基础环境：
Z100L加速卡 * 2
镜像 image.sourcefind.cn:5000/dcu/admin/base/jupyterlab-deepseek-r1-distill-qwen-7b：pytorch2.3.0-py3.10-dtk24.04.3-model
CPU:Hygon C86 7390 32-core Processor * 2
内存：512G

具体报错为：

root@node02:~# ./start.sh
Started vllm with PID 34
Checking if output.log has content...
Waiting for output to appear in vllm-output.log...
Waiting for output to appear in vllm-output.log...
Waiting for output to appear in vllm-output.log...
Waiting for output to appear in vllm-output.log...
INFO 02-07 17:32:54 api_server.py:526] vLLM API server version 0.6.2
INFO 02-07 17:32:54 api_server.py:527] args: Namespace(host=None, port=8000, uvicorn_log_level='info', allow_credentials=False, allowed_origins=['*'], allowed_methods=['*'], allowed_headers=['*'], api_key=None, lora_modules=None, prompt_adapters=None, chat_template=None, response_role='assistant', ssl_keyfile=None, ssl_certfile=None, ssl_ca_certs=None, ssl_cert_reqs=0, root_path=None, middleware=[], return_tokens_as_token_ids=False, disable_frontend_multiprocessing=False, enable_auto_tool_choice=False, tool_call_parser=None, model='DeepSeek-R1-Distill-Qwen-7B', tokenizer=None, skip_tokenizer_init=False, revision=None, code_revision=None, tokenizer_revision=None, tokenizer_mode='auto', trust_remote_code=True, download_dir=None, load_format='auto', config_format='auto', dtype='float16', kv_cache_dtype='auto', quantization_param_path=None, max_model_len=None, guided_decoding_backend='outlines', distributed_executor_backend=None, worker_use_ray=False, pipeline_parallel_size=1, tensor_parallel_size=1, max_parallel_loading_workers=None, ray_workers_use_nsight=False, block_size=16, enable_prefix_caching=False, disable_sliding_window=False, use_v2_block_manager=False, num_lookahead_slots=0, seed=0, swap_space=4, cpu_offload_gb=0, gpu_memory_utilization=0.9, num_gpu_blocks_override=None, max_num_batched_tokens=None, max_num_seqs=256, max_logprobs=20, disable_log_stats=False, quantization=None, rope_scaling=None, rope_theta=None, enforce_eager=True, max_context_len_to_capture=None, max_seq_len_to_capture=8192, disable_custom_all_reduce=False, tokenizer_pool_size=0, tokenizer_pool_type='ray', tokenizer_pool_extra_config=None, limit_mm_per_prompt=None, mm_processor_kwargs=None, enable_lora=False, max_loras=1, max_lora_rank=16, merge_lora=False, lora_target_modules=None, lora_extra_vocab_size=256, lora_dtype='auto', long_lora_scaling_factors=None, max_cpu_loras=None, fully_sharded_loras=False, enable_prompt_adapter=False, max_prompt_adapters=1, max_prompt_adapter_token=0, device='auto', num_scheduler_steps=1, multi_step_stream_outputs=False, scheduler_delay_factor=0.0, enable_chunked_prefill=None, speculative_model=None, speculative_model_quantization=None, num_speculative_tokens=None, num_speculative_heads=None, speculative_draft_tensor_parallel_size=None, speculative_max_model_len=None, speculative_disable_by_batch_size=None, ngram_prompt_lookup_max=None, ngram_prompt_lookup_min=None, spec_decoding_acceptance_method='rejection_sampler', tree_style_spec_decoding=False, typical_acceptance_sampler_posterior_threshold=None, typical_acceptance_sampler_posterior_alpha=None, disable_logprobs_during_spec_decoding=None, model_loader_extra_config=None, ignore_patterns=[], preemption_mode=None, served_model_name=None, qlora_adapter_name_or_path=None, otlp_traces_endpoint=None, collect_detailed_traces=None, disable_async_output_proc=False, override_neuron_config=None, disable_log_requests=False, max_log_len=None, disable_fastapi_docs=False)
INFO 02-07 17:32:54 api_server.py:164] Multiprocessing frontend to use ipc:///tmp/495a3f92-43fb-4f4c-ae63-986a0dfc9f6d for IPC Path.
INFO 02-07 17:32:55 api_server.py:177] Started engine process with PID 113
WARNING 02-07 17:32:55 config.py:1687] Casting torch.bfloat16 to torch.float16.
INFO 02-07 17:32:55 config.py:937] Disabled the custom all-reduce kernel because it is not supported on hcus.
WARNING 02-07 17:32:55 arg_utils.py:953] Chunked prefill is enabled by default for models with max_model_len > 32K. Currently, chunked prefill might not work with some features or models. If you encounter any issues, please disable chunked prefill by setting --enable-chunked-prefill=False.
