Buckets:
| (MainProcess pid=44936) INFO 10-31 06:56:27 [model.py:733] Resolved architecture: Qwen3_5ForConditionalGeneration | |
| (MainProcess pid=44936) INFO 10-31 06:56:27 [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, 1167.98it/s] | |
| (MainProcess pid=44936) INFO 10-31 06:56:29 [scheduler.py:307] Chunked prefill is enabled with max_num_batched_tokens=2048. | |
| (MainProcess pid=44936) INFO 10-31 06:56:29 [config.py:688] Mamba cache mode is set to 'align' for Qwen3_5ForConditionalGeneration by default when prefix caching is enabled | |
| Parse safetensors files: 0%| | 0/1 [00:00, ?it/s] Parse safetensors files: 100%|██████████| 1/1 [00:00<00:00, 6.11it/s] Parse safetensors files: 100%|██████████| 1/1 [00:00<00:00, 6.09it/s] | |
| (MainProcess pid=44936) INFO 10-31 06:56:30 [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=44936) WARNING 10-31 06:56:41 [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=45494) INFO 10-31 06:56:53 [core.py:133] Initializing a V1 LLM engine (v0.32.0) with config: model='Qwen/Qwen3.5-0.8B', speculative_config=None, tokenizer='Qwen/Qwen3.5-0.8B', 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/Qwen3.5-0.8B, 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, 272, 288, 304, 320, 336, 352, 368, 384, 400, 416, 432, 448, 464, 480, 496, 512], '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': 512, '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=45494) INFO 10-31 06:56:55 [parallel_state.py:1906] world_size=1 rank=0 local_rank=0 distributed_init_method=file:///tmp/vllm_dist_03e14da7159f41738f36c39d53d7c123 backend=nccl | |
| (EngineCore pid=45494) INFO 10-31 06:56:55 [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=45494) INFO 10-31 06:56:55 [gpu_worker.py:518] Using V2 Model Runner | |
| (EngineCore pid=45494) INFO 10-31 06:56:56 [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, 924.96it/s] | |
| (EngineCore pid=45494) INFO 10-31 06:56:57 [cuda.py:565] Using backend AttentionBackendEnum.FLASH_ATTN for vit attention | |
| (EngineCore pid=45494) INFO 10-31 06:56:57 [mm_encoder_attention.py:372] Using AttentionBackendEnum.FLASH_ATTN for MMEncoderAttention. | |
| (EngineCore pid=45494) INFO 10-31 06:56:58 [qwen_gdn_linear_attn.py:221] Using Triton/FLA GDN prefill kernel (requested=auto, head_k_dim=128). | |
| (EngineCore pid=45494) INFO 10-31 06:56:58 [cuda.py:506] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION']. | |
| (EngineCore pid=45494) INFO 10-31 06:56:58 [flash_attn.py:1117] Using FlashAttention version 2 | |
| (EngineCore pid=45494) INFO 10-31 06:56:59 [base_loader.py:82] Initializing model took 2.163349 seconds | |
| (EngineCore pid=45494) INFO 10-31 06:56:59 [weight_utils.py:1108] Filesystem type for checkpoints: XFS. Checkpoint size: 1.63 GiB. Available RAM: 182.05 GiB. | |
| (EngineCore pid=45494) INFO 10-31 06:56:59 [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 | 0/1 [00:00, ?it/s] | |
| (EngineCore pid=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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=45494) WARNING 10-31 06:56:59 [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 | 1/1 [00:00<00:00, 1.33it/s] | |
| Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00, 1.33it/s] | |
| (EngineCore pid=45494) INFO 10-31 06:56:59 [default_loader.py:501] Loading weights took 0.76 seconds | |
| [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. | |
| (EngineCore pid=45494) INFO 10-31 06:57:01 [model_runner.py:424] Model loading took 1.63 GiB memory and 4.655278 seconds | |
| (EngineCore pid=45494) INFO 10-31 06:57:01 [topk_topp_sampler.py:124] Using FlashInfer for top-p & top-k sampling. | |
| (EngineCore pid=45494) INFO 10-31 06:57:01 [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=45494) INFO 10-31 06:57:01 [interface.py:1096] Setting attention block size to 544 tokens to ensure that attention page size is >= mamba page size. | |
| (EngineCore pid=45494) INFO 10-31 06:57:01 [interface.py:1120] Padding mamba page size by 2.64% to ensure that mamba page size and attention page size are exactly equal. | |
| (EngineCore pid=45494) INFO 10-31 06:57:01 [utils.py:347] Using LBNHC KV cache layout. | |
| (MainProcess pid=44936) INFO 10-31 06:57:01 [base.py:280] Multi-modal warmup completed in 20.179s | |
| (EngineCore pid=45494) INFO 10-31 06:57:09 [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=45494) INFO 10-31 06:57:11 [monitor.py:40] [backbone] Starting torch.compile | |
| (EngineCore pid=45494) INFO 10-31 06:57:15 [backends.py:1104] [backbone] Using cache directory: /root/.cache/vllm/torch_compile_cache/a09563f6c2/rank_0_0/backbone for vLLM's torch.compile | |
| (EngineCore pid=45494) INFO 10-31 06:57:15 [backends.py:1166] [backbone] Dynamo bytecode transform time: 3.25 s | |
| (EngineCore pid=45494) INFO 10-31 06:57:30 [backends.py:398] [backbone] Compiling a graph for compile range (1, 2048) takes 15.32 s | |
| (EngineCore pid=45494) INFO 10-31 06:57:33 [backends.py:927] collected artifacts: 25 entries, 11 artifacts, 17073159 bytes total | |
| (EngineCore pid=45494) INFO 10-31 06:57:33 [decorators.py:743] [backbone] saved AOT compiled function to /root/.cache/vllm/torch_compile_cache/torch_aot_compile/e6a7cf21b233ded846a30264795443f0c57059c148a944f486b258b1ac4ad238/rank_0_0/model | |
| (EngineCore pid=45494) INFO 10-31 06:57:33 [monitor.py:62] [backbone] torch.compile took 21.39 s in total | |
| (EngineCore pid=45494) INFO 10-31 06:59:11 [monitor.py:91] [backbone] Initial profiling/warmup run took 98.53 s | |
| Capturing CUDA graphs (PIECEWISE): 0%| | 0/51 [00:00<?, ?it/s] Capturing CUDA graphs (PIECEWISE): 2%|▏ | 1/51 [00:03<02:56, 3.54s/it] Capturing CUDA graphs (PIECEWISE): 4%|▍ | 2/51 [00:04<01:30, 1.86s/it] Capturing CUDA graphs (PIECEWISE): 6%|▌ | 3/51 [00:04<00:54, 1.13s/it] Capturing CUDA graphs (PIECEWISE): 8%|▊ | 4/51 [00:04<00:37, 1.27it/s] Capturing CUDA graphs (PIECEWISE): 10%|▉ | 5/51 [00:05<00:27, 1.67it/s] Capturing CUDA graphs (PIECEWISE): 12%|█▏ | 6/51 [00:05<00:21, 2.09it/s] Capturing CUDA graphs (PIECEWISE): 14%|█▎ | 7/51 [00:05<00:17, 2.56it/s] Capturing CUDA graphs (PIECEWISE): 16%|█▌ | 