llms: add cpu support

pilot/model/adapter.py

@@ -9,6 +9,8 @@ from transformers import (
     AutoModel
 )
 
+from pilot.configs.model_config import DEVICE
+
 class BaseLLMAdaper:
     """The Base class for multi model, in our project.
     We will support those model, which performance resemble ChatGPT  """
@@ -64,10 +66,17 @@ class ChatGLMAdapater(BaseLLMAdaper):
 
     def loader(self, model_path: str, from_pretrained_kwargs: dict):
         tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
-        model = AutoModel.from_pretrained(
+
-            model_path, trust_remote_code=True, **from_pretrained_kwargs
+        if DEVICE != "cuda":
-        ).half().cuda()
+            model = AutoModel.from_pretrained(
-        return model, tokenizer
+                model_path, trust_remote_code=True, **from_pretrained_kwargs
+            ).float()
+            return model, tokenizer
+        else:
+            model = AutoModel.from_pretrained(
+                model_path, trust_remote_code=True, **from_pretrained_kwargs
+            ).half().cuda()
+            return model, tokenizer
         
 class CodeGenAdapter(BaseLLMAdaper):
     pass

pilot/server/llmserver.py

@@ -155,6 +155,7 @@ if __name__ == "__main__":
     model_path = LLM_MODEL_CONFIG[CFG.LLM_MODEL]
     print(model_path, DEVICE)
     
+    
     worker = ModelWorker(
         model_path=model_path, 
         model_name=CFG.LLM_MODEL, 

requirements.txt

@@ -42,6 +42,7 @@ tenacity==8.2.2
 peft
 pycocoevalcap
 sentence-transformers
+cpm_kernels
 umap-learn
 notebook
 gradio==3.23