WARNING 12-09 21:18:19 [argparse_utils.py:195] With `vllm serve`, you should provide the model as a positional argument or in a config file instead of via the `--model` option. The `--model` option will be removed in v0.13. (APIServer pid=35981) INFO 12-09 21:18:19 [api_server.py:1772] vLLM API server version 0.12.0 (APIServer pid=35981) INFO 12-09 21:18:19 [utils.py:253] non-default args: {'model_tag': 'cpatonn/Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit', 'host': '127.0.0.1', 'model': 'cpatonn/Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit', 'trust_remote_code': True, 'max_model_len': 24576, 'max_num_seqs': 64} (APIServer pid=35981) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored. (APIServer pid=35981) INFO 12-09 21:18:25 [model.py:637] Resolved architecture: Qwen3MoeForCausalLM (APIServer pid=35981) INFO 12-09 21:18:25 [model.py:1750] Using max model len 24576 (APIServer pid=35981) INFO 12-09 21:18:26 [scheduler.py:228] Chunked prefill is enabled with max_num_batched_tokens=2048. (EngineCore_DP0 pid=36287) INFO 12-09 21:18:33 [core.py:93] Initializing a V1 LLM engine (v0.12.0) with config: model='cpatonn/Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit', speculative_config=None, tokenizer='cpatonn/Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=24576, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, disable_custom_all_reduce=False, quantization=compressed-tensors, enforce_eager=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01), seed=0, served_model_name=cpatonn/Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'level': None, 'mode': , 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['none'], 'splitting_ops': ['vllm::unified_attention', 'vllm::unified_attention_with_output', 'vllm::unified_mla_attention', 'vllm::unified_mla_attention_with_output', 'vllm::mamba_mixer2', 'vllm::mamba_mixer', 'vllm::short_conv', 'vllm::linear_attention', 'vllm::plamo2_mamba_mixer', 'vllm::gdn_attention_core', 'vllm::kda_attention', 'vllm::sparse_attn_indexer'], 'compile_mm_encoder': False, 'compile_sizes': [], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': , 'cudagraph_num_of_warmups': 1, 'cudagraph_capture_sizes': [1, 2, 4, 8, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120, 128], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'eliminate_noops': True, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False}, 'max_cudagraph_capture_size': 128, 'dynamic_shapes_config': {'type': }, 'local_cache_dir': None} (EngineCore_DP0 pid=36287) INFO 12-09 21:18:34 [parallel_state.py:1200] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://172.17.0.4:58697 backend=nccl (EngineCore_DP0 pid=36287) INFO 12-09 21:18:34 [parallel_state.py:1408] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank 0 (EngineCore_DP0 pid=36287) INFO 12-09 21:18:34 [gpu_model_runner.py:3467] Starting to load model cpatonn/Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit... (EngineCore_DP0 pid=36287) INFO 12-09 21:18:35 [compressed_tensors_wNa16.py:114] Using MarlinLinearKernel for CompressedTensorsWNA16 (EngineCore_DP0 pid=36287) INFO 12-09 21:18:35 [cuda.py:411] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'] (EngineCore_DP0 pid=36287) INFO 12-09 21:18:35 [layer.py:379] Enabled separate cuda stream for MoE shared_experts (EngineCore_DP0 pid=36287) INFO 12-09 21:18:35 [compressed_tensors_moe.py:167] Using CompressedTensorsWNA16MarlinMoEMethod (EngineCore_DP0 pid=36287) WARNING 12-09 21:18:35 [compressed_tensors.py:721] Acceleration for non-quantized schemes is not supported by Compressed Tensors. Falling back to UnquantizedLinearMethod (EngineCore_DP0 pid=36287) Loading safetensors checkpoint shards: 0% Completed | 0/4 [00:00