Modified beam search with RNNLM rescoring (#1002)

* add RNNLM rescore

* add shallow fusion and lm rescore for streaming zipformer

* minor fix

* update RESULTS.md

* fix yesno workflow, change from ubuntu-18.04 to ubuntu-latest

.github/workflows/run-yesno-recipe.yml

@@ -35,7 +35,7 @@ jobs:
       matrix:
         # os: [ubuntu-18.04, macos-10.15]
         # TODO: enable macOS for CPU testing
-        os: [ubuntu-18.04]
+        os: [ubuntu-latest]
         python-version: [3.8]
       fail-fast: false
 

egs/librispeech/ASR/RESULTS.md

@@ -76,6 +76,64 @@ for m in greedy_search modified_beam_search fast_beam_search; do
     --num-decode-streams 2000
 done
 ```
+We also support decoding with neural network LMs. After combining with language models, the WERs are
+| decoding method      | chunk size | test-clean | test-other | comment             | decoding mode        |
+|----------------------|------------|------------|------------|---------------------|----------------------|
+| modified beam search | 320ms      | 3.11       | 7.93       | --epoch 30 --avg 9  | simulated streaming  |
+| modified beam search + RNNLM shallow fusion | 320ms      | 2.58       | 6.65       | --epoch 30 --avg 9  | simulated streaming  |
+| modified beam search + RNNLM nbest rescore | 320ms      | 2.59       | 6.86       | --epoch 30 --avg 9  | simulated streaming  |
+
+Please use the following command for RNNLM shallow fusion:
+```bash
+for lm_scale in $(seq 0.15 0.01 0.38); do
+    for beam_size in 4 8 12; do
+        ./pruned_transducer_stateless7_streaming/decode.py \
+            --epoch 99 \
+            --avg 1 \
+            --use-averaged-model False \
+            --beam-size $beam_size \
+            --exp-dir ./pruned_transducer_stateless7_streaming/exp-large-LM \
+            --max-duration 600 \
+            --decode-chunk-len 32 \
+            --decoding-method modified_beam_search_lm_shallow_fusion \
+            --use-shallow-fusion 1 \
+            --lm-type rnn \
+            --lm-exp-dir rnn_lm/exp \
+            --lm-epoch 99 \
+            --lm-scale $lm_scale \
+            --lm-avg 1 \
+            --rnn-lm-embedding-dim 2048 \
+            --rnn-lm-hidden-dim 2048 \
+            --rnn-lm-num-layers 3 \
+            --lm-vocab-size 500
+    done
+done
+```
+
+Please use the following command for RNNLM rescore:
+```bash
+./pruned_transducer_stateless7_streaming/decode.py \
+    --epoch 30 \
+    --avg 9 \
+    --use-averaged-model True \
+    --beam-size 8 \
+    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
+    --max-duration 600 \
+    --decode-chunk-len 32 \
+    --decoding-method modified_beam_search_lm_rescore \
+    --use-shallow-fusion 0 \
+    --lm-type rnn \
+    --lm-exp-dir rnn_lm/exp \
+    --lm-epoch 99 \
+    --lm-avg 1 \
+    --rnn-lm-embedding-dim 2048 \
+    --rnn-lm-hidden-dim 2048 \
+    --rnn-lm-num-layers 3 \
+    --lm-vocab-size 500
+```
+
+A well-trained RNNLM can be found here: <https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm/tree/main>.
+
 
 #### Smaller model
 
@@ -540,9 +598,9 @@ for m in greedy_search fast_beam_search modified_beam_search ; do
 done
 ```
 
-Note that a small change is made to the `pruned_transducer_stateless7/decoder.py` in 
-this [PR](/ceph-data4/yangxiaoyu/softwares/icefall_development/icefall_random_padding/egs/librispeech/ASR/pruned_transducer_stateless7/exp_960h_no_paddingidx_ngpu4/tensorboard) to address the 
-problem of emitting the first symbol at the very beginning. If you need a 
+Note that a small change is made to the `pruned_transducer_stateless7/decoder.py` in
+this [PR](/ceph-data4/yangxiaoyu/softwares/icefall_development/icefall_random_padding/egs/librispeech/ASR/pruned_transducer_stateless7/exp_960h_no_paddingidx_ngpu4/tensorboard) to address the
+problem of emitting the first symbol at the very beginning. If you need a
 model without this issue, please download the model from here: <https://huggingface.co/marcoyang/icefall-asr-librispeech-pruned-transducer-stateless7-2023-03-10>
 
