[Bug]: DeepSeek-R1-AWQ broken in nightly

[eldarkurtic]

Your current environment

Model: cognitivecomputations/DeepSeek-R1-AWQ
Served on 8xH100s: vllm serve "cognitivecomputations/DeepSeek-R1-AWQ" -tp 8 --gpu-memory-utilization 0.8 --max-model-len 4096 --enable-chunked-prefill --trust-remote-code --max-num-batched-tokens 4096 --dtype float16 --port 1234
Prompted with:

curl -X POST http://localhost:1234/v1/chat/completions   -H "Content-Type: application/json"   -d '{
    "model": "cognitivecomputations/DeepSeek-R1-AWQ",
    "messages": [
      {"role": "user", "content": "What is the capital of France?"}
    ],
    "temperature": 0.7,
    "max_tokens": 512
  }'

🐛 Describe the bug

In vllm==0.7.3, everything looks good:

{"id":"chatcmpl-acf69072a2a545c4baedcf93efbde64e","object":"chat.completion","created":1742275076,"model":"local-awq","choices":[{"index":0,"message":{"role":"assistant","reasoning_content":null,"content":"Okay, so the question is asking for the capital of France. Let me think. I know France is a country in Europe. I've heard of Paris being mentioned a lot in relation to France. Wait, is Paris the capital? I remember seeing the Eiffel Tower in Paris, so maybe that's the capital. But let me double-check in my mind. Are there other major cities in France that could be the capital? Lyon, Marseille, maybe Toulouse? No, I think Paris is the most well-known. Also, I think the government and president are based in Paris. Yeah, I'm pretty sure it's Paris.\n</think>\n\nThe capital of France is **Paris**. Known for its iconic landmarks such as the Eiffel Tower, Louvre Museum, and Notre-Dame Cathedral, Paris serves as the country's political, cultural, and economic center. 🇫🇷","tool_calls":[]},"logprobs":null,"finish_reason":"stop","stop_reason":null}],"usage":{"prompt_tokens":12,"total_tokens":192,"completion_tokens":180,"prompt_tokens_details":null},"prompt_logprobs":null}

In vllm==v0.8.0rc3.dev5+g5eeabc2a (V1) the model produces garbage:

