如图，模型可以正常载入，但在推理时报错。使用的镜像为image.sourcefind.cn:5000/dcu/admin/base/vllm:0.9.2-ubuntu22.04-dtk25.04.1-rc5-rocblas104381-0915-das1.6-py3.10-20250916-rc2-ds3.2，vllm启动命令为： ``` vllm serve /root/Workspace/Llama-3.1-8B-Instruct \ --host 0.0.0.0 \ --port 28000 \ --tensor-parallel-size 8 \ --pipeline-parallel-size 1 \ --gpu-memory-utilization 0.85 \ --max-num-batched-tokens 32768 \ --max-num-seqs 1024 \ --max-model-len 65536 \ --dtype float16 \ --enable-prefix-caching \ --enable-chunked-prefill \ --api-key "sk-8af5bf335e628cfa4904c6ff1b673174" ``` 控制台报错输出为： ``` RROR 10-19 16:33:18 [dump_input.py:69] Dumping input data for V1 LLM engine (v0.9.2) with config: model='/root/Workspace/Llama-3.1-8B-Instruct', speculative_config=None, tokenizer='/root/Workspace/Llama-3.1-8B-Instruct', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config={}, tokenizer_revision=None, trust_remote_code=False, dtype=torch.float16, max_seq_len=65536, download_dir=None, load_format=auto, tensor_parallel_size=8, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, kv_cache_dtype=auto, device_config=cuda, decoding_config=DecodingConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_backend=''), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None), seed=0, served_model_name=/root/Workspace/Llama-3.1-8B-Instruct, num_scheduler_steps=1, multi_step_stream_outputs=True, enable_prefix_caching=True, chunked_prefill_enabled=True, use_async_output_proc=True, pooler_config=None, compilation_config={"level":3,"debug_dump_path":"","cache_dir":"","backend":"","custom_ops":[],"splitting_ops":["vllm.unified_attention","vllm.unified_attention_with_output"],"use_inductor":true,"compile_sizes":[],"inductor_compile_config":{},"inductor_passes":{},"use_cudagraph":true,"cudagraph_num_of_warmups":1,"cudagraph_capture_sizes":[512,504,496,488,480,472,464,456,448,440,432,424,416,408,400,392,384,376,368,360,352,344,336,328,320,312,304,296,288,280,272,264,256,248,240,232,224,216,208,200,192,184,176,168,160,152,144,136,128,120,112,104,96,88,80,72,64,56,48,40,32,24,16,8,4,2,1],"cudagraph_copy_inputs":false,"full_cuda_graph":false,"max_capture_size":512,"local_cache_dir":null}, ERROR 10-19 16:33:18 [dump_input.py:76] Dumping scheduler output for model execution: SchedulerOutput(scheduled_new_reqs=[NewRequestData(req_id=chatcmpl-ee691e225b28486987514520f95257cb,prompt_token_ids_len=41,mm_inputs=[],mm_hashes=[],mm_positions=[],sampling_params=SamplingParams(n=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.0, temperature=0.6, top_p=0.9, top_k=0, min_p=0.0, seed=None, stop=[], stop_token_ids=[128008, 128001], bad_words=[], include_stop_str_in_output=False, ignore_eos=False, max_tokens=100, min_tokens=0, logprobs=None, prompt_logprobs=None, skip_special_tokens=True, spaces_between_special_tokens=True, truncate_prompt_tokens=None, guided_decoding=None, extra_args=None),block_ids=([1],),num_computed_tokens=0,lora_request=None)], scheduled_cached_reqs=CachedRequestData(req_ids=[], resumed_from_preemption=[], new_token_ids=[], new_block_ids=[], num_computed_tokens=[]), num_scheduled_tokens={chatcmpl-ee691e225b28486987514520f95257cb: 41}, total_num_scheduled_tokens=41, scheduled_spec_decode_tokens={}, scheduled_encoder_inputs={}, num_common_prefix_blocks=[1], finished_req_ids=[], free_encoder_input_ids=[], structured_output_request_ids={}, grammar_bitmask=null, kv_connector_metadata=null) ERROR 10-19 16:33:18 [dump_input.py:79] Dumping scheduler stats: SchedulerStats(num_running_reqs=1, num_waiting_reqs=0, kv_cache_usage=4.1732743510602965e-05, prefix_cache_stats=PrefixCacheStats(reset=False, requests=1, queries=41, hits=0), spec_decoding_stats=None, num_corrupted_reqs=0) ERROR 