 ### LibriSpeech BPE training results (Pruned Stateless LSTM RNN-T + gradient filter)

egs/librispeech/ASR/lstm_transducer_stateless2/decode.py

@@ -925,7 +925,6 @@ def main():
         )
         LM.to(device)
         LM.eval()
-
     else:
         LM = None
 

egs/librispeech/ASR/pruned_transducer_stateless2/beam_search.py

@@ -1059,6 +1059,204 @@ def modified_beam_search(
         )
 
 
+def modified_beam_search_lm_rescore(
+    model: Transducer,
+    encoder_out: torch.Tensor,
+    encoder_out_lens: torch.Tensor,
+    LM: LmScorer,
+    lm_scale_list: List[int],
+    beam: int = 4,
+    temperature: float = 1.0,
+    return_timestamps: bool = False,
+) -> Union[List[List[int]], DecodingResults]:
+    """Beam search in batch mode with --max-sym-per-frame=1 being hardcoded.
+    Rescore the final results with RNNLM and return the one with the highest score
+
+    Args:
+      model:
+        The transducer model.
+      encoder_out:
+        Output from the encoder. Its shape is (N, T, C).
+      encoder_out_lens:
+        A 1-D tensor of shape (N,), containing number of valid frames in
+        encoder_out before padding.
+      beam:
+        Number of active paths during the beam search.
+      temperature:
+        Softmax temperature.
+      LM:
+        A neural network language model
+      return_timestamps:
+        Whether to return timestamps.
+    Returns:
+      If return_timestamps is False, return the decoded result.
+      Else, return a DecodingResults object containing
+      decoded result and corresponding timestamps.
+    """
+    assert encoder_out.ndim == 3, encoder_out.shape
+    assert encoder_out.size(0) >= 1, encoder_out.size(0)
+
+    packed_encoder_out = torch.nn.utils.rnn.pack_padded_sequence(
+        input=encoder_out,
+        lengths=encoder_out_lens.cpu(),
+        batch_first=True,
+        enforce_sorted=False,
+    )
+
+    blank_id = model.decoder.blank_id
+    unk_id = getattr(model, "unk_id", blank_id)
+    context_size = model.decoder.context_size
+    device = next(model.parameters()).device
+
+    batch_size_list = packed_encoder_out.batch_sizes.tolist()
+    N = encoder_out.size(0)
+    assert torch.all(encoder_out_lens > 0), encoder_out_lens
+    assert N == batch_size_list[0], (N, batch_size_list)
+
+    B = [HypothesisList() for _ in range(N)]
+    for i in range(N):
+        B[i].add(
+            Hypothesis(
+                ys=[blank_id] * context_size,
+                log_prob=torch.zeros(1, dtype=torch.float32, device=device),
+                timestamp=[],
+            )
+        )
+
+    encoder_out = model.joiner.encoder_proj(packed_encoder_out.data)
+
+    offset = 0
+    finalized_B = []
+    for (t, batch_size) in enumerate(batch_size_list):
+        start = offset
+        end = offset + batch_size
+        current_encoder_out = encoder_out.data[start:end]
+        current_encoder_out = current_encoder_out.unsqueeze(1).unsqueeze(1)
+        # current_encoder_out's shape is (batch_size, 1, 1, encoder_out_dim)
+        offset = end
+
+        finalized_B = B[batch_size:] + finalized_B
+        B = B[:batch_size]
+
+        hyps_shape = get_hyps_shape(B).to(device)
+
+        A = [list(b) for b in B]
+        B = [HypothesisList() for _ in range(batch_size)]
+
+        ys_log_probs = torch.cat(
+            [hyp.log_prob.reshape(1, 1) for hyps in A for hyp in hyps]
+        )  # (num_hyps, 1)
+
+        decoder_input = torch.tensor(
+            [hyp.ys[-context_size:] for hyps in A for hyp in hyps],
+            device=device,
+            dtype=torch.int64,
+        )  # (num_hyps, context_size)
+
+        decoder_out = model.decoder(decoder_input, need_pad=False).unsqueeze(1)
+        decoder_out = model.joiner.decoder_proj(decoder_out)
+        # decoder_out is of shape (num_hyps, 1, 1, joiner_dim)
+
+        # Note: For torch 1.7.1 and below, it requires a torch.int64 tensor
