[Core] Refactor model loading code (#4097)

vllm/model_executor/models/commandr.py

@@ -20,7 +20,7 @@
 
 # This file is based on the LLama model definition file in transformers
 """PyTorch Cohere model."""
-from typing import List, Optional, Tuple
+from typing import Iterable, List, Optional, Tuple
 
 import torch
 import torch.utils.checkpoint
@@ -41,10 +41,9 @@ from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.sampler import Sampler
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     VocabParallelEmbedding)
+from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.model_executor.utils import set_weight_attrs
-from vllm.model_executor.weight_utils import (default_weight_loader,
-                                              hf_model_weights_iterator)
 from vllm.sequence import SamplerOutput
 
 
@@ -335,13 +334,7 @@ class CohereForCausalLM(nn.Module):
         next_tokens = self.sampler(logits, sampling_metadata)
         return next_tokens
 
-    def load_weights(
-        self,
-        model_name_or_path: str,
-        cache_dir: Optional[str] = None,
-        load_format: str = "auto",
-        revision: Optional[str] = None,
-    ):
+    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
         stacked_params_mapping = [
             # (param_name, shard_name, shard_id)
             ("qkv_proj", "q_proj", "q"),
@@ -352,8 +345,7 @@ class CohereForCausalLM(nn.Module):
         ]
         params_dict = dict(self.named_parameters())
         loaded_params = set()
-        for name, loaded_weight in hf_model_weights_iterator(
-                model_name_or_path, cache_dir, load_format, revision):
+        for name, loaded_weight in weights:
             for param_name, shard_name, shard_id in stacked_params_mapping:
                 if shard_name not in name:
                     continue