10-19 16:33:18 [core.py:591] EngineCore encountered a fatal error. ERROR 10-19 16:33:18 [core.py:591] Traceback (most recent call last): ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/core.py", line 582, in run_engine_core ERROR 10-19 16:33:18 [core.py:591] engine_core.run_busy_loop() ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/core.py", line 609, in run_busy_loop ERROR 10-19 16:33:18 [core.py:591] self._process_engine_step() ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/core.py", line 634, in _process_engine_step ERROR 10-19 16:33:18 [core.py:591] outputs, model_executed = self.step_fn() ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/core.py", line 239, in step ERROR 10-19 16:33:18 [core.py:591] model_output = self.execute_model(scheduler_output) ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/core.py", line 223, in execute_model ERROR 10-19 16:33:18 [core.py:591] raise err ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/core.py", line 214, in execute_model ERROR 10-19 16:33:18 [core.py:591] return self.model_executor.execute_model(scheduler_output) ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/executor/multiproc_executor.py", line 158, in execute_model ERROR 10-19 16:33:18 [core.py:591] (output, ) = self.collective_rpc( ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/executor/multiproc_executor.py", line 215, in collective_rpc ERROR 10-19 16:33:18 [core.py:591] result = get_response(w, dequeue_timeout) ERROR 10-19 16:33:18 [core.py:591] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/executor/multiproc_executor.py", line 202, in get_response ERROR 10-19 16:33:18 [core.py:591] raise RuntimeError( ERROR 10-19 16:33:18 [core.py:591] RuntimeError: Worker failed with error 'HIP error: invalid device function ERROR 10-19 16:33:18 [core.py:591] HIP kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. ERROR 10-19 16:33:18 [core.py:591] For debugging consider passing AMD_SERIALIZE_KERNEL=3 ERROR 10-19 16:33:18 [core.py:591] Compile with `TORCH_USE_HIP_DSA` to enable device-side assertions. ERROR 10-19 16:33:18 [core.py:591] ', please check the stack trace above for the root cause ERROR 10-19 16:33:18 [async_llm.py:419] AsyncLLM output_handler failed. ERROR 10-19 16:33:18 [async_llm.py:419] Traceback (most recent call last): ERROR 10-19 16:33:18 [async_llm.py:419] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/async_llm.py", line 378, in output_handler ERROR 10-19 16:33:18 [async_llm.py:419] outputs = await engine_core.get_output_async() ERROR 10-19 16:33:18 [async_llm.py:419] File "/usr/local/lib/python3.10/dist-packages/vllm/v1/engine/core_client.py", line 740, in get_output_async ERROR 10-19 16:33:18 [async_llm.py:419] raise self._format_exception(outputs) from None ERROR 10-19 16:33:18 [async_llm.py:419] vllm.v1.engine.exceptions.EngineDeadError: EngineCore encountered an issue. See stack trace (above) for the root cause. INFO 10-19 16:33:18 [async_llm.py:345] Request chatcmpl-ee691e225b28486987514520f95257cb failed (engine dead). INFO: 10.63.123.218:54018 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error INFO: Shutting down INFO: Waiting for application shutdown. INFO: Application shutdown complete. INFO: Finished server process [37] /usr/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 1 leaked shared_memory objects to clean up at shutdown warnings.warn('resource_tracker: There appear to be %d ' root@gpuserver1:/workspace# exit exit ```