Add Consistency-Regularized CTC

egs/librispeech/ASR/zipformer/model.py

@@ -24,7 +24,8 @@
 from encoder_interface import EncoderInterface
 from scaling import ScaledLinear
 
-from icefall.utils import add_sos, make_pad_mask
+from icefall.utils import add_sos, make_pad_mask, time_warp
+from lhotse.dataset import SpecAugment
 
 
 class AsrModel(nn.Module):
@@ -181,6 +182,49 @@ def forward_ctc(
         )
         return ctc_loss
 
+    def forward_cr_ctc(
+        self,
+        encoder_out: torch.Tensor,
+        encoder_out_lens: torch.Tensor,
+        targets: torch.Tensor,
+        target_lengths: torch.Tensor,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """Compute CTC loss with consistency regularization loss.
+        Args:
+          encoder_out:
+            Encoder output, of shape (2 * N, T, C).
+          encoder_out_lens:
+            Encoder output lengths, of shape (2 * N,).
+          targets:
+            Target Tensor of shape (2 * sum(target_lengths)). The targets are assumed
+            to be un-padded and concatenated within 1 dimension.
+        """
+        # Compute CTC loss
+        ctc_output = self.ctc_output(encoder_out)  # (2 * N, T, C)
+        ctc_loss = torch.nn.functional.ctc_loss(
+            log_probs=ctc_output.permute(1, 0, 2),  # (T, 2 * N, C)
+            targets=targets.cpu(),
+            input_lengths=encoder_out_lens.cpu(),
+            target_lengths=target_lengths.cpu(),
+            reduction="sum",
+        )
+
+        # Compute consistency regularization loss
+        exchanged_targets = ctc_output.detach().chunk(2, dim=0)
+        exchanged_targets = torch.cat(
+            [exchanged_targets[1], exchanged_targets[0]], dim=0
+        )  # exchange: [x1, x2] -> [x2, x1]
+        cr_loss = nn.functional.kl_div(
+            input=ctc_output,
+            target=exchanged_targets,
+            reduction="none",
+            log_target=True,
+        )  # (2 * N, T, C)
+        length_mask = make_pad_mask(encoder_out_lens).unsqueeze(-1)
+        cr_loss = cr_loss.masked_fill(length_mask, 0.0).sum()
+
+        return ctc_loss, cr_loss
+
     def forward_transducer(
         self,
         encoder_out: torch.Tensor,
@@ -296,7 +340,12 @@ def forward(
         prune_range: int = 5,
         am_scale: float = 0.0,
         lm_scale: float = 0.0,
-    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
+        use_cr_ctc: bool = False,
+        use_spec_aug: bool = False,
+        spec_augment: Optional[SpecAugment] = None,
+        supervision_segments: Optional[torch.Tensor] = None,
+        time_warp_factor: Optional[int] = 80,
+    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
         """
         Args:
           x:
@@ -316,9 +365,26 @@ def forward(
           lm_scale:
             The scale to smooth the loss with lm (output of predictor network)
             part
+          use_cr_ctc:
+            Whether use consistency-regularized CTC.
+          use_spec_aug:
+            Whether apply spec-augment manually, used only if use_cr_ctc is True.
+          spec_augment:
+            The SpecAugment instance that returns time masks,
+            used only if use_cr_ctc is True.
+          supervision_segments:
+            An int tensor of shape ``(S, 3)``. ``S`` is the number of
+            supervision segments that exist in ``features``.
+            Used only if use_cr_ctc is True.
+          time_warp_factor:
+            Parameter for the time warping; larger values mean more warping.
+            Set to ``None``, or less than ``1``, to disable.
+            Used only if use_cr_ctc is True.
+
         Returns:
-          Return the transducer losses and CTC loss,
-          in form of (simple_loss, pruned_loss, ctc_loss, attention_decoder_loss)
+          Return the transducer losses, CTC loss, AED loss,
+          and consistency-regularization loss in form of
+          (simple_loss, pruned_loss, ctc_loss, attention_decoder_loss, cr_loss)
 
         Note:
            Regarding am_scale & lm_scale, it will make the loss-function one of
@@ -334,6 +400,24 @@ def forward(
 
         device = x.device
 
+        if use_cr_ctc:
+            assert self.use_ctc
+            if use_spec_aug:
+                assert spec_augment is not None and spec_augment.time_warp_factor < 1
+                # Apply time warping before input duplicating
+                assert supervision_segments is not None
+                x = time_warp(
+                    x,
+                    time_warp_factor=time_warp_factor,
+                    supervision_segments=supervision_segments,
+                )
+                # Independently apply frequency masking and time masking to the two copies
+                x = spec_augment(x.repeat(2, 1, 1))
+            else:
+                x = x.repeat(2, 1, 1)
+            x_lens = x_lens.repeat(2)
+            y = k2.ragged.cat([y, y], axis=0)
+
         # Compute encoder outputs
         encoder_out, encoder_out_lens = self.forward_encoder(x, x_lens)
 