INFO 02-07 17:32:55 config.py:1019] Chunked prefill is enabled with max_num_batched_tokens=512.
WARNING 02-07 17:32:55 config.py:398] To see benefits of async output processing, enable CUDA graph. Since, enforce-eager is enabled, async output processor cannot be used
WARNING 02-07 17:33:02 config.py:1687] Casting torch.bfloat16 to torch.float16.
INFO 02-07 17:33:02 config.py:937] Disabled the custom all-reduce kernel because it is not supported on hcus.
WARNING 02-07 17:33:02 arg_utils.py:953] Chunked prefill is enabled by default for models with max_model_len > 32K. Currently, chunked prefill might not work with some features or models. If you encounter any issues, please disable chunked prefill by setting --enable-chunked-prefill=False.
INFO 02-07 17:33:02 config.py:1019] Chunked prefill is enabled with max_num_batched_tokens=512.
WARNING 02-07 17:33:02 config.py:398] To see benefits of async output processing, enable CUDA graph. Since, enforce-eager is enabled, async output processor cannot be used
INFO 02-07 17:33:02 llm_engine.py:226] Initializing an LLM engine (v0.6.2) with config: model='DeepSeek-R1-Distill-Qwen-7B', speculative_config=None, tokenizer='DeepSeek-R1-Distill-Qwen-7B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config=None, rope_scaling=None, rope_theta=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.float16, max_seq_len=131072, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=True, quantization=None, enforce_eager=True, kv_cache_dtype=auto, quantization_param_path=None, device_config=cuda, decoding_config=DecodingConfig(guided_decoding_backend='outlines'), observability_config=ObservabilityConfig(otlp_traces_endpoint=None, collect_model_forward_time=False, collect_model_execute_time=False), seed=0, served_model_name=DeepSeek-R1-Distill-Qwen-7B, use_v2_block_manager=False, num_scheduler_steps=1, multi_step_stream_outputs=False, enable_prefix_caching=False, use_async_output_proc=False, use_cached_outputs=True, mm_processor_kwargs=None)
INFO 02-07 17:33:03 selector.py:121] Using ROCmFlashAttention backend.
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0207 17:33:03.805110 113 ProcessGroupNCCL.cpp:821] [PG 0 Rank 0] ProcessGroupNCCL initialization options: NCCL version: 2.18.3, size: 1, global rank: 0, TORCH_NCCL_ASYNC_ERROR_HANDLING: 3, TORCH_NCCL_DUMP_ON_TIMEOUT: 0, TORCH_NCCL_WAIT_TIMEOUT_DUMP_MILSEC: 60000, TORCH_NCCL_DESYNC_DEBUG: 0, TORCH_NCCL_ENABLE_TIMING: 0, TORCH_NCCL_BLOCKING_WAIT: 0, TIMEOUT(ms): 600000, USE_HIGH_PRIORITY_STREAM: 0, SPLIT_FROM: 0, SPLIT_COLOR: 0, TORCH_DISTRIBUTED_DEBUG: OFF, TORCH_NCCL_ENABLE_MONITORING: 1, TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: 600, TORCH_NCCL_TRACE_BUFFER_SIZE: 0, TORCH_NCCL_COORD_CHECK_MILSEC: 1000, ID=407042000
I0207 17:33:03.805871 113 ProcessGroupNCCL.cpp:821] [PG 1 Rank 0] ProcessGroupNCCL initialization options: NCCL version: 2.18.3, size: 1, global rank: 0, TORCH_NCCL_ASYNC_ERROR_HANDLING: 3, TORCH_NCCL_DUMP_ON_TIMEOUT: 0, TORCH_NCCL_WAIT_TIMEOUT_DUMP_MILSEC: 60000, TORCH_NCCL_DESYNC_DEBUG: 0, TORCH_NCCL_ENABLE_TIMING: 0, TORCH_NCCL_BLOCKING_WAIT: 0, TIMEOUT(ms): 600000, USE_HIGH_PRIORITY_STREAM: 0, SPLIT_FROM: 0x1842f7d0, SPLIT_COLOR: 3389850942126204093, TORCH_DISTRIBUTED_DEBUG: OFF, TORCH_NCCL_ENABLE_MONITORING: 1, TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: 600, TORCH_NCCL_TRACE_BUFFER_SIZE: 0, TORCH_NCCL_COORD_CHECK_MILSEC: 1000, ID=407331248
I0207 17:33:03.818078 113 ProcessGroupNCCL.cpp:821] [PG 2 Rank 0] ProcessGroupNCCL initialization options: NCCL version: 2.18.3, size: 1, global rank: 0, TORCH_NCCL_ASYNC_ERROR_HANDLING: 3, TORCH_NCCL_DUMP_ON_TIMEOUT: 0, TORCH_NCCL_WAIT_TIMEOUT_DUMP_MILSEC: 60000, TORCH_NCCL_DESYNC_DEBUG: 0, TORCH_NCCL_ENABLE_TIMING: 0, TORCH_NCCL_BLOCKING_WAIT: 0, TIMEOUT(ms): 600000, USE_HIGH_PRIORITY_STREAM: 0, SPLIT_FROM: 0x1842f7d0, SPLIT_COLOR: 3389850942126204093, TORCH_DISTRIBUTED_DEBUG: OFF, TORCH_NCCL_ENABLE_MONITORING: 1, TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: 600, TORCH_NCCL_TRACE_BUFFER_SIZE: 0, TORCH_NCCL_COORD_CHECK_MILSEC: 1000, ID=396030864
