Buckets:
| (MainProcess pid=25496) INFO 10-31 00:49:49 [model.py:733] Resolved architecture: Qwen3_5ForConditionalGeneration | |
| (MainProcess pid=25496) INFO 10-31 00:49:49 [model.py:2152] Using max model len 32768 | |
| Fetching 12 files: 0%| | 0/12 [00:00, ?it/s] Fetching 12 files: 100%|██████████| 12/12 [00:00<00:00, 1166.68it/s] | |
| (MainProcess pid=25496) INFO 10-31 00:49:51 [scheduler.py:307] Chunked prefill is enabled with max_num_batched_tokens=2048. | |
| (MainProcess pid=25496) INFO 10-31 00:49:51 [config.py:688] Mamba cache mode is set to 'align' for Qwen3_5ForConditionalGeneration by default when prefix caching is enabled | |
| Parse safetensors files: 0%| | 0/2 [00:00, ?it/s] Parse safetensors files: 50%|█████ | 1/2 [00:00<00:00, 4.73it/s] Parse safetensors files: 100%|██████████| 2/2 [00:00<00:00, 7.62it/s] | |
| (MainProcess pid=25496) INFO 10-31 00:49:52 [kernel.py:467] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'], gelu_and_mul_sparse=['triton', 'native']) | |
| (MainProcess pid=25496) WARNING 10-31 00:50:03 [system_utils.py:153] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: CUDA is initialized | |
| (EngineCore pid=25921) INFO 10-31 00:50:14 [core.py:133] Initializing a V1 LLM engine (v0.32.0) with config: model='Qwen/Qwen3.5-4B', speculative_config=None, tokenizer='Qwen/Qwen3.5-4B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=main, tokenizer_revision=main, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=32768, 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, quantization_config=None, enforce_eager=False, aux_output_config=AuxOutputConfig(enable_return_routed_experts=False, max_bytes=None), 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, per_request_spec_decode_metrics='none', kv_cache_metrics=False, kv_cache_metrics_sample=0.01, custom_histogram_buckets=None, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False, jit_monitor_mode='warn', jit_monitor_verbose=False), seed=17, served_model_name=Qwen/Qwen3.5-4B, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'mode': <CompilationMode.VLLM_COMPILE: 3>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['none'], 'ir_enable_torch_wrap': True, 'splitting_ops': ['vllm::unified_attention_with_output', 'vllm::unified_mla_attention_with_output', 'vllm::mamba_mixer2', 'vllm::mamba_mixer', 'vllm::short_conv', 'vllm::qwen4_exp_ple_short_conv', 'vllm::qwen4_exp_qsa_with_output', 'vllm::qwen4_exp_qsa_prepare_with_output', 'vllm::linear_attention', 'vllm::qwen_gdn_attention_core', 'vllm::qwen_gdn_attention_core_fused_norm_packed', 'vllm::gdn_attention_core_xpu', 'vllm::olmo_hybrid_gdn_full_forward', 'vllm::sparse_attn_indexer', 'vllm::rocm_aiter_sparse_attn_indexer', 'vllm::deepseek_v4_attention', 'vllm::hpc_rope_norm_forward', 'vllm::unified_kv_cache_update', 'vllm::unified_mla_kv_cache_update'], 'compile_mm_encoder': False, 'cudagraph_mm_encoder': False, 'encoder_cudagraph_token_budgets': [], 'encoder_cudagraph_max_vision_items_per_batch': 0, 'encoder_cudagraph_max_frames_per_batch': None, 'cudagraph_decoder_replay': True, 'decoder_replay_cudagraph_capture_sizes': [], 'decoder_replay_trim_threshold': 768, 'compile_sizes': [], 'compile_ranges_endpoints': [2048], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.FULL_AND_PIECEWISE: (2, 1)>, '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, 136, 144, 152, 160, 168, 176, 184, 192, 200, 208, 216, 224, 232, 240, 248, 