fix #2782 #2798

src/llmtuner/chat/vllm_engine.py

@@ -31,9 +31,11 @@ class VllmEngine(BaseEngine):
             model=model_args.model_name_or_path,
             trust_remote_code=True,
             max_model_len=model_args.vllm_maxlen,
-            tensor_parallel_size=get_device_count(),
+            tensor_parallel_size=get_device_count() or 1,
+            gpu_memory_utilization=model_args.vllm_gpu_util,
             disable_log_stats=True,
             disable_log_requests=True,
+            enforce_eager=model_args.vllm_enforce_eager,
         )
         self.model = AsyncLLMEngine.from_engine_args(engine_args)
         self.tokenizer = load_tokenizer(model_args)

src/llmtuner/hparams/model_args.py

@@ -89,6 +89,14 @@ class ModelArguments:
         default=2048,
         metadata={"help": "Maximum input length of the vLLM engine."},
     )
+    vllm_gpu_util: float = field(
+        default=0.9,
+        metadata={"help": "The fraction of GPU memory in (0,1) to be used for the vLLM engine."},
+    )
+    vllm_enforce_eager: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to disable CUDA graph in the vLLM engine."},
+    )
     hf_hub_token: Optional[str] = field(
         default=None,
         metadata={"help": "Auth token to log in with Hugging Face Hub."},

src/llmtuner/hparams/parser.py

@@ -124,7 +124,7 @@ def get_train_args(args: Optional[Dict[str, Any]] = None) -> _TRAIN_CLS:
     if training_args.do_train and training_args.predict_with_generate:
         raise ValueError("`predict_with_generate` cannot be set as True while training.")
 
-    if training_args.do_train and model_args.use_unsloth and not is_unsloth_available:
+    if training_args.do_train and model_args.use_unsloth and not is_unsloth_available():
         raise ValueError("Unsloth was not installed: https://github.com/unslothai/unsloth")
 
     if finetuning_args.use_dora:
@@ -141,6 +141,13 @@ def get_train_args(args: Optional[Dict[str, Any]] = None) -> _TRAIN_CLS:
         if training_args.fp16 or training_args.bf16:
             raise ValueError("Turn off mixed precision training when using `pure_bf16`.")
 
+    if (
+        finetuning_args.use_galore
+        and finetuning_args.galore_layerwise
+        and training_args.parallel_mode.value == "distributed"
+    ):
+        raise ValueError("Distributed training does not support layer-wise GaLore.")
+
     if model_args.infer_backend == "vllm":
         raise ValueError("vLLM backend is only available for API, CLI and Web.")