{"id":"chatcmpl-b6e33ee4c1d64a3ab785cd54e2ed11fa","object":"chat.completion","created":1742274408,"model":"local-awq","choices":[{"index":0,"message":{"role":"assistant","reasoning_content":null,"content":"<think>\n\n Bible_\n\n.<think>\n\n* R<think> down\n\n** \n\n\n\n\n<think>\n\n<think></think><think><think></think><think></think><think>\n\n�</think><｜place▁holder▁no▁9｜><think><think><think></think><think></think><think></think><think><｜place▁holder▁no▁772｜><think>\n\n�</think><｜place▁holder▁no▁5｜><think><think></think><think></think><think></think><｜place▁holder▁no▁796｜><think><think><｜place▁holder▁no▁7｜>⠀<think><｜place▁holder▁no▁9｜><think><think></think>\n<think>\n\n�\n\n##</think>\n<think></think><think></think>\n\n�\n</think><think></think>\n\n><think>\n\n<think>\n\n<think>\n\n</think>\n\n<think></think><think></think>\n\n>Hey?\n\n</think><think><｜place▁holder▁no▁791｜><think>\n\n<think>\n\n\n\n</think><think>\n\n<｜place▁holder▁no▁795｜>\n\n<think>\n</think>\n\n<think>\n\n<think></think>**<｜place▁holder▁no▁795｜>\n\n<｜place▁holder▁no▁14｜>\n\n<｜place▁holder▁no▁795｜>\n\n<｜place▁holder▁no▁370｜>\n<｜place▁holder▁no▁795｜>\n\n<｜place▁holder▁no▁795｜>\n\n\n\n\n\n<think></\n\n<think>\n\n<｜place▁holder▁no▁20｜><think><think><think><think></think></think></think><think>\n\n<｜place▁holder▁no▁795｜>\n\n<｜place▁holder▁no▁284｜>\n\n\n\n<｜place▁holder▁no▁792｜>.<think></think><｜place▁holder▁no▁793｜></think></think><think><｜place▁holder▁no▁10｜><think></think><think>\n-<｜place▁holder▁no▁794｜>;\\<think></think><think></think>\n\n\n\n\n<｜place▁holder▁no▁765｜><｜place▁holder▁no▁793｜>endum<｜place▁holder▁no▁793｜></think><think><｜place▁holder▁no▁794｜><｜place▁holder▁no▁795｜><｜place▁holder▁no▁794｜>\n\n\n:\n\n�\n\n<think>\n\n\n</think><think><｜place▁holder▁no▁772｜><think><｜place▁holder▁no▁797｜>\n\n<｜place▁holder▁no▁10｜><think></think><think></think><think>\n\n<｜place▁holder▁no▁558｜></think><think></think></think><｜place▁holder▁no▁795｜><think><｜place▁holder▁no▁795｜><think>\n\n°<｜place▁holder▁no▁793｜><think></think><｜place▁holder▁no▁793｜><think></think><think>\nH;**\n\n<think><｜place▁holder▁no▁11｜><｜place▁holder▁no▁794｜><think>®<think><think><｜place▁holder▁no▁791｜></think><｜place▁holder▁no▁29｜><think>\n<think><｜place▁holder▁no▁792｜><think><｜place▁holder▁no▁792｜><think><｜place▁holder▁no▁5｜></think><｜place▁holder▁no▁147｜><｜place▁holder▁no▁794｜>\n\nH<｜place▁holder▁no▁795｜><think></think><｜place▁holder▁no▁365｜>\n\n,<think>\n<｜place▁holder▁no▁794｜>;\n\n</think><think><think>\n\n<｜place▁holder▁no▁792｜><think><｜place▁holder▁no▁793｜>;<think><｜place▁holder▁no▁793｜>\n\n<｜place▁holder▁no▁10｜>,\n\n</think><think>;<｜place▁holder▁no▁793｜>\n\\u\n\n</think><｜place▁holder▁no▁793｜><｜place▁holder▁no▁793｜><｜place▁holder▁no▁794｜>\n\nThe\n\n9<｜place▁holder▁no▁793｜><think></think><｜place▁holder▁no▁793｜>\n\n<｜place▁holder▁no▁795｜>,<｜place▁holder▁no▁793｜></think>\n\n**<｜place▁holder▁no▁793｜><think>\n\n\\<｜place▁holder▁no▁793｜><｜place▁holder▁no▁795｜><｜place▁holder▁no▁793｜>\n\n><think><｜place▁holder▁no▁794｜><｜place▁holder▁no▁792｜><think>\n\n<｜place▁holder▁no▁792｜>\n\n<｜place▁holder▁no▁795｜>\n\n<｜place▁holder▁no▁10｜>\n\n,<｜place▁holder▁no▁793｜>  \n  \n,<｜place▁holder▁no▁793｜><\n\n>,<｜place▁holder▁no▁793｜>\n\n><think>\n\n>,; hydroxide</think><think>\n\n\"\n\n<｜place▁holder▁no▁761｜>\n\n°.<｜place▁holder▁no▁775｜><｜place▁holder▁no▁793｜>endum\n\n,<think></think><｜place▁holder▁no▁795｜>&#\n\n\n<｜place▁holder▁no▁40｜><｜place▁holder▁no▁791｜></think><think></think><｜place▁holder▁no▁793｜>\n\n,\n-<｜place▁holder▁no▁796｜><｜place▁holder▁no▁793｜></think>\n\n.<｜place▁holder▁no▁27｜> \n\n</\n\n</think><｜place▁holder▁no▁792｜>&#R<｜place▁holder▁no▁792｜><think><｜place▁holder▁no▁796｜><｜place▁holder▁no▁27｜>\n\n_\n\n \n</think><｜place▁holder▁no▁793｜>°,</think><think>\n\n><｜place▁holder▁no▁793｜>\n\n,�<｜place▁holder▁no▁793｜>\n\n -</think><｜place▁holder▁no▁59｜></think><｜place▁holder▁no▁793｜>\n\n⁻<｜place▁holder▁no▁792｜><｜place▁holder▁no▁793｜><｜place▁holder▁no▁10｜><think></think><｜place▁holder▁no▁797｜>,\n\n.**</think>\t<think><｜place▁holder▁no▁793｜><｜place▁holder▁no▁793｜>,<think>\n\n<｜place▁holder▁no▁170｜>\u0007</think><｜place▁holder▁no▁793｜>\n\n,<｜place▁holder▁no▁793｜>:>\n\n</think><think>\n\n,<｜place▁holder▁no▁795｜><｜place▁holder▁no▁795｜><｜place▁holder▁no▁794｜></think>\n\n,:<｜place▁holder▁no▁793｜>:<think></think><think>\n\n,.\n\n,\n\n,\n\n,\n\n,\n\n,.ate\n\n;<｜place▁holder▁no▁795｜>","tool_calls":[]},"logprobs":null,"finish_reason":"stop","stop_reason":null}],"usage":{"prompt_tokens":12,"total_tokens":484,"completion_tokens":472,"prompt_tokens_details":null},"prompt_logprobs":null}