+        # as index, so we use `to(torch.int64)` below.
+        current_encoder_out = torch.index_select(
+            current_encoder_out,
+            dim=0,
+            index=hyps_shape.row_ids(1).to(torch.int64),
+        )  # (num_hyps, 1, 1, encoder_out_dim)
+
+        logits = model.joiner(
+            current_encoder_out,
+            decoder_out,
+            project_input=False,
+        )  # (num_hyps, 1, 1, vocab_size)
+
+        logits = logits.squeeze(1).squeeze(1)  # (num_hyps, vocab_size)
+
+        log_probs = (logits / temperature).log_softmax(dim=-1)  # (num_hyps, vocab_size)
+
+        log_probs.add_(ys_log_probs)
+
+        vocab_size = log_probs.size(-1)
+
+        log_probs = log_probs.reshape(-1)
+
+        row_splits = hyps_shape.row_splits(1) * vocab_size
+        log_probs_shape = k2.ragged.create_ragged_shape2(
+            row_splits=row_splits, cached_tot_size=log_probs.numel()
+        )
+        ragged_log_probs = k2.RaggedTensor(shape=log_probs_shape, value=log_probs)
+
+        for i in range(batch_size):
+            topk_log_probs, topk_indexes = ragged_log_probs[i].topk(beam)
+
+            with warnings.catch_warnings():
+                warnings.simplefilter("ignore")
+                topk_hyp_indexes = (topk_indexes // vocab_size).tolist()
+                topk_token_indexes = (topk_indexes % vocab_size).tolist()
+
+            for k in range(len(topk_hyp_indexes)):
+                hyp_idx = topk_hyp_indexes[k]
+                hyp = A[i][hyp_idx]
+
+                new_ys = hyp.ys[:]
+                new_token = topk_token_indexes[k]
+                new_timestamp = hyp.timestamp[:]
+                if new_token not in (blank_id, unk_id):
+                    new_ys.append(new_token)
+                    new_timestamp.append(t)
+
+                new_log_prob = topk_log_probs[k]
+                new_hyp = Hypothesis(
+                    ys=new_ys, log_prob=new_log_prob, timestamp=new_timestamp
+                )
+                B[i].add(new_hyp)
+
+    B = B + finalized_B
+
+    # get the am_scores for n-best list
+    hyps_shape = get_hyps_shape(B)
+    am_scores = torch.tensor([hyp.log_prob.item() for b in B for hyp in b])
+    am_scores = k2.RaggedTensor(value=am_scores, shape=hyps_shape).to(device)
+
+    # now LM rescore
+    # prepare input data to LM
+    candidate_seqs = [hyp.ys[context_size:] for b in B for hyp in b]
+    possible_seqs = k2.RaggedTensor(candidate_seqs)
+    row_splits = possible_seqs.shape.row_splits(1)
+    sentence_token_lengths = row_splits[1:] - row_splits[:-1]
+    possible_seqs_with_sos = add_sos(possible_seqs, sos_id=1)
+    possible_seqs_with_eos = add_eos(possible_seqs, eos_id=1)
+    sentence_token_lengths += 1
+
+    x = possible_seqs_with_sos.pad(mode="constant", padding_value=blank_id)
+    y = possible_seqs_with_eos.pad(mode="constant", padding_value=blank_id)
+    x = x.to(device).to(torch.int64)
+    y = y.to(device).to(torch.int64)
+    sentence_token_lengths = sentence_token_lengths.to(device).to(torch.int64)
+
+    lm_scores = LM.lm(x=x, y=y, lengths=sentence_token_lengths)
+    assert lm_scores.ndim == 2
+    lm_scores = -1 * lm_scores.sum(dim=1)
+
+    ans = {}
+    unsorted_indices = packed_encoder_out.unsorted_indices.tolist()
+
+    # get the best hyp with different lm_scale
+    for lm_scale in lm_scale_list:
+        key = f"nnlm_scale_{lm_scale}"
+        tot_scores = am_scores.values + lm_scores * lm_scale
+        ragged_tot_scores = k2.RaggedTensor(shape=am_scores.shape, value=tot_scores)
+        max_indexes = ragged_tot_scores.argmax().tolist()
+        unsorted_hyps = [candidate_seqs[idx] for idx in max_indexes]
+        hyps = []
+        for idx in unsorted_indices:
+            hyps.append(unsorted_hyps[idx])
+
+        ans[key] = hyps
+    return ans
+
+
 def _deprecated_modified_beam_search(
     model: Transducer,
     encoder_out: torch.Tensor,

