diff --git a/tests/models/transformers/test_backend.py b/tests/models/transformers/test_backend.py index 621c249137..0429f3da2c 100644 --- a/tests/models/transformers/test_backend.py +++ b/tests/models/transformers/test_backend.py @@ -479,11 +479,11 @@ def replace_vocab_embeddings(model, **config_kwargs): stub.config = PretrainedConfig( vocab_size=VOCAB_SIZE, num_positions=NUM_POSITIONS, **config_kwargs ) - ids = Base._vocab_embedding_ids(stub) + embeddings = Base._vocab_embeddings(stub) replaced = [] for name, module in list(model.named_modules()): - if id(module) in ids: + if module in embeddings: replaced.append(replace(module)) attrsetter(name)(model, replaced[-1]) return replaced diff --git a/vllm/model_executor/models/transformers/base.py b/vllm/model_executor/models/transformers/base.py index 0377b5708a..11b7c8466c 100644 --- a/vllm/model_executor/models/transformers/base.py +++ b/vllm/model_executor/models/transformers/base.py @@ -416,8 +416,8 @@ class Base( # attrsetter in case the module is nested (e.g. "text_model.norm") attrsetter(name)(module, PPMissingLayer()) - def _vocab_embedding_ids(self) -> set[int]: - """ids of the `nn.Embedding`s in `self.model` which hold a vocab table.""" + def _vocab_embeddings(self) -> set[nn.Embedding]: + """The `nn.Embedding`s in `self.model` which hold a vocab table.""" def vocab_sizes(config: PretrainedConfig): for key, value in vars(config).items(): @@ -429,17 +429,17 @@ class Base( sizes = set(vocab_sizes(self.config)) # `get_input_embeddings` may return a module which composes the `nn.Embedding`, # or a `PPMissingLayer` if the embeddings are not on this pipeline stage - ids = { - id(module) + embeddings = { + module for module in self.model.get_input_embeddings().modules() if isinstance(module, nn.Embedding) } - ids.update( - id(module) + embeddings.update( + module for module in self.model.modules() if isinstance(module, nn.Embedding) and module.num_embeddings in sizes ) - return ids + return embeddings def recursive_replace(self): """Recursively replace modules in the model as needed. @@ -473,7 +473,7 @@ class Base( # Detect fusable patterns once per module class (cached, so this is cheap) fusers = Fusers(self.model, self.vllm_config) - vocab_embedding_ids = self._vocab_embedding_ids() + vocab_embeddings = self._vocab_embeddings() def register_fusion(fuser: BaseFuser, prefix: str): """Register a fused layer's mappings just before it is built.""" @@ -524,7 +524,7 @@ class Base( ) elif isinstance(child_module, (nn.Conv2d, nn.Conv3d)): new_module = replace_conv_class(child_module) - elif id(child_module) in vocab_embedding_ids: + elif child_module in vocab_embeddings: new_module = replace_embedding_class( child_module, self.quant_config, prefix=qual_name )