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[eldarkurtic]

FYI: I've also tried vllm==v0.8.0rc3.dev5+g5eeabc2a(V0 with VLLM_USE_V1=0), and in this case it can't even load the model.

❯ VLLM_USE_V1=0 vllm serve "local-awq"  -tp 8 --gpu-memory-utilization 0.8 --max-model-len 4096 --enable-chunked-prefill --trust-remote-code --max-num-batched-tokens 4096 --dtype float16 --port 1234
INFO 03-18 05:28:45 [__init__.py:256] Automatically detected platform cuda.
INFO 03-18 05:28:46 [api_server.py:972] vLLM API server version 0.8.0rc3.dev5+g5eeabc2a
INFO 03-18 05:28:46 [api_server.py:973] args: Namespace(subparser='serve', model_tag='local-awq', config='', host=None, port=1234, 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, chat_template_content_format='auto', response_role='assistant', ssl_keyfile=None, ssl_certfile=None, ssl_ca_certs=None, enable_ssl_refresh=False, ssl_cert_reqs=0, root_path=None, middleware=[], return_tokens_as_token_ids=False, disable_frontend_multiprocessing=False, enable_request_id_headers=False, enable_auto_tool_choice=False, tool_call_parser=None, tool_parser_plugin='', model='local-awq', task='auto', tokenizer=None, hf_config_path=None, skip_tokenizer_init=False, revision=None, code_revision=None, tokenizer_revision=None, tokenizer_mode='auto', trust_remote_code=True, allowed_local_media_path=None, download_dir=None, load_format='auto', config_format=<ConfigFormat.AUTO: 'auto'>, dtype='float16', kv_cache_dtype='auto', max_model_len=4096, guided_decoding_backend='xgrammar', logits_processor_pattern=None, model_impl='auto', distributed_executor_backend=None, pipeline_parallel_size=1, tensor_parallel_size=8, enable_expert_parallel=False, max_parallel_loading_workers=None, ray_workers_use_nsight=False, block_size=None, enable_prefix_caching=None, disable_sliding_window=False, use_v2_block_manager=True, num_lookahead_slots=0, seed=None, swap_space=4, cpu_offload_gb=0, gpu_memory_utilization=0.8, num_gpu_blocks_override=None, max_num_batched_tokens=4096, max_num_partial_prefills=1, max_long_partial_prefills=1, long_prefill_token_threshold=0, max_num_seqs=None, max_logprobs=20, disable_log_stats=False, quantization=None, rope_scaling=None, rope_theta=None, hf_overrides=None, enforce_eager=False, 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, disable_mm_preprocessor_cache=False, enable_lora=False, enable_lora_bias=False, max_loras=1, max_lora_rank=16, 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, use_tqdm_on_load=True, multi_step_stream_outputs=True, scheduler_delay_factor=0.0, enable_chunked_prefill=True, speculative_model=None, speculative_model_quantization=None, num_speculative_tokens=None, speculative_disable_mqa_scorer=False, 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', 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, show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, disable_async_output_proc=False, scheduling_policy='fcfs', scheduler_cls='vllm.core.scheduler.Scheduler', override_neuron_config=None, override_pooler_config=None, compilation_config=None, kv_transfer_config=None, worker_cls='auto', worker_extension_cls='', generation_config='auto', override_generation_config=None, enable_sleep_mode=False, calculate_kv_scales=False, additional_config=None, enable_reasoning=False, reasoning_parser=None, disable_log_requests=False, max_log_len=None, disable_fastapi_docs=False, enable_prompt_tokens_details=False, enable_server_load_tracking=False, dispatch_function=<function ServeSubcommand.cmd at 0x7fab3e78d440>)
INFO 03-18 05:28:46 [config.py:208] Replacing legacy 'type' key with 'rope_type'
WARNING 03-18 05:28:46 [config.py:2583] Casting torch.bfloat16 to torch.float16.
INFO 03-18 05:28:52 [config.py:583] This model supports multiple tasks: {'generate', 'classify', 'score', 'reward', 'embed'}. Defaulting to 'generate'.
INFO 03-18 05:28:53 [awq_marlin.py:114] The model is convertible to awq_marlin during runtime. Using awq_marlin kernel.
INFO 03-18 05:28:53 [config.py:1499] Defaulting to use mp for distributed inference