8/51 [00:05<00:14, 2.92it/s] Capturing CUDA graphs (PIECEWISE): 18%|█▊ | 9/51 [00:06<00:14, 2.95it/s] Capturing CUDA graphs (PIECEWISE): 20%|█▉ | 10/51 [00:06<00:12, 3.21it/s] Capturing CUDA graphs (PIECEWISE): 22%|██▏ | 11/51 [00:06<00:11, 3.47it/s] Capturing CUDA graphs (PIECEWISE): 24%|██▎ | 12/51 [00:06<00:10, 3.84it/s] Capturing CUDA graphs (PIECEWISE): 25%|██▌ | 13/51 [00:06<00:09, 4.14it/s] Capturing CUDA graphs (PIECEWISE): 27%|██▋ | 14/51 [00:07<00:08, 4.61it/s] Capturing CUDA graphs (PIECEWISE): 29%|██▉ | 15/51 [00:07<00:07, 5.01it/s] Capturing CUDA graphs (PIECEWISE): 31%|███▏ | 16/51 [00:07<00:06, 5.45it/s] Capturing CUDA graphs (PIECEWISE): 33%|███▎ | 17/51 [00:08<00:14, 2.27it/s] Capturing CUDA graphs (PIECEWISE): 35%|███▌ | 18/51 [00:08<00:12, 2.71it/s] Capturing CUDA graphs (PIECEWISE): 41%|████ | 21/51 [00:08<00:05, 5.40it/s] Capturing CUDA graphs (PIECEWISE): 47%|████▋ | 24/51 [00:08<00:03, 8.33it/s] Capturing CUDA graphs (PIECEWISE): 53%|█████▎ | 27/51 [00:08<00:02, 11.21it/s] Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 29/51 [00:09<00:01, 11.72it/s] Capturing CUDA graphs (PIECEWISE): 63%|██████▎ | 32/51 [00:09<00:01, 13.72it/s] Capturing CUDA graphs (PIECEWISE): 69%|██████▊ | 35/51 [00:09<00:01, 15.57it/s] Capturing CUDA graphs (PIECEWISE): 73%|███████▎ | 37/51 [00:09<00:00, 15.75it/s] Capturing CUDA graphs (PIECEWISE): 78%|███████▊ | 40/51 [00:09<00:00, 16.45it/s] Capturing CUDA graphs (PIECEWISE): 82%|████████▏ | 42/51 [00:09<00:00, 15.78it/s] Capturing CUDA graphs (PIECEWISE): 86%|████████▋ | 44/51 [00:10<00:00, 15.11it/s] Capturing CUDA graphs (PIECEWISE): 90%|█████████ | 46/51 [00:10<00:00, 14.63it/s] Capturing CUDA graphs (PIECEWISE): 94%|█████████▍| 48/51 [00:10<00:00, 14.28it/s] Capturing CUDA graphs (PIECEWISE): 98%|█████████▊| 50/51 [00:10<00:00, 13.98it/s] Capturing CUDA graphs (PIECEWISE): 100%|██████████| 51/51 [00:10<00:00, 4.76it/s] | |
| Capturing CUDA graphs (FULL): 0%| | 0/2 [00:00<?, ?it/s] Capturing CUDA graphs (FULL): 100%|██████████| 2/2 [00:00<00:00, 29.82it/s] | |
| (EngineCore pid=45494) INFO 10-31 06:59:26 [model_runner.py:1127] Graph capturing finished in 12 secs, took 0.30 GiB | |
| (EngineCore pid=45494) INFO 10-31 06:59:26 [gpu_worker.py:759] Available KV cache memory: 10.76 GiB | |
| (EngineCore pid=45494) INFO 10-31 06:59:26 [gpu_worker.py:774] CUDA graph memory profiling is enabled (default since v0.21.0). The current --gpu-memory-utilization=0.2200 is equivalent to --gpu-memory-utilization=0.2158 without CUDA graph memory profiling. To maintain the same effective KV cache size as before, increase --gpu-memory-utilization to 0.2242. To disable, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0. | |
| (EngineCore pid=45494) INFO 10-31 06:59:26 [kv_cache_utils.py:2327] GPU KV cache size: 845,120 tokens, Maximum concurrency for 32,768 tokens per request: 25.79x | |
| (EngineCore pid=45494) INFO 10-31 06:59:26 [kv_cache_utils.py:853] kv cache group sizes (544, 544, 544, 544) | |
| (EngineCore pid=45494) INFO 10-31 06:59:26 [kv_cache_utils.py:854] kv lcm block sizes 544 | |
