support freezing modules

egs/librispeech/ASR/whisper/train.py

@@ -88,15 +88,6 @@ from icefall.utils import (
 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
@@ -226,6 +217,13 @@ def get_parser():
         help="Whether to use half precision training.",
     )
 
+    parser.add_argument(
+        "--freeze-modules",
+        type=str,
+        default=None,
+        help="Which modules to freeze during finetune"
+    )
+
     parser = deepspeed.add_config_arguments(parser)
 
     return parser
@@ -583,6 +581,9 @@ def train_one_epoch(
         be set to 0.
     """
     model.train()
+    for name, module in model.named_modules():
+        if name.startswith(params.freeze_modules):
+            module.eval()
 
     tot_loss = MetricsTracker()
 
@@ -630,7 +631,6 @@ def train_one_epoch(
                 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)
@@ -739,8 +739,19 @@ def run(rank, world_size, args):
     replace_whisper_encoder_forward()
     model = whisper.load_model(params.model_name, "cpu")
     del model.alignment_heads
+
+    if params.freeze_modules is not None:        
+        for name, p in model.named_parameters():
+            if name.startswith(params.freeze_modules):
+                p.requires_grad = False
+                logging.info(f"Do not update {name}")
+        for name, module in model.named_modules():
+            if name.startswith(params.freeze_modules):
+                module.eval()
+
     num_param = sum([p.numel() for p in model.parameters()])
-    logging.info(f"Number of model parameters: {num_param}")
+    num_trainable = sum([p.numel() if p.requires_grad else 0 for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}. Total trainable parameters: {num_trainable}")
 
     tokenizer = whisper.tokenizer.get_tokenizer(
         model.is_multilingual,