egs/librispeech/ASR/pruned_transducer_stateless7_streaming/decode.py

@@ -122,6 +122,8 @@ from beam_search import (
     greedy_search,
     greedy_search_batch,
     modified_beam_search,
+    modified_beam_search_lm_rescore,
+    modified_beam_search_lm_shallow_fusion,
 )
 from train import add_model_arguments, get_params, get_transducer_model
 
@@ -132,6 +134,7 @@ from icefall.checkpoint import (
     load_checkpoint,
 )
 from icefall.lexicon import Lexicon
+from icefall.lm_wrapper import LmScorer
 from icefall.utils import (
     AttributeDict,
     setup_logger,
@@ -307,6 +310,32 @@ def get_parser():
         fast_beam_search_nbest_LG, and fast_beam_search_nbest_oracle""",
     )
 
+    parser.add_argument(
+        "--use-shallow-fusion",
+        type=str2bool,
+        default=False,
+        help="""Use neural network LM for shallow fusion.
+        If you want to use LODR, you will also need to set this to true
+        """,
+    )
+
+    parser.add_argument(
+        "--lm-type",
+        type=str,
+        default="rnn",
+        help="Type of NN lm",
+        choices=["rnn", "transformer"],
+    )
+
+    parser.add_argument(
+        "--lm-scale",
+        type=float,
+        default=0.3,
+        help="""The scale of the neural network LM
+        Used only when `--use-shallow-fusion` is set to True.
+        """,
+    )
+
     add_model_arguments(parser)
 
     return parser
@@ -319,6 +348,7 @@ def decode_one_batch(
     batch: dict,
     word_table: Optional[k2.SymbolTable] = None,
     decoding_graph: Optional[k2.Fsa] = None,
+    LM: Optional[LmScorer] = None,
 ) -> Dict[str, List[List[str]]]:
     """Decode one batch and return the result in a dict. The dict has the
     following format:
@@ -443,6 +473,26 @@ def decode_one_batch(
         )
         for hyp in sp.decode(hyp_tokens):
             hyps.append(hyp.split())
+    elif params.decoding_method == "modified_beam_search_lm_shallow_fusion":
+        hyp_tokens = modified_beam_search_lm_shallow_fusion(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+            LM=LM,
+        )
+        for hyp in sp.decode(hyp_tokens):
+            hyps.append(hyp.split())
+    elif params.decoding_method == "modified_beam_search_lm_rescore":
+        lm_scale_list = [0.01 * i for i in range(10, 50)]
+        ans_dict = modified_beam_search_lm_rescore(
+            model=model,
+            encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
+            beam=params.beam_size,
+            LM=LM,
+            lm_scale_list=lm_scale_list,
+        )
     else:
         batch_size = encoder_out.size(0)
 