256], '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, 'fuse_allreduce_rms': False, 'enable_qk_norm_rope_fusion': False, 'fuse_rope_kvcache_cat_mla': False, 'fuse_act_padding': False, 'fuse_qk_norm_rope_kvcache': False}, 'max_cudagraph_capture_size': 256, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': False, 'static_all_moe_layers': []}, kernel_config=KernelConfig(ir_op_priority=IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'], gelu_and_mul_sparse=['triton', 'native']), enable_flashinfer_autotune=True, enable_rocm_segmented_attn_autotune=False, enable_cutedsl_warmup=True, enable_jit_warmup=True, moe_backend='auto', sparse_indexer_topk_backend='auto', linear_backend='auto', linear_backend_per_quant=None) | |
| (EngineCore pid=25921) INFO 10-31 00:50:15 [parallel_state.py:1906] world_size=1 rank=0 local_rank=0 distributed_init_method=file:///tmp/vllm_dist_1cdc2a918df5485caa3258a5a3b3664c backend=nccl | |
| (EngineCore pid=25921) INFO 10-31 00:50:15 [parallel_state.py:2350] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, ETP rank 0, EP rank N/A, EPLB rank N/A | |
| (EngineCore pid=25921) INFO 10-31 00:50:15 [gpu_worker.py:518] Using V2 Model Runner | |
| (EngineCore pid=25921) INFO 10-31 00:50:17 [model_runner.py:392] Loading model from scratch... | |
| Fetching 12 files: 0%| | 0/12 [00:00, ?it/s] Fetching 12 files: 100%|██████████| 12/12 [00:00<00:00, 1077.81it/s] | |
| (EngineCore pid=25921) INFO 10-31 00:50:18 [cuda.py:565] Using backend AttentionBackendEnum.FLASH_ATTN for vit attention | |
| (EngineCore pid=25921) INFO 10-31 00:50:18 [mm_encoder_attention.py:372] Using AttentionBackendEnum.FLASH_ATTN for MMEncoderAttention. | |
| (EngineCore pid=25921) INFO 10-31 00:50:18 [qwen_gdn_linear_attn.py:221] Using Triton/FLA GDN prefill kernel (requested=auto, head_k_dim=128). | |
| (EngineCore pid=25921) INFO 10-31 00:50:19 [cuda.py:506] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION']. | |
| (EngineCore pid=25921) INFO 10-31 00:50:19 [flash_attn.py:1117] Using FlashAttention version 2 | |
| (EngineCore pid=25921) INFO 10-31 00:50:19 [base_loader.py:82] Initializing model took 2.131808 seconds | |
| (EngineCore pid=25921) INFO 10-31 00:50:19 [weight_utils.py:1108] Filesystem type for checkpoints: XFS. Checkpoint size: 8.68 GiB. Available RAM: 193.47 GiB. | |
| (EngineCore pid=25921) INFO 10-31 00:50:19 [weight_utils.py:1131] Auto-prefetch is disabled because the filesystem (XFS) is not a recognized network FS (NFS/Lustre). If you want to force prefetching, start vLLM with --safetensors-load-strategy=prefetch. | |
| Loading safetensors checkpoint shards: 0% Completed | 0/2 [00:00, ?it/s] | |
| Loading safetensors checkpoint shards: 50% Completed | 1/2 [00:00<00:00, 2.67it/s] | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.0.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.1.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.10.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.12.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.13.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.14.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.16.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.17.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.18.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.2.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.20.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.21.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.22.