I0207 17:33:03.820130 113 ProcessGroupNCCL.cpp:821] [PG 3 Rank 0] ProcessGroupNCCL initialization options: NCCL version: 2.18.3, size: 1, global rank: 0, TORCH_NCCL_ASYNC_ERROR_HANDLING: 3, TORCH_NCCL_DUMP_ON_TIMEOUT: 0, TORCH_NCCL_WAIT_TIMEOUT_DUMP_MILSEC: 60000, TORCH_NCCL_DESYNC_DEBUG: 0, TORCH_NCCL_ENABLE_TIMING: 0, TORCH_NCCL_BLOCKING_WAIT: 0, TIMEOUT(ms): 600000, USE_HIGH_PRIORITY_STREAM: 0, SPLIT_FROM: 0x1842f7d0, SPLIT_COLOR: 3389850942126204093, TORCH_DISTRIBUTED_DEBUG: OFF, TORCH_NCCL_ENABLE_MONITORING: 1, TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: 600, TORCH_NCCL_TRACE_BUFFER_SIZE: 0, TORCH_NCCL_COORD_CHECK_MILSEC: 1000, ID=364555120
INFO 02-07 17:33:03 model_runner.py:1041] Starting to load model DeepSeek-R1-Distill-Qwen-7B...
INFO 02-07 17:33:03 selector.py:121] Using ROCmFlashAttention backend.
Loading safetensors checkpoint shards: 0% Completed | 0/2 [00:00<?, ?it/s]
Loading safetensors checkpoint shards: 50% Completed | 1/2 [00:01<00:01, 1.69s/it]
Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:03<00:00, 1.54s/it]
Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:03<00:00, 1.56s/it]

INFO 02-07 17:33:08 model_runner.py:1052] Loading model weights took 14.2716 GB
INFO 02-07 17:33:11 model_runner_base.py:120] Writing input of failed execution to /tmp/err_execute_model_input_20250207-173311.pkl...
INFO 02-07 17:33:11 model_runner_base.py:149] Completed writing input of failed execution to /tmp/err_execute_model_input_20250207-173311.pkl.
Process SpawnProcess-1:
Traceback (most recent call last):
File "/opt/conda/lib/python3.10/site-packages/vllm/worker/model_runner_base.py", line 116, in _wrapper
return func(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/worker/model_runner.py", line 1622, in execute_model
hidden_or_intermediate_states = model_executable(
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/model_executor/models/qwen2.py", line 400, in forward
hidden_states = self.model(input_ids, positions, kv_caches,
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/model_executor/models/qwen2.py", line 297, in forward
hidden_states, residual = layer(
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/model_executor/models/qwen2.py", line 222, in forward
hidden_states = self.self_attn(
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/model_executor/models/qwen2.py", line 169, in forward
attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/attention/layer.py", line 98, in forward
return self.impl.forward(query,
File "/opt/conda/lib/python3.10/site-packages/vllm/attention/backends/rocm_flash_attn.py", line 572, in forward
out = self.attn_func(
File "/opt/conda/lib/python3.10/site-packages/flash_attn/flash_attn_interface.py", line 1125, in flash_attn_varlen_func
return FlashAttnVarlenFunc.apply(
File "/opt/conda/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
return super().apply(*args, **kwargs) # type: ignore[misc]
File "/opt/conda/lib/python3.10/site-packages/flash_attn/flash_attn_interface.py", line 621, in forward
out, q, k, v, out_padded, softmax_lse, S_dmask, rng_state = _flash_attn_varlen_forward(
File "/opt/conda/lib/python3.10/site-packages/flash_attn/flash_attn_interface.py", line 89, in _flash_attn_varlen_forward
out, q, k, v, out_padded, softmax_lse, S_dmask, rng_state = flash_attn_cuda.varlen_fwd(
RuntimeError: HIP error: invalid device function
HIP kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing AMD_SERIALIZE_KERNEL=3.