INFO 03-18 05:28:53 [config.py:1677] Chunked prefill is enabled with max_num_batched_tokens=4096.
INFO 03-18 05:28:53 [cuda.py:159] Forcing kv cache block size to 64 for FlashMLA backend.
INFO 03-18 05:28:53 [api_server.py:236] Started engine process with PID 3278262
INFO 03-18 05:28:56 [__init__.py:256] Automatically detected platform cuda.
INFO 03-18 05:28:59 [llm_engine.py:241] Initializing a V0 LLM engine (v0.8.0rc3.dev5+g5eeabc2a) with config: model='local-awq', speculative_config=None, tokenizer='local-awq', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.float16, max_seq_len=4096, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=8, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=awq_marlin, enforce_eager=False, kv_cache_dtype=auto,  device_config=cuda, decoding_config=DecodingConfig(guided_decoding_backend='xgrammar', reasoning_backend=None), observability_config=ObservabilityConfig(show_hidden_metrics=False, otlp_traces_endpoint=None, collect_model_forward_time=False, collect_model_execute_time=False), seed=None, served_model_name=local-awq, num_scheduler_steps=1, multi_step_stream_outputs=True, enable_prefix_caching=None, chunked_prefill_enabled=True, use_async_output_proc=True, disable_mm_preprocessor_cache=False, mm_processor_kwargs=None, pooler_config=None, compilation_config={"splitting_ops":[],"compile_sizes":[],"cudagraph_capture_sizes":[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],"max_capture_size":256}, use_cached_outputs=True,
WARNING 03-18 05:28:59 [multiproc_worker_utils.py:310] Reducing Torch parallelism from 192 threads to 1 to avoid unnecessary CPU contention. Set OMP_NUM_THREADS in the external environment to tune this value as needed.
INFO 03-18 05:28:59 [custom_cache_manager.py:19] Setting Triton cache manager to: vllm.triton_utils.custom_cache_manager:CustomCacheManager
ERROR 03-18 05:28:59 [engine.py:443] Can't pickle <class 'transformers_modules.local-awq.configuration_deepseek.DeepseekV3Config'>: it's not the same object as transformers_modules.local-awq.configuration_deepseek.DeepseekV3Config
ERROR 03-18 05:28:59 [engine.py:443] Traceback (most recent call last):
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/multiprocessing/engine.py", line 431, in run_mp_engine
ERROR 03-18 05:28:59 [engine.py:443]     engine = MQLLMEngine.from_vllm_config(
ERROR 03-18 05:28:59 [engine.py:443]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/multiprocessing/engine.py", line 126, in from_vllm_config
ERROR 03-18 05:28:59 [engine.py:443]     return cls(
ERROR 03-18 05:28:59 [engine.py:443]            ^^^^
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/multiprocessing/engine.py", line 80, in __init__
ERROR 03-18 05:28:59 [engine.py:443]     self.engine = LLMEngine(*args, **kwargs)
ERROR 03-18 05:28:59 [engine.py:443]                   ^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/llm_engine.py", line 280, in __init__
ERROR 03-18 05:28:59 [engine.py:443]     self.model_executor = executor_class(vllm_config=vllm_config, )
ERROR 03-18 05:28:59 [engine.py:443]                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/executor_base.py", line 271, in __init__
ERROR 03-18 05:28:59 [engine.py:443]     super().__init__(*args, **kwargs)
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/executor_base.py", line 52, in __init__
ERROR 03-18 05:28:59 [engine.py:443]     self._init_executor()
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/mp_distributed_executor.py", line 90, in _init_executor
ERROR 03-18 05:28:59 [engine.py:443]     worker = ProcessWorkerWrapper(result_handler,
ERROR 03-18 05:28:59 [engine.py:443]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 03-18 05:28:59 [engine.py:443]   File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/multiproc_worker_utils.py", line 171, in __init__
ERROR 03-18 05:28:59 [engine.py:443]     self.process.start()
ERROR 03-18 05:28:59 [engine.py:443]   File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 121, in start