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.24.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.25.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.26.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.28.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.29.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.30.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.4.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.5.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.6.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.8.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| (EngineCore pid=25921) WARNING 10-31 00:50:20 [utils.py:333] Attempted to load weight layers.9.linear_attn.norm.weight with dtype torch.float32 into parameter with dtype torch.bfloat16 | |
| Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:00<00:00, 2.68it/s] | |
| Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:00<00:00, 2.68it/s] | |
| (EngineCore pid=25921) INFO 10-31 00:50:20 [default_loader.py:501] Loading weights took 0.75 seconds | |
| (EngineCore pid=25921) INFO 10-31 00:50:21 [model_runner.py:424] Model loading took 8.56 GiB memory and 4.702506 seconds | |
| (EngineCore pid=25921) INFO 10-31 00:50:21 [topk_topp_sampler.py:124] Using FlashInfer for top-p & top-k sampling. | |
| (EngineCore pid=25921) INFO 10-31 00:50:21 [torch_utils.py:295] Reducing Torch threads from 128 to 1 for serving. Set OMP_NUM_THREADS in the external environment to override. | |
| (EngineCore pid=25921) INFO 10-31 00:50:21 [interface.py:1096] Setting attention block size to 528 tokens to ensure that attention page size is >= mamba page size. | |
| (EngineCore pid=25921) INFO 10-31 00:50:21 [interface.py:1120] Padding mamba page size by 0.76% to ensure that mamba page size and attention page size are exactly equal. | |
| (EngineCore pid=25921) INFO 10-31 00:50:21 [utils.py:347] Using LBNHC KV cache layout. | |
| [transformers] Qwen3VL video processing does not apply the per-frame pixel cap the reference implementation (qwen-vl-utils) applies, so some videos cost far more tokens than they would there. In v5.22 the capped behavior will become the default and `cap_pixels_per_frame` will be removed. Pass `cap_pixels_per_frame=True` to adopt the reference behavior now, or `False` to keep the current behavior and silence this warning. | |
| (MainProcess pid=25496) INFO 10-31 00:50:23 [base.py:280] Multi-modal warmup completed in 20.065s | |
| (EngineCore pid=25921) INFO 10-31 00:50:29 [encoder_runner.py:131] Encoder cache will be initialized with a budget of 16384 tokens, and profiled with 1 image items of the maximum feature size. | |
| (EngineCore pid=25921) INFO 10-31 00:50:34 [monitor.py:40] [backbone] Starting torch.compile | |
| (EngineCore pid=25921) INFO 10-31 00:50:39 [backends.py:1104] [backbone] Using cache directory: /root/.cache/vllm/torch_compile_cache/c4af2daa17/rank_0_0/backbone for vLLM's torch.compile | |
| (EngineCore pid=25921) INFO 10-31 00:50:39 [backends.py:1166] [backbone] Dynamo bytecode transform time: 4.61 s | |
| (EngineCore pid=25921) INFO 10-31 00:50:56 [backends.py:398] [backbone] Compiling a graph for compile range (1, 2048) takes 17.28 s | |
| (EngineCore pid=25921) INFO 10-31 00:50:59 [backends.py:927] collected artifacts: 33 entries, 13 artifacts, 18863803 bytes total | |