Compile with TORCH_USE_HIP_DSA  to enable device-side assertions.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "/opt/conda/lib/python3.10/multiprocessing/process.py", line 314, in _bootstrap
self.run()
File "/opt/conda/lib/python3.10/multiprocessing/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/engine/multiprocessing/engine.py", line 388, in run_mp_engine
engine = MQLLMEngine.from_engine_args(engine_args=engine_args,
File "/opt/conda/lib/python3.10/site-packages/vllm/engine/multiprocessing/engine.py", line 138, in from_engine_args
return cls(
File "/opt/conda/lib/python3.10/site-packages/vllm/engine/multiprocessing/engine.py", line 78, in init
self.engine = LLMEngine(*args,
File "/opt/conda/lib/python3.10/site-packages/vllm/engine/llm_engine.py", line 339, in init
self._initialize_kv_caches()
File "/opt/conda/lib/python3.10/site-packages/vllm/engine/llm_engine.py", line 474, in _initialize_kv_caches
self.model_executor.determine_num_available_blocks())
File "/opt/conda/lib/python3.10/site-packages/vllm/executor/gpu_executor.py", line 114, in determine_num_available_blocks
return self.driver_worker.determine_num_available_blocks()
File "/opt/conda/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/worker/worker.py", line 223, in determine_num_available_blocks
self.model_runner.profile_run()
File "/opt/conda/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/worker/model_runner.py", line 1263, in profile_run
self.execute_model(model_input, kv_caches, intermediate_tensors)
File "/opt/conda/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/opt/conda/lib/python3.10/site-packages/vllm/worker/model_runner_base.py", line 152, in _wrapper
raise type(err)(
RuntimeError: Error in model execution (input dumped to /tmp/err_execute_model_input_20250207-173311.pkl): HIP error: invalid device function
HIP kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing AMD_SERIALIZE_KERNEL=3.
Compile with TORCH_USE_HIP_DSA  to enable device-side assertions.

I0207 17:33:12.397046 113 ProcessGroupNCCL.cpp:1126] [PG 0 Rank 0] ProcessGroupNCCL destructor entered.
I0207 17:33:12.397102 113 ProcessGroupNCCL.cpp:1111] [PG 0 Rank 0] Launching ProcessGroupNCCL abort asynchrounously.
I0207 17:33:12.397302 113 ProcessGroupNCCL.cpp:1016] [PG 0 Rank 0] future is successfully executed for: ProcessGroup abort
I0207 17:33:12.397315 113 ProcessGroupNCCL.cpp:1117] [PG 0 Rank 0] ProcessGroupNCCL aborts successfully.
I0207 17:33:12.397392 113 ProcessGroupNCCL.cpp:1149] [PG 0 Rank 0] ProcessGroupNCCL watchdog thread joined.
I0207 17:33:12.397584 113 ProcessGroupNCCL.cpp:1153] [PG 0 Rank 0] ProcessGroupNCCL heart beat monitor thread joined.
Traceback (most recent call last):
File "/opt/conda/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/opt/conda/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/opt/conda/lib/python3.10/site-packages/vllm/entrypoints/openai/api_server.py", line 571, in <module> 
uvloop.run(run_server(args))
File "/opt/conda/lib/python3.10/site-packages/uvloop/init.py", line 82, in run
return loop.run_until_complete(wrapper())
File "uvloop/loop.pyx", line 1518, in uvloop.loop.Loop.run_until_complete
File "/opt/conda/lib/python3.10/site-packages/uvloop/init.py", line 61, in wrapper
return await main
File "/opt/conda/lib/python3.10/site-packages/vllm/entrypoints/openai/api_server.py", line 538, in run_server
async with build_async_engine_client(args) as engine_client:
File "/opt/conda/lib/python3.10/contextlib.py", line 199, in aenter
return await anext(self.gen)
File "/opt/conda/lib/python3.10/site-packages/vllm/entrypoints/openai/api_server.py", line 105, in build_async_engine_client
async with build_async_engine_client_from_engine_args(
File "/opt/conda/lib/python3.10/contextlib.py", line 199, in aenter
return await anext(self.gen)
File "/opt/conda/lib/python3.10/site-packages/vllm/entrypoints/openai/api_server.py", line 192, in build_async_engine_client_from_engine_args
raise RuntimeError(
RuntimeError: Engine process failed to start 

在社区搜索发现相似问题，说是Z100L不支持vllm0.6.2需要降级，我在DAS下重新下载了低版本的vllm然后再容器内pip install vllm0.5.0结果报错更多了，不确定方向是否正确，希望能得到帮助。感谢！