.cur_epoch}.pt",
+                    tag=f"epoch-{params.cur_epoch}",
+                )
+        else:
+            save_checkpoint(
+                params=params,
+                model=model,
+                model_avg=model_avg,
+                optimizer=optimizer,
+                scheduler=scheduler,
+                sampler=train_dl.sampler,
+                scaler=scaler,
+                rank=rank,
+            )
+
+    logging.info("Done!")
+
+    if world_size > 1 and not params.deepspeed:
+        torch.distributed.barrier()
+        cleanup_dist()
+
+
+def display_and_save_batch(
+    batch: dict,
+    params: AttributeDict,
+) -> None:
+    """Display the batch statistics and save the batch into disk.
+
+    Args:
+      batch:
+        A batch of data. See `lhotse.dataset.K2SpeechRecognitionDataset()`
+        for the content in it.
+      params:
+        Parameters for training. See :func:`get_params`.
+    """
+    from lhotse.utils import uuid4
+
+    filename = f"{params.exp_dir}/batch-{uuid4()}.pt"
+    logging.info(f"Saving batch to {filename}")
+    torch.save(batch, filename)
+
+    supervisions = batch["supervisions"]
+    features = batch["inputs"]
+
+    logging.info(f"features shape: {features.shape}")
+
+
+def main():
+    parser = get_parser()
+    AishellAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    world_size = get_world_size()
+    rank = get_rank()
+
+    torch.set_num_threads(1)
+    torch.set_num_interop_threads(1)
+    run(rank=rank, world_size=world_size, args=args)
+
+
+if __name__ == "__main__":
+    main()

egs/aishell/ASR/whisper/whisper_encoder_forward_monkey_patch.py

@@ -0,0 +1,29 @@
+import torch
+import torch.nn.functional as F
+import whisper
+
+
+def forward(self, x: torch.Tensor):
+    """
+    x : torch.Tensor, shape = (batch_size, n_mels, n_ctx)
+        the mel spectrogram of the audio
+    """
+    x = F.gelu(self.conv1(x))
+    x = F.gelu(self.conv2(x))
+    x = x.permute(0, 2, 1)
+
+    x = (x + self.positional_embedding[: x.shape[1], :]).to(x.dtype)
+
+    for block in self.blocks:
+        x = block(x)
+
+    x = self.ln_post(x)
+    return x
+
+
+def replace_whisper_encoder_forward():
+    """
+    This function monkey patches the forward method of the whisper encoder.
+    To be called before the model is loaded, it changes whisper to process audio with any length < 30s.
+    """
+    whisper.model.AudioEncoder.forward = forward

egs/librispeech/ASR/local/compute_fbank_musan.py

@@ -22,16 +22,25 @@ It looks for manifests in the directory data/manifests.
 
 The generated fbank features are saved in data/fbank.
 """
-
+import argparse
 import logging
 import os
 from pathlib import Path
 
 import torch
-from lhotse import CutSet, Fbank, FbankConfig, LilcomChunkyWriter, MonoCut, combine
+from lhotse import (
+    CutSet,
+    Fbank,
+    FbankConfig,
+    LilcomChunkyWriter,
+    MonoCut,
+    WhisperFbank,
+    WhisperFbankConfig,
+    combine,
+)
 from lhotse.recipes.utils import read_manifests_if_cached
 
-from icefall.utils import get_executor
+from icefall.utils import get_executor, str2bool
 
 # Torch's multithreaded behavior needs to be disabled or
 # it wastes a lot of CPU and slow things down.
@@ -45,11 +54,12 @@ def is_cut_long(c: MonoCut) -> bool:
     return c.duration > 5
 
 
-def compute_fbank_musan():
+def compute_fbank_musan(
+    num_mel_bins: int = 80, whisper_fbank: bool = False, output_dir: str = "data/fbank"
+):
     src_dir = Path("data/manifests")
-    output_dir = Path("data/fbank")
+    output_dir = Path(output_dir)
     num_jobs = min(15, os.cpu_count())
-    num_mel_bins = 80
 
     dataset_parts = (
         "music",
@@ -81,7 +91,12 @@ def compute_fbank_musan():
 
     logging.info("Extracting features for Musan")
 
-    extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
+    if whisper_fbank:
+        extractor = WhisperFbank(
+            WhisperFbankConfig(num_filters=num_mel_bins, device="cuda")
+        )
+    else:
+        extractor = Fbank(FbankConfig(num_mel_bins=num_mel_bins))
 
     with get_executor() as ex:  # Initialize the executor only once.
         # create chunks of Musan with duration 5 - 10 seconds
@@ -102,8 +117,36 @@ def compute_fbank_musan():
         musan_cuts.to_file(musan_cuts_path)
 
 
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--num-mel-bins",
+        type=int,
+        default=80,
+        help="""The number of mel bins for Fbank""",
+    )
+    parser.add_argument(
+        "--whisper-fbank",
+        type=str2bool,
+        default=False,
+        help="Use WhisperFbank instead of Fbank. Default: False.",
+    )
+    parser.add_argument(
+        "--output-dir",
+        type=str,
+        default="data/fbank",
+        help="Output directory. Default: data/fbank.",
+    )
+    return parser.parse_args()
+
+
 if __name__ == "__main__":
     formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
 
     logging.basicConfig(format=formatter, level=logging.INFO)
-    compute_fbank_musan()
+    args = get_args()
+    compute_fbank_musan(
+        num_mel_bins=args.num_mel_bins,
+        whisper_fbank=args.whisper_fbank,
+        output_dir=args.output_dir,
+    )

icefall/dist.py

@@ -22,7 +22,7 @@ from torch import distributed as dist
 
 
 def setup_dist(
-    rank, world_size, master_port=None, use_ddp_launch=False, master_addr=None
+    rank=None, world_size=None, master_port=None, use_ddp_launch=False, master_addr=None
 ):
     """
     rank and world_size are used only if use_ddp_launch is False.