(APIServer pid=2742420) INFO 08-16 04:17:56 [utils.py:299] (APIServer pid=2742420) INFO 08-16 04:17:56 [utils.py:299] █ █ █▄ ▄█ (APIServer pid=2742420) INFO 08-16 04:17:56 [utils.py:299] ▄▄ ▄█ █ █ █ ▀▄▀ █ version 0.19.0 (APIServer pid=2742420) INFO 08-16 04:17:56 [utils.py:299] █▄█▀ █ █ █ █ model Qwen/Qwen3.5-4B (APIServer pid=2742420) INFO 08-16 04:17:56 [utils.py:299] ▀▀ ▀▀▀▀▀ ▀▀▀▀▀ ▀ ▀ (APIServer pid=2742420) INFO 08-16 04:17:56 [utils.py:299] (APIServer pid=2742420) INFO 08-16 04:17:56 [utils.py:233] non-default args: {'model_tag': 'Qwen/Qwen3.5-4B', 'lora_modules': [LoRAModulePath(name='exec', path='runs_stage1/exec_v1', base_model_name=None)], 'host': '127.0.0.1', 'port': 8005, 'model': 'Qwen/Qwen3.5-4B', 'max_model_len': 16384, 'enforce_eager': True, 'gpu_memory_utilization': 0.5, 'enable_lora': True, 'max_lora_rank': 32} (APIServer pid=2742420) INFO 08-16 04:17:56 [arg_utils.py:665] HF_HUB_OFFLINE is True, replace model_id [Qwen/Qwen3.5-4B] to model_path [/home/lukas/.cache/huggingface/hub/models--Qwen--Qwen3.5-4B/snapshots/851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a] (APIServer pid=2742420) INFO 08-16 04:17:56 [model.py:549] Resolved architecture: Qwen3_5ForConditionalGeneration (APIServer pid=2742420) INFO 08-16 04:17:56 [model.py:1678] Using max model len 16384 (APIServer pid=2742420) [transformers] `Qwen2VLImageProcessorFast` is deprecated. The `Fast` suffix for image processors has been removed; use `Qwen2VLImageProcessor` instead. (APIServer pid=2742420) INFO 08-16 04:17:57 [config.py:281] Setting attention block size to 528 tokens to ensure that attention page size is >= mamba page size. (APIServer pid=2742420) INFO 08-16 04:17:57 [config.py:312] Padding mamba page size by 0.76% to ensure that mamba page size and attention page size are exactly equal. (APIServer pid=2742420) INFO 08-16 04:17:57 [vllm.py:790] Asynchronous scheduling is enabled. (APIServer pid=2742420) WARNING 08-16 04:17:57 [vllm.py:848] Enforce eager set, disabling torch.compile and CUDAGraphs. This is equivalent to setting -cc.mode=none -cc.cudagraph_mode=none (APIServer pid=2742420) WARNING 08-16 04:17:57 [vllm.py:859] Inductor compilation was disabled by user settings, optimizations settings that are only active during inductor compilation will be ignored. (APIServer pid=2742420) INFO 08-16 04:17:57 [vllm.py:1025] Cudagraph is disabled under eager mode (APIServer pid=2742420) INFO 08-16 04:17:57 [compilation.py:290] Enabled custom fusions: norm_quant, act_quant (APIServer pid=2742420) [transformers] The `use_fast` parameter is deprecated and will be removed in a future version. Use `backend="torchvision"` instead of `use_fast=True`, or `backend="pil"` instead of `use_fast=False`. (EngineCore pid=2743086) INFO 08-16 04:18:21 [core.py:105] Initializing a V1 LLM engine (v0.19.0) with config: model='/home/lukas/.cache/huggingface/hub/models--Qwen--Qwen3.5-4B/snapshots/851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a', speculative_config=None, tokenizer='/home/lukas/.cache/huggingface/hub/models--Qwen--Qwen3.5-4B/snapshots/851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=16384, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, decode_context_parallel_size=1, dcp_comm_backend=ag_rs, disable_custom_all_reduce=False, quantization=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', 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, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=0, served_model_name=/home/lukas/.cache/huggingface/hub/models--Qwen--Qwen3.5-4B/snapshots/851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a, enable_prefix_caching=False, enable_chunked_prefill=True, pooler_config=None, compilation_config={'mode': , 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'splitting_ops': [], 'compile_mm_encoder': False, 'cudagraph_mm_encoder': False, 'encoder_cudagraph_token_budgets': [], 'encoder_cudagraph_max_images_per_batch': 0, 'compile_sizes': [], 'compile_ranges_endpoints': [2048], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'size_asserts': False, 'alignment_asserts': False, 'scalar_asserts': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': , 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': True, 'fuse_act_quant': True, 'fuse_attn_quant': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False}, 'max_cudagraph_capture_size': 0, 'dynamic_shapes_config': {'type': , 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []} (EngineCore pid=2743086) [transformers] `Qwen2VLImageProcessorFast` is deprecated. The `Fast` suffix for image processors has been removed; use `Qwen2VLImageProcessor` instead. (EngineCore pid=2743086) INFO 08-16 04:18:23 [parallel_state.py:1400] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://192.168.62.173:37405 backend=nccl (EngineCore pid=2743086) INFO 08-16 04:18:23 [parallel_state.py:1716] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank N/A, EPLB rank N/A (EngineCore pid=2743086) [transformers] The `use_fast` parameter is deprecated and will be removed in a future version. Use `backend="torchvision"` instead of `use_fast=True`, or `backend="pil"` instead of `use_fast=False`. (EngineCore pid=2743086) INFO 08-16 04:18:30 [gpu_model_runner.py:4735] Starting to load model /home/lukas/.cache/huggingface/hub/models--Qwen--Qwen3.5-4B/snapshots/851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a... (EngineCore pid=2743086) INFO 08-16 04:18:30 [cuda.py:390] Using backend AttentionBackendEnum.FLASH_ATTN for vit attention (EngineCore pid=2743086) INFO 08-16 04:18:30 [mm_encoder_attention.py:230] Using AttentionBackendEnum.FLASH_ATTN for MMEncoderAttention. (EngineCore pid=2743086) INFO 08-16 04:18:30 [gdn_linear_attn.py:147] Using Triton/FLA GDN prefill kernel (EngineCore pid=2743086) INFO 08-16 04:18:32 [cuda.py:334] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION']. (EngineCore pid=2743086) INFO 08-16 04:18:32 [flash_attn.py:596] Using FlashAttention version 2 (EngineCore pid=2743086) INFO 08-16 04:18:32 [weight_utils.py:848] Prefetching checkpoint files into page cache started (in background) (EngineCore pid=2743086) Loading safetensors checkpoint shards: 0% Completed | 0/2 [00:00