fix: fix server for triton kernel

ktransformers/server/backend/interfaces/ktransformers.py

@@ -16,6 +16,8 @@ from ktransformers.local_chat import custom_models, default_optimize_rules
 from ktransformers.util.utils import get_device
 
 
+warm_uped = False
+
 class KTransformersThreadContext(TransformersThreadContext):
     pass
 
@@ -74,10 +76,13 @@ class KTransformersInterface(TransformersInterface):
         self._infer_lock = asyncio.Lock()
 
     def decode_one_tokens(self):
+        global warm_uped
+
         device_map = self.model.gguf_loader.tensor_device_map
         torch_device = get_device("blk.0.self_attn", device_map)
         torch_device = "cuda:0" if torch_device == "cuda" else torch_device
-        if self.args.use_cuda_graph:
+        torch.cuda.set_device(torch_device)
+        if warm_uped and self.args.use_cuda_graph:
             if not hasattr(self, "cuda_graph_runner"):
                 self.cuda_graph_runner = CUDAGraphRunner()
                 self.cuda_graph_runner.capture(
@@ -113,6 +118,7 @@ class KTransformersInterface(TransformersInterface):
         else:
             logits = self.model(self.current_ids, return_dict=False)[0]
         logits = logits[0, -1, :]
+        warm_uped = True
 
         return self.logits_to_token(logits)
 
@@ -176,6 +182,7 @@ class KTransformersInterface(TransformersInterface):
         if not (type(self) is TransformersInterface):
             input_ids = input_ids.to("cpu")
         inputs_embeds = self.model.model.embed_tokens(input_ids).to(device)
+        torch.cuda.set_device(device)
         if self.use_static_cache:
             logits = self.model(
                 inputs_embeds=inputs_embeds,

ktransformers/server/main.py

@@ -106,9 +106,6 @@ def custom_openapi(app):
 def main():
     cfg = Config()
 
-    # Temporarily disable cuda graph by default because of a bug in the prefix cache.
-    cfg.use_cuda_graph = False
-
     arg_parser = ArgumentParser(cfg)
 
     # 初始化消息