| (EngineCore pid=45494) INFO 10-31 06:59:27 [kernel_warmup.py:176] JIT kernel warmup starting. | |
| (EngineCore pid=45494) INFO 10-31 06:59:27 [kernel_warmup.py:189] JIT kernel warmup finished in 0.00s. | |
| (EngineCore pid=45494) INFO 10-31 06:59:27 [qwen_triton_warmup.py:296] Warming up Qwen GDN Triton kernels for model_type=qwen3_5_text. | |
| (EngineCore pid=45494) INFO 10-31 06:59:28 [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=45494) INFO 10-31 06:59:29 [mamba_triton_warmup.py:42] Warmed Mamba batch_memcpy_kernel. | |
| Capturing CUDA graphs (PIECEWISE): 0%| | 0/51 [00:00<?, ?it/s] Capturing CUDA graphs (PIECEWISE): 2%|▏ | 1/51 [00:00<00:14, 3.42it/s] Capturing CUDA graphs (PIECEWISE): 4%|▍ | 2/51 [00:00<00:13, 3.53it/s] Capturing CUDA graphs (PIECEWISE): 6%|▌ | 3/51 [00:00<00:13, 3.59it/s] Capturing CUDA graphs (PIECEWISE): 8%|▊ | 4/51 [00:01<00:12, 3.89it/s] Capturing CUDA graphs (PIECEWISE): 10%|▉ | 5/51 [00:01<00:11, 3.86it/s] Capturing CUDA graphs (PIECEWISE): 12%|█▏ | 6/51 [00:01<00:11, 3.92it/s] Capturing CUDA graphs (PIECEWISE): 14%|█▎ | 7/51 [00:01<00:11, 3.94it/s] Capturing CUDA graphs (PIECEWISE): 16%|█▌ | 8/51 [00:02<00:10, 4.00it/s] Capturing CUDA graphs (PIECEWISE): 18%|█▊ | 9/51 [00:02<00:11, 3.79it/s] Capturing CUDA graphs (PIECEWISE): 20%|█▉ | 10/51 [00:02<00:10, 3.78it/s] Capturing CUDA graphs (PIECEWISE): 22%|██▏ | 11/51 [00:02<00:10, 3.96it/s] Capturing CUDA graphs (PIECEWISE): 24%|██▎ | 12/51 [00:03<00:09, 4.22it/s] Capturing CUDA graphs (PIECEWISE): 25%|██▌ | 13/51 [00:03<00:08, 4.46it/s] Capturing CUDA graphs (PIECEWISE): 27%|██▋ | 14/51 [00:03<00:07, 4.87it/s] Capturing CUDA graphs (PIECEWISE): 29%|██▉ | 15/51 [00:03<00:06, 5.20it/s] Capturing CUDA graphs (PIECEWISE): 31%|███▏ | 16/51 [00:03<00:06, 5.63it/s] Capturing CUDA graphs (PIECEWISE): 35%|███▌ | 18/51 [00:03<00:04, 7.65it/s] Capturing CUDA graphs (PIECEWISE): 39%|███▉ | 20/51 [00:04<00:03, 9.19it/s] Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 22/51 [00:04<00:02, 10.31it/s] Capturing CUDA graphs (PIECEWISE): 47%|████▋ | 24/51 [00:04<00:02, 11.14it/s] Capturing CUDA graphs (PIECEWISE): 51%|█████ | 26/51 [00:04<00:02, 11.76it/s] Capturing CUDA graphs (PIECEWISE): 55%|█████▍ | 28/51 [00:04<00:01, 12.21it/s] Capturing CUDA graphs (PIECEWISE): 59%|█████▉ | 30/51 [00:04<00:01, 12.46it/s] Capturing CUDA graphs (PIECEWISE): 63%|██████▎ | 32/51 [00:04<00:01, 12.88it/s] Capturing CUDA graphs (PIECEWISE): 67%|██████▋ | 34/51 [00:05<00:01, 13.02it/s] Capturing CUDA graphs (PIECEWISE): 71%|███████ | 36/51 [00:05<00:01, 13.12it/s] Capturing CUDA graphs (PIECEWISE): 75%|███████▍ | 38/51 [00:05<00:00, 13.19it/s] Capturing CUDA graphs (PIECEWISE): 78%|███████▊ | 40/51 [00:05<00:00, 13.13it/s] Capturing CUDA graphs (PIECEWISE): 82%|████████▏ | 42/51 [00:05<00:00, 13.22it/s] Capturing CUDA graphs (PIECEWISE): 86%|████████▋ | 44/51 [00:05<00:00, 13.29it/s] Capturing CUDA graphs (PIECEWISE): 90%|█████████ | 46/51 [00:05<00:00, 13.36it/s] Capturing CUDA graphs (PIECEWISE): 94%|█████████▍| 48/51 [00:06<00:00, 13.34it/s] Capturing CUDA graphs (PIECEWISE): 98%|█████████▊| 50/51 [00:06<00:00, 13.70it/s] Capturing CUDA graphs (PIECEWISE): 100%|██████████| 51/51 [00:06<00:00, 8.11it/s] | |