| (EngineCore pid=25921) INFO 10-31 00:50:59 [decorators.py:743] [backbone] saved AOT compiled function to /root/.cache/vllm/torch_compile_cache/torch_aot_compile/cc0f0cae53dc3e1375a6d2ed4c88238808e3367e1d9abaa0419cc3d58ebc1b12/rank_0_0/model | |
| (EngineCore pid=25921) INFO 10-31 00:50:59 [monitor.py:62] [backbone] torch.compile took 24.79 s in total | |
| (EngineCore pid=25921) INFO 10-31 00:52:25 [monitor.py:91] [backbone] Initial profiling/warmup run took 85.73 s | |
| Capturing CUDA graphs (PIECEWISE): 0%| | 0/35 [00:00<?, ?it/s] Capturing CUDA graphs (PIECEWISE): 3%|▎ | 1/35 [00:00<00:30, 1.12it/s] Capturing CUDA graphs (PIECEWISE): 6%|▌ | 2/35 [00:01<00:26, 1.22it/s] Capturing CUDA graphs (PIECEWISE): 9%|▊ | 3/35 [00:01<00:18, 1.75it/s] Capturing CUDA graphs (PIECEWISE): 11%|█▏ | 4/35 [00:02<00:15, 2.03it/s] Capturing CUDA graphs (PIECEWISE): 14%|█▍ | 5/35 [00:02<00:13, 2.24it/s] Capturing CUDA graphs (PIECEWISE): 17%|█▋ | 6/35 [00:02<00:11, 2.63it/s] Capturing CUDA graphs (PIECEWISE): 20%|██ | 7/35 [00:03<00:10, 2.77it/s] Capturing CUDA graphs (PIECEWISE): 23%|██▎ | 8/35 [00:03<00:08, 3.12it/s] Capturing CUDA graphs (PIECEWISE): 26%|██▌ | 9/35 [00:03<00:08, 3.00it/s] Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 10/35 [00:04<00:08, 2.95it/s] Capturing CUDA graphs (PIECEWISE): 31%|███▏ | 11/35 [00:04<00:07, 3.25it/s] Capturing CUDA graphs (PIECEWISE): 34%|███▍ | 12/35 [00:04<00:07, 3.28it/s] Capturing CUDA graphs (PIECEWISE): 37%|███▋ | 13/35 [00:04<00:06, 3.67it/s] Capturing CUDA graphs (PIECEWISE): 40%|████ | 14/35 [00:07<00:21, 1.02s/it] Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 15/35 [00:07<00:15, 1.31it/s] Capturing CUDA graphs (PIECEWISE): 46%|████▌ | 16/35 [00:08<00:11, 1.71it/s] Capturing CUDA graphs (PIECEWISE): 49%|████▊ | 17/35 [00:09<00:12, 1.39it/s] Capturing CUDA graphs (PIECEWISE): 51%|█████▏ | 18/35 [00:09<00:09, 1.76it/s] Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 20/35 [00:09<00:05, 2.81it/s] Capturing CUDA graphs (PIECEWISE): 63%|██████▎ | 22/35 [00:09<00:03, 4.13it/s] Capturing CUDA graphs (PIECEWISE): 69%|██████▊ | 24/35 [00:09<00:01, 5.73it/s] Capturing CUDA graphs (PIECEWISE): 74%|███████▍ | 26/35 [00:09<00:01, 7.39it/s] Capturing CUDA graphs (PIECEWISE): 80%|████████ | 28/35 [00:10<00:00, 8.35it/s] Capturing CUDA graphs (PIECEWISE): 86%|████████▌ | 30/35 [00:10<00:00, 10.09it/s] Capturing CUDA graphs (PIECEWISE): 91%|█████████▏| 32/35 [00:10<00:00, 11.78it/s] Capturing CUDA graphs (PIECEWISE): 97%|█████████▋| 34/35 [00:10<00:00, 13.05it/s] Capturing CUDA graphs (PIECEWISE): 100%|██████████| 35/35 [00:10<00:00, 3.31it/s] | |
| Capturing CUDA graphs (FULL): 0%| | 0/2 [00:00<?, ?it/s] Capturing CUDA graphs (FULL): 100%|██████████| 2/2 [00:00<00:00, 23.26it/s] | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [model_runner.py:1127] Graph capturing finished in 12 secs, took 0.28 GiB | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [gpu_worker.py:759] Available KV cache memory: 58.36 GiB | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [gpu_worker.py:774] CUDA graph memory profiling is enabled (default since v0.21.0). The current --gpu-memory-utilization=0.9000 is equivalent to --gpu-memory-utilization=0.8960 without CUDA graph memory profiling. To maintain the same effective KV cache size as before, increase --gpu-memory-utilization to 0.9040. To disable, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0. | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [kv_cache_utils.py:2327] GPU KV cache size: 1,719,607 tokens, Maximum concurrency for 32,768 tokens per request: 52.48x | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [kv_cache_utils.py:853] kv cache group sizes (528, 528, 528, 528) | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [kv_cache_utils.py:854] kv lcm block sizes 528 | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [kernel_warmup.py:176] JIT kernel warmup starting. | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [kernel_warmup.py:189] JIT kernel warmup finished in 0.00s. | |