ERROR 03-18 05:28:59 [engine.py:443]     self._popen = self._Popen(self)
ERROR 03-18 05:28:59 [engine.py:443]                   ^^^^^^^^^^^^^^^^^
ERROR 03-18 05:28:59 [engine.py:443]   File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/context.py", line 289, in _Popen
ERROR 03-18 05:28:59 [engine.py:443]     return Popen(process_obj)
ERROR 03-18 05:28:59 [engine.py:443]            ^^^^^^^^^^^^^^^^^^
ERROR 03-18 05:28:59 [engine.py:443]   File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_spawn_posix.py", line 32, in __init__
ERROR 03-18 05:28:59 [engine.py:443]     super().__init__(process_obj)
ERROR 03-18 05:28:59 [engine.py:443]   File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_fork.py", line 19, in __init__
ERROR 03-18 05:28:59 [engine.py:443]     self._launch(process_obj)
ERROR 03-18 05:28:59 [engine.py:443]   File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_spawn_posix.py", line 47, in _launch
ERROR 03-18 05:28:59 [engine.py:443]     reduction.dump(process_obj, fp)
ERROR 03-18 05:28:59 [engine.py:443]   File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/reduction.py", line 60, in dump
ERROR 03-18 05:28:59 [engine.py:443]     ForkingPickler(file, protocol).dump(obj)
ERROR 03-18 05:28:59 [engine.py:443] _pickle.PicklingError: Can't pickle <class 'transformers_modules.local-awq.configuration_deepseek.DeepseekV3Config'>: it's not the same object as transformers_modules.local-awq.configuration_deepseek.DeepseekV3Config
Process SpawnProcess-1:
Traceback (most recent call last):
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/multiprocessing/engine.py", line 445, in run_mp_engine
    raise e
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/multiprocessing/engine.py", line 431, in run_mp_engine
    engine = MQLLMEngine.from_vllm_config(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/multiprocessing/engine.py", line 126, in from_vllm_config
    return cls(
           ^^^^
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/multiprocessing/engine.py", line 80, in __init__
    self.engine = LLMEngine(*args, **kwargs)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/engine/llm_engine.py", line 280, in __init__
    self.model_executor = executor_class(vllm_config=vllm_config, )
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/executor_base.py", line 271, in __init__
    super().__init__(*args, **kwargs)
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/executor_base.py", line 52, in __init__
    self._init_executor()
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/mp_distributed_executor.py", line 90, in _init_executor
    worker = ProcessWorkerWrapper(result_handler,
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/nvme2/eldar/hf_home/test/lib/python3.12/site-packages/vllm/executor/multiproc_worker_utils.py", line 171, in __init__
    self.process.start()
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 121, in start
    self._popen = self._Popen(self)
                  ^^^^^^^^^^^^^^^^^
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/context.py", line 289, in _Popen
    return Popen(process_obj)
           ^^^^^^^^^^^^^^^^^^
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_spawn_posix.py", line 32, in __init__
    super().__init__(process_obj)
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_fork.py", line 19, in __init__
    self._launch(process_obj)
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_spawn_posix.py", line 47, in _launch
    reduction.dump(process_obj, fp)
  File "/home/eldar/.local/share/uv/python/cpython-3.12.9-linux-x86_64-gnu/lib/python3.12/multiprocessing/reduction.py", line 60, in dump
    ForkingPickler(file, protocol).dump(obj)
_pickle.PicklingError: Can't pickle <class 'transformers_modules.local-awq.configuration_deepseek.DeepseekV3Config'>: it's not the same object as transformers_modules.local-awq.configuration_deepseek.DeepseekV3Config

[NaiveYan]

Same issue occurs in both the v0.8.0 and v0.8.1 release versions.

[wangjiyang]

Same issue here.