@@ -351,21 +435,36 @@ def forward(
                 am_scale=am_scale,
                 lm_scale=lm_scale,
             )
+            if use_cr_ctc:
+                simple_loss = simple_loss * 0.5
+                pruned_loss = pruned_loss * 0.5
         else:
             simple_loss = torch.empty(0)
             pruned_loss = torch.empty(0)
 
         if self.use_ctc:
             # Compute CTC loss
             targets = y.values
-            ctc_loss = self.forward_ctc(
-                encoder_out=encoder_out,
-                encoder_out_lens=encoder_out_lens,
-                targets=targets,
-                target_lengths=y_lens,
-            )
+            if not use_cr_ctc:
+                ctc_loss = self.forward_ctc(
+                    encoder_out=encoder_out,
+                    encoder_out_lens=encoder_out_lens,
+                    targets=targets,
+                    target_lengths=y_lens,
+                )
+                cr_loss = torch.empty(0)
+            else:
+                ctc_loss, cr_loss = self.forward_cr_ctc(
+                    encoder_out=encoder_out,
+                    encoder_out_lens=encoder_out_lens,
+                    targets=targets,
+                    target_lengths=y_lens,
+                )
+                ctc_loss = ctc_loss * 0.5
+                cr_loss = cr_loss * 0.5
         else:
             ctc_loss = torch.empty(0)
+            cr_loss = torch.empty(0)
 
         if self.use_attention_decoder:
             attention_decoder_loss = self.attention_decoder.calc_att_loss(
@@ -374,7 +473,9 @@ def forward(
                 ys=y.to(device),
                 ys_lens=y_lens.to(device),
             )
+            if use_cr_ctc:
+                attention_decoder_loss = attention_decoder_loss * 0.5
         else:
             attention_decoder_loss = torch.empty(0)
 
-        return simple_loss, pruned_loss, ctc_loss, attention_decoder_loss
+        return simple_loss, pruned_loss, ctc_loss, attention_decoder_loss, cr_loss

egs/librispeech/ASR/zipformer/train.py

@@ -72,6 +72,7 @@
 from decoder import Decoder
 from joiner import Joiner
 from lhotse.cut import Cut
+from lhotse.dataset import SpecAugment
 from lhotse.dataset.sampling.base import CutSampler
 from lhotse.utils import fix_random_seed
 from model import AsrModel
@@ -304,6 +305,13 @@ def add_model_arguments(parser: argparse.ArgumentParser):
         help="If True, use attention-decoder head.",
     )
 
+    parser.add_argument(
+        "--use-cr-ctc",
+        type=str2bool,
+        default=False,
+        help="If True, use consistency-regularized CTC.",
+    )
+
 
 def get_parser():
     parser = argparse.ArgumentParser(
@@ -449,6 +457,20 @@ def get_parser():
         help="Scale for CTC loss.",
     )
 
+    parser.add_argument(
+        "--cr-loss-scale",
+        type=float,
+        default=0.2,
+        help="Scale for consistency-regularization loss.",
+    )
+
+    parser.add_argument(
+        "--time-mask-ratio",
+        type=float,
+        default=2.5,
+        help="When using cr-ctc, we increase the amount of time-masking in SpecAugment.",
+    )
+
     parser.add_argument(
         "--attention-decoder-loss-scale",
         type=float,
@@ -717,6 +739,24 @@ def get_model(params: AttributeDict) -> nn.Module:
     return model
 
 
+def get_spec_augment(params: AttributeDict) -> SpecAugment:
+    num_frame_masks = int(10 * params.time_mask_ratio)
+    max_frames_mask_fraction = 0.15 * params.time_mask_ratio
+    logging.info(
+        f"num_frame_masks: {num_frame_masks}, "
+        f"max_frames_mask_fraction: {max_frames_mask_fraction}"
+    )
+    spec_augment = SpecAugment(
+        time_warp_factor=0,  # Do time warping in model.py
+        num_frame_masks=num_frame_masks,  # default: 10
+        features_mask_size=27,
+        num_feature_masks=2,
+        frames_mask_size=100,
+        max_frames_mask_fraction=max_frames_mask_fraction,  # default: 0.15
+    )
+    return spec_augment
+
+
 def load_checkpoint_if_available(
     params: AttributeDict,
     model: nn.Module,
@@ -839,6 +879,7 @@ def compute_loss(
     sp: spm.SentencePieceProcessor,
     batch: dict,
     is_training: bool,
+    spec_augment: Optional[SpecAugment] = None,
 ) -> Tuple[Tensor, MetricsTracker]:
     """
     Compute loss given the model and its inputs.
@@ -855,8 +896,8 @@ def compute_loss(
         True for training. False for validation. When it is True, this
         function enables autograd during computation; when it is False, it
         disables autograd.
-      warmup: a floating point value which increases throughout training;
-        values >= 1.0 are fully warmed up and have all modules present.
+      spec_augment:
+        The SpecAugment instance used only when use_cr_ctc is True.
     """
     device = model.device if isinstance(model, DDP) else next(model.parameters()).device
     feature = batch["inputs"]
@@ -874,14 +915,34 @@ def compute_loss(
     y = sp.encode(texts, out_type=int)
     y = k2.RaggedTensor(y)
 