| (EngineCore pid=25921) INFO 10-31 00:52:39 [qwen_triton_warmup.py:296] Warming up Qwen GDN Triton kernels for model_type=qwen3_5_text. | |
| (EngineCore pid=25921) INFO 10-31 00:52:41 [qwen_vl_triton_warmup.py:57] Warmed position embedding and vision rotary kernels on grids=[(1, 16, 16), (1, 16, 2), (1, 2, 16), (1, 2, 2)]. | |
| (EngineCore pid=25921) INFO 10-31 00:52:41 [mamba_triton_warmup.py:42] Warmed Mamba batch_memcpy_kernel. | |
| Capturing CUDA graphs (PIECEWISE): 0%| | 0/35 [00:00<?, ?it/s] Capturing CUDA graphs (PIECEWISE): 3%|▎ | 1/35 [00:00<00:11, 2.99it/s] Capturing CUDA graphs (PIECEWISE): 6%|▌ | 2/35 [00:00<00:12, 2.59it/s] Capturing CUDA graphs (PIECEWISE): 9%|▊ | 3/35 [00:01<00:12, 2.59it/s] Capturing CUDA graphs (PIECEWISE): 11%|█▏ | 4/35 [00:01<00:11, 2.65it/s] Capturing CUDA graphs (PIECEWISE): 14%|█▍ | 5/35 [00:01<00:10, 2.76it/s] Capturing CUDA graphs (PIECEWISE): 17%|█▋ | 6/35 [00:02<00:10, 2.83it/s] Capturing CUDA graphs (PIECEWISE): 20%|██ | 7/35 [00:02<00:09, 3.10it/s] Capturing CUDA graphs (PIECEWISE): 23%|██▎ | 8/35 [00:02<00:08, 3.26it/s] Capturing CUDA graphs (PIECEWISE): 26%|██▌ | 9/35 [00:03<00:08, 2.95it/s] Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 10/35 [00:03<00:07, 3.18it/s] Capturing CUDA graphs (PIECEWISE): 31%|███▏ | 11/35 [00:03<00:07, 3.20it/s] Capturing CUDA graphs (PIECEWISE): 34%|███▍ | 12/35 [00:03<00:06, 3.55it/s] Capturing CUDA graphs (PIECEWISE): 37%|███▋ | 13/35 [00:04<00:06, 3.51it/s] Capturing CUDA graphs (PIECEWISE): 40%|████ | 14/35 [00:04<00:05, 3.98it/s] Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 15/35 [00:04<00:05, 4.00it/s] Capturing CUDA graphs (PIECEWISE): 46%|████▌ | 16/35 [00:04<00:04, 4.43it/s] Capturing CUDA graphs (PIECEWISE): 51%|█████▏ | 18/35 [00:04<00:02, 6.62it/s] Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 20/35 [00:05<00:02, 7.38it/s] Capturing CUDA graphs (PIECEWISE): 63%|██████▎ | 22/35 [00:05<00:01, 9.07it/s] Capturing CUDA graphs (PIECEWISE): 69%|██████▊ | 24/35 [00:05<00:01, 10.69it/s] Capturing CUDA graphs (PIECEWISE): 74%|███████▍ | 26/35 [00:05<00:00, 11.95it/s] Capturing CUDA graphs (PIECEWISE): 80%|████████ | 28/35 [00:05<00:00, 11.12it/s] Capturing CUDA graphs (PIECEWISE): 86%|████████▌ | 30/35 [00:05<00:00, 12.50it/s] Capturing CUDA graphs (PIECEWISE): 91%|█████████▏| 32/35 [00:05<00:00, 13.45it/s] Capturing CUDA graphs (PIECEWISE): 97%|█████████▋| 34/35 [00:06<00:00, 11.06it/s] Capturing CUDA graphs (PIECEWISE): 100%|██████████| 35/35 [00:06<00:00, 5.58it/s] | |
| Capturing CUDA graphs (FULL): 0%| | 0/19 [00:00<?, ?it/s] Capturing CUDA graphs (FULL): 16%|█▌ | 3/19 [00:00<00:00, 25.15it/s] Capturing CUDA graphs (FULL): 32%|███▏ | 6/19 [00:00<00:00, 25.44it/s] Capturing CUDA graphs (FULL): 47%|████▋ | 9/19 [00:00<00:00, 22.67it/s] Capturing CUDA graphs (FULL): 63%|██████▎ | 12/19 [00:00<00:00, 22.18it/s] Capturing CUDA graphs (FULL): 79%|███████▉ | 15/19 [00:00<00:00, 23.26it/s] Capturing CUDA graphs (FULL): 95%|█████████▍| 18/19 [00:00<00:00, 23.42it/s] Capturing CUDA graphs (FULL): 100%|██████████| 19/19 [00:00<00:00, 23.23it/s] | |
