support musa backend for MooreThreads

fastchat/model/model_adapter.py

@@ -271,6 +271,12 @@ def load_model(
             import torch_npu
         except ImportError:
             warnings.warn("Ascend Extension for PyTorch is not installed.")
+    elif device == "musa":
+        kwargs = {"torch_dtype": torch.float16}
+        try:
+            import torch_musa
+        except ImportError:
+            warnings.warn("Musa Extension for PyTorch is not installed.")
     else:
         raise ValueError(f"Invalid device: {device}")
 
@@ -377,6 +383,7 @@ def load_model(
         "mps",
         "xpu",
         "npu",
+        "musa",
     ):
         model.to(device)
 
@@ -495,7 +502,7 @@ def add_model_args(parser):
     parser.add_argument(
         "--device",
         type=str,
-        choices=["cpu", "cuda", "mps", "xpu", "npu"],
+        choices=["cpu", "cuda", "mps", "xpu", "npu", "musa"],
         default="cuda",
         help="The device type",
     )

fastchat/serve/model_worker.py

@@ -89,6 +89,7 @@ def __init__(
             xft_config=xft_config,
             debug=debug,
         )
+        self.model_path = model_path
         self.device = device
         if self.tokenizer.pad_token == None:
             self.tokenizer.pad_token = self.tokenizer.eos_token
@@ -165,6 +166,9 @@ def __process_embed_chunk(self, input_ids, attention_mask, **model_type_dict):
             else:
                 data = model_output.hidden_states[-1]
 
+        #import pdb
+        #pdb.set_trace()
+        #sum_embeddings = data[:, 0]
         if hasattr(self.model, "use_cls_pooling") and self.model.use_cls_pooling:
             sum_embeddings = data[:, 0]
         else:
@@ -245,9 +249,10 @@ def get_embeddings(self, params):
                             )
                             + tokenizer.cls_token_id
                         )
-                        chunk_input_ids = torch.cat(
-                            [cls_tokens, chunk_input_ids], dim=-1
-                        )
+                        if "bge-m3" not in self.model_path.lower():
+                            chunk_input_ids = torch.cat(
+                                [cls_tokens, chunk_input_ids], dim=-1
+                            )
                         mask = torch.ones(
                             (chunk_attention_mask.size(0), 1),
                             dtype=chunk_attention_mask.dtype,
@@ -260,6 +265,8 @@ def get_embeddings(self, params):
                     chunk_embeddings, token_num = self.__process_embed_chunk(
                         chunk_input_ids, chunk_attention_mask, **model_type_dict
                     )
+                    import pdb
+                    pdb.set_trace()
                     if (
                         hasattr(self.model, "use_cls_pooling")
                         and self.model.use_cls_pooling
@@ -271,6 +278,8 @@ def get_embeddings(self, params):
 
                 all_embeddings_tensor = torch.stack(all_embeddings)
                 embedding = torch.sum(all_embeddings_tensor, dim=0) / all_token_num
+                import pdb
+                pdb.set_trace()
                 normalized_embeddings = F.normalize(embedding, p=2, dim=1)
 
                 ret["token_num"] = all_token_num