+    use_cr_ctc = params.use_cr_ctc
+    use_spec_aug = use_cr_ctc and is_training
+    if use_spec_aug:
+        supervision_intervals = batch["supervisions"]
+        supervision_segments = torch.stack(
+            [
+                supervision_intervals["sequence_idx"],
+                supervision_intervals["start_frame"],
+                supervision_intervals["num_frames"],
+            ],
+            dim=1,
+        )  # shape: (S, 3)
+    else:
+        supervision_segments = None
+
     with torch.set_grad_enabled(is_training):
-        simple_loss, pruned_loss, ctc_loss, attention_decoder_loss = model(
+        simple_loss, pruned_loss, ctc_loss, attention_decoder_loss, cr_loss = model(
             x=feature,
             x_lens=feature_lens,
             y=y,
             prune_range=params.prune_range,
             am_scale=params.am_scale,
             lm_scale=params.lm_scale,
+            use_cr_ctc=use_cr_ctc,
+            use_spec_aug=use_spec_aug,
+            spec_augment=spec_augment,
+            supervision_segments=supervision_segments,
+            time_warp_factor=params.spec_aug_time_warp_factor,
         )
 
         loss = 0.0
@@ -904,6 +965,8 @@ def compute_loss(
 
         if params.use_ctc:
             loss += params.ctc_loss_scale * ctc_loss
+            if use_cr_ctc:
+                loss += params.cr_loss_scale * cr_loss
 
         if params.use_attention_decoder:
             loss += params.attention_decoder_loss_scale * attention_decoder_loss
@@ -922,6 +985,8 @@ def compute_loss(
         info["pruned_loss"] = pruned_loss.detach().cpu().item()
     if params.use_ctc:
         info["ctc_loss"] = ctc_loss.detach().cpu().item()
+        if params.use_cr_ctc:
+            info["cr_loss"] = cr_loss.detach().cpu().item()
     if params.use_attention_decoder:
         info["attn_decoder_loss"] = attention_decoder_loss.detach().cpu().item()
 
@@ -971,6 +1036,7 @@ def train_one_epoch(
     train_dl: torch.utils.data.DataLoader,
     valid_dl: torch.utils.data.DataLoader,
     scaler: GradScaler,
+    spec_augment: Optional[SpecAugment] = None,
     model_avg: Optional[nn.Module] = None,
     tb_writer: Optional[SummaryWriter] = None,
     world_size: int = 1,
@@ -997,6 +1063,8 @@ def train_one_epoch(
         Dataloader for the validation dataset.
       scaler:
         The scaler used for mix precision training.
+      spec_augment:
+        The SpecAugment instance used only when use_cr_ctc is True.
       model_avg:
         The stored model averaged from the start of training.
       tb_writer:
@@ -1043,6 +1111,7 @@ def save_bad_model(suffix: str = ""):
                     sp=sp,
                     batch=batch,
                     is_training=True,
+                    spec_augment=spec_augment,
                 )
             # summary stats
             tot_loss = (tot_loss * (1 - 1 / params.reset_interval)) + loss_info
@@ -1238,6 +1307,13 @@ def run(rank, world_size, args):
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
+    if params.use_cr_ctc:
+        assert params.use_ctc
+        assert not params.enable_spec_aug  # we will do spec_augment in model.py
+        spec_augment = get_spec_augment(params)
+    else:
+        spec_augment = None
+
     assert params.save_every_n >= params.average_period
     model_avg: Optional[nn.Module] = None
     if rank == 0:
@@ -1360,6 +1436,7 @@ def remove_short_and_long_utt(c: Cut):
             optimizer=optimizer,
             sp=sp,
             params=params,
+            spec_augment=spec_augment,
         )
 
     scaler = GradScaler(enabled=params.use_autocast, init_scale=1.0)
@@ -1387,6 +1464,7 @@ def remove_short_and_long_utt(c: Cut):
             train_dl=train_dl,
             valid_dl=valid_dl,
             scaler=scaler,
+            spec_augment=spec_augment,
             tb_writer=tb_writer,
             world_size=world_size,
             rank=rank,
@@ -1452,6 +1530,7 @@ def scan_pessimistic_batches_for_oom(
     optimizer: torch.optim.Optimizer,
     sp: spm.SentencePieceProcessor,
     params: AttributeDict,
+    spec_augment: Optional[SpecAugment] = None,
 ):
     from lhotse.dataset import find_pessimistic_batches
 
@@ -1471,6 +1550,7 @@ def scan_pessimistic_batches_for_oom(
                     sp=sp,
                     batch=batch,
                     is_training=True,
+                    spec_augment=spec_augment,
                 )
             loss.backward()
             optimizer.zero_grad()