| (EngineCore pid=25921) INFO 10-31 00:53:08 [model_runner.py:1127] Graph capturing finished in 8 secs, took 0.14 GiB | |
| (EngineCore pid=25921) INFO 10-31 00:53:08 [gpu_worker.py:971] CUDA graph pool memory: 0.14 GiB (actual), 0.32 GiB (estimated), difference: 0.18 GiB (132.9%). | |
| (EngineCore pid=25921) INFO 10-31 00:53:08 [gpu_worker.py:1034] Free memory on device (78.84/79.25 GiB) on startup. Desired GPU memory utilization is (0.9, 71.33 GiB). Actual usage is 11.05 GiB for consumed memory (weights + non-torch), 1.92 GiB for peak activation, and 0.14 GiB for CUDAGraph memory. Replace gpu_memory_utilization config with `--kv-cache-memory=62358280090` (58.08 GiB) to fit into requested memory, or `--kv-cache-memory=70424346112` (65.59 GiB) to fully utilize gpu memory. Current kv cache memory in use is 58.36 GiB. | |
| (EngineCore pid=25921) INFO 10-31 00:53:09 [jit_monitor.py:84] Kernel JIT monitor activated; monitored JIT compilations during inference will use mode=warn. | |
| (EngineCore pid=25921) INFO 10-31 00:53:09 [core.py:395] init engine (profile, create kv cache, warmup model) took 167.63 s (compilation: 24.79 s) | |
| Parse safetensors files: 0%| | 0/2 [00:00<?, ?it/s] Parse safetensors files: 50%|█████ | 1/2 [00:00<00:00, 3.68it/s] Parse safetensors files: 100%|██████████| 2/2 [00:00<00:00, 7.14it/s] | |
| (EngineCore pid=25921) INFO 10-31 00:53:09 [kernel.py:467] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'], gelu_and_mul_sparse=['triton', 'native']) | |
| (MainProcess pid=25496) WARNING 10-31 00:53:10 [input_processor.py:408] Passing raw prompts to InputProcessor is deprecated and will be removed in the future. You should instead pass the outputs of Renderer.render_cmpl() or Renderer.render_chat(). | |
| (EngineCore pid=25921) WARNING 10-31 00:53:10 [jit_monitor.py:140] Triton kernel JIT compilation during inference: _topp_sb_stats_kernel. This causes a latency spike; consider extending warmup to cover this shape/config. | |
| (EngineCore pid=25921) WARNING 10-31 00:53:11 [jit_monitor.py:140] Triton kernel JIT compilation during inference: _topp_sb_step_kernel. This causes a latency spike; consider extending warmup to cover this shape/config. | |
| (EngineCore pid=25921) WARNING 10-31 00:53:12 [jit_monitor.py:140] Triton kernel JIT compilation during inference: _topp_sb_mask_kernel. This causes a latency spike; consider extending warmup to cover this shape/config. | |
| progress {"time": "00:54:09", "elapsed_min": 1.0, "submitted": 128, "finished": 18, "in_flight": 110, "gen_tokens": 57273, "tok_per_s": 955, "chain_tokens_median": 734.0, "chain_tokens_p90": 1303.8, "chain_tokens_max": 2064, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "00:55:09", "elapsed_min": 2.0, "submitted": 128, "finished": 31, "in_flight": 97, "gen_tokens": 160684, "tok_per_s": 1339, "chain_tokens_median": 866.5, "chain_tokens_p90": 2762.9, "chain_tokens_max": 4367, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "00:56:09", "elapsed_min": 3.0, "submitted": 128, "finished": 61, "in_flight": 67, "gen_tokens": 408301, "tok_per_s": 2268, "chain_tokens_median": 969.0, "chain_tokens_p90": 3920.2000000000003, "chain_tokens_max": 6404, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "00:57:09", "elapsed_min": 4.0, "submitted": 128, "finished": 90, "in_flight": 38, "gen_tokens": 