@@ -481,6 +531,13 @@ def decode_one_batch(
                 key += f"_ngram_lm_scale_{params.ngram_lm_scale}"
 
         return {key: hyps}
+    elif params.decoding_method == "modified_beam_search_lm_rescore":
+        ans = dict()
+        assert ans_dict is not None
+        for key, hyps in ans_dict.items():
+            hyps = [sp.decode(hyp).split() for hyp in hyps]
+            ans[f"beam_size_{params.beam_size}_{key}"] = hyps
+        return ans
     else:
         return {f"beam_size_{params.beam_size}": hyps}
 
@@ -492,6 +549,7 @@ def decode_dataset(
     sp: spm.SentencePieceProcessor,
     word_table: Optional[k2.SymbolTable] = None,
     decoding_graph: Optional[k2.Fsa] = None,
+    LM: Optional[LmScorer] = None,
 ) -> Dict[str, List[Tuple[str, List[str], List[str]]]]:
     """Decode dataset.
 
@@ -541,6 +599,7 @@ def decode_dataset(
             decoding_graph=decoding_graph,
             word_table=word_table,
             batch=batch,
+            LM=LM,
         )
 
         for name, hyps in hyps_dict.items():
@@ -603,6 +662,7 @@ def save_results(
 def main():
     parser = get_parser()
     LibriSpeechAsrDataModule.add_arguments(parser)
+    LmScorer.add_arguments(parser)
     args = parser.parse_args()
     args.exp_dir = Path(args.exp_dir)
 
@@ -617,6 +677,8 @@ def main():
         "fast_beam_search_nbest_LG",
         "fast_beam_search_nbest_oracle",
         "modified_beam_search",
+        "modified_beam_search_lm_shallow_fusion",
+        "modified_beam_search_lm_rescore",
     )
     params.res_dir = params.exp_dir / params.decoding_method
 
@@ -642,6 +704,14 @@ def main():
         params.suffix += f"-context-{params.context_size}"
         params.suffix += f"-max-sym-per-frame-{params.max_sym_per_frame}"
 
+    if params.use_shallow_fusion:
+        params.suffix += f"-{params.lm_type}-lm-scale-{params.lm_scale}"
+
+        if "LODR" in params.decoding_method:
+            params.suffix += (
+                f"-LODR-{params.tokens_ngram}gram-scale-{params.ngram_lm_scale}"
+            )
+
     if params.use_averaged_model:
         params.suffix += "-use-averaged-model"
 
@@ -751,6 +821,19 @@ def main():
     model.to(device)
     model.eval()
 
+    # only load the neural network LM if required
+    if params.use_shallow_fusion or "lm" in params.decoding_method:
+        LM = LmScorer(
+            lm_type=params.lm_type,
+            params=params,
+            device=device,
+            lm_scale=params.lm_scale,
+        )
+        LM.to(device)
+        LM.eval()
+    else:
+        LM = None
+
     if "fast_beam_search" in params.decoding_method:
         if params.decoding_method == "fast_beam_search_nbest_LG":
             lexicon = Lexicon(params.lang_dir)
@@ -792,6 +875,7 @@ def main():
             sp=sp,
             word_table=word_table,
             decoding_graph=decoding_graph,
+            LM=LM,
         )
 
         save_results(

icefall/rnn_lm/model.py

@@ -154,17 +154,18 @@ class RnnLmModel(torch.nn.Module):
         self.cache = {}
 
     def score_token(self, x: torch.Tensor, x_lens: torch.Tensor, state=None):
-        """Score a batch of tokens
+        """Score a batch of tokens, i.e each sample in the batch should be a
+        single token. For example, x = torch.tensor([[5],[10],[20]])
+
 
         Args:
             x (torch.Tensor):
                 A batch of tokens
             x_lens (torch.Tensor):
                 The length of tokens in the batch before padding
-            state (_type_, optional):
+            state (optional):
                 Either None or a tuple of two torch.Tensor. Each tensor has
-                the shape of (hidden_dim)
-
+                the shape of (num_layers, bs, hidden_dim)
 
         Returns:
             _type_: _description_