744826, "tok_per_s": 3103, "chain_tokens_median": 1202.5, "chain_tokens_p90": 4983.400000000002, "chain_tokens_max": 8258, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "00:58:09", "elapsed_min": 5.0, "submitted": 128, "finished": 111, "in_flight": 17, "gen_tokens": 1030362, "tok_per_s": 3434, "chain_tokens_median": 1366.5, "chain_tokens_p90": 5461.599999999999, "chain_tokens_max": 9949, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "00:59:09", "elapsed_min": 6.0, "submitted": 128, "finished": 124, "in_flight": 4, "gen_tokens": 1338435, "tok_per_s": 3718, "chain_tokens_median": 1763.0, "chain_tokens_p90": 6118.0, "chain_tokens_max": 11487, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "01:00:09", "elapsed_min": 7.0, "submitted": 128, "finished": 125, "in_flight": 3, "gen_tokens": 1367754, "tok_per_s": 3256, "chain_tokens_median": 1798.5, "chain_tokens_p90": 6196.1, "chain_tokens_max": 11487, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "01:01:09", "elapsed_min": 8.0, "submitted": 128, "finished": 126, "in_flight": 2, "gen_tokens": 1417464, "tok_per_s": 2953, "chain_tokens_median": 1853.5, "chain_tokens_p90": 6267.4, "chain_tokens_max": 19323, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| progress {"time": "01:02:09", "elapsed_min": 9.0, "submitted": 128, "finished": 126, "in_flight": 2, "gen_tokens": 1417464, "tok_per_s": 2625, "chain_tokens_median": 1853.5, "chain_tokens_p90": 6267.4, "chain_tokens_max": 19323, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=25496) INFO 10-31 01:03:02 [core_client.py:770] [shutdown] MPClient: start timeout=default | |
| (MainProcess pid=25496) INFO 10-31 01:03:02 [core_client.py:772] [shutdown] MPClient: stopping engine manager | |
| (MainProcess pid=25496) INFO 10-31 01:03:02 [utils.py:628] [shutdown] Process manager: send sigterm to process EngineCore | |
| (EngineCore pid=25921) INFO 10-31 01:03:02 [core.py:1454] [shutdown] EngineCore: trigger received signal=SIGTERM | |
| (EngineCore pid=25921) INFO 10-31 01:03:02 [core.py:1635] [shutdown] EngineCore: start mode=abort timeout=0s | |
| (EngineCore pid=25921) INFO 10-31 01:03:02 [core.py:1666] [shutdown] EngineCore: request processing complete; starting resource teardown | |
| (EngineCore pid=25921) INFO 10-31 01:03:02 [core.py:1499] [shutdown] EngineCore: exiting busy loop | |
| (MainProcess pid=25496) INFO 10-31 01:03:06 [core_client.py:774] [shutdown] MPClient: engine manager stopped | |
| (MainProcess pid=25496) INFO 10-31 01:03:06 [core_client.py:775] [shutdown] MPClient: cleaning up background resources | |
| (MainProcess pid=25496) INFO 10-31 01:03:06 [core_client.py:777] [shutdown] MPClient: complete | |
| { | |
| "items": 128, | |
| "samples": 512, | |
| "tokens_total": 1562525, | |
| "tokens_median": 1878.0, | |
| "tokens_p90": 6413.900000000001, | |
| "tokens_max": 32471, | |
| "parse_rate": 0.9921875, | |
| "looped": 4, | |
| "items_with_correct_chain": 111, | |
| "majority_accuracy": 0.8125, | |
| "seconds": 798.2385666370392, | |
| "gen_tok_per_s": 1957.4661828015678, | |
| "engine": { | |
| "submitted": 128, | |
| "finished": 128, | |
| "gen_tokens": 1562525, | |
| "t0": 1793407989.7934816, | |
| "seconds": 592.3575947284698, | |
| "tok_per_s": 2637.806983324396 | |
| }, | |
| "N_fit": 1587 | |
| } | |
Xet Storage Details
- Size:
- 29.4 kB
- Xet hash:
- 7eb098c0c3d552ea54d32ee221d0843e2966f2cfae077aab7c6f3f5e0f7a8823
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.