| Capturing CUDA graphs (FULL): 0%| | 0/35 [00:00<?, ?it/s] Capturing CUDA graphs (FULL): 11%|█▏ | 4/35 [00:00<00:00, 34.69it/s] Capturing CUDA graphs (FULL): 23%|██▎ | 8/35 [00:00<00:00, 34.71it/s] Capturing CUDA graphs (FULL): 34%|███▍ | 12/35 [00:00<00:00, 34.69it/s] Capturing CUDA graphs (FULL): 46%|████▌ | 16/35 [00:00<00:00, 34.93it/s] Capturing CUDA graphs (FULL): 57%|█████▋ | 20/35 [00:00<00:00, 35.30it/s] Capturing CUDA graphs (FULL): 69%|██████▊ | 24/35 [00:00<00:00, 34.91it/s] Capturing CUDA graphs (FULL): 80%|████████ | 28/35 [00:00<00:00, 35.14it/s] Capturing CUDA graphs (FULL): 91%|█████████▏| 32/35 [00:00<00:00, 35.35it/s] Capturing CUDA graphs (FULL): 100%|██████████| 35/35 [00:00<00:00, 35.24it/s] | |
| (EngineCore pid=45494) INFO 10-31 06:59:55 [model_runner.py:1127] Graph capturing finished in 8 secs, took 0.20 GiB | |
| (EngineCore pid=45494) INFO 10-31 06:59:55 [gpu_worker.py:971] CUDA graph pool memory: 0.2 GiB (actual), 0.34 GiB (estimated), difference: 0.13 GiB (67.0%). | |
| (EngineCore pid=45494) INFO 10-31 06:59:55 [gpu_worker.py:1034] Free memory on device (57.96/79.25 GiB) on startup. Desired GPU memory utilization is (0.22, 17.44 GiB). Actual usage is 5.11 GiB for consumed memory (weights + non-torch), 1.56 GiB for peak activation, and 0.2 GiB for CUDAGraph memory. Replace gpu_memory_utilization config with `--kv-cache-memory=11181304853` (10.41 GiB) to fit into requested memory, or `--kv-cache-memory=54691659264` (50.94 GiB) to fully utilize gpu memory. Current kv cache memory in use is 10.76 GiB. | |
| (EngineCore pid=45494) INFO 10-31 06:59:56 [jit_monitor.py:84] Kernel JIT monitor activated; monitored JIT compilations during inference will use mode=warn. | |
| (EngineCore pid=45494) INFO 10-31 06:59:57 [core.py:395] init engine (profile, create kv cache, warmup model) took 175.50 s (compilation: 21.39 s) | |
| Parse safetensors files: 0%| | 0/1 [00:00<?, ?it/s] Parse safetensors files: 100%|██████████| 1/1 [00:00<00:00, 7.08it/s] Parse safetensors files: 100%|██████████| 1/1 [00:00<00:00, 7.05it/s] | |
| (EngineCore pid=45494) INFO 10-31 06:59:57 [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=44936) WARNING 10-31 06:59:57 [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=45494) WARNING 10-31 06:59:59 [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=45494) WARNING 10-31 07:00:00 [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=45494) WARNING 10-31 07:00:01 [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. | |
| (EngineCore pid=45494) WARNING 10-31 07:00:10 [jit_monitor.py:140] Triton kernel JIT compilation during inference: _topk_log_softmax_kernel. This causes a latency spike; consider extending warmup to cover this shape/config. | |
| progress {"time": "07:00:57", "elapsed_min": 1.0, "submitted": 24, "finished": 8, "in_flight": 16, "gen_tokens": 9524, "tok_per_s": 159, "chain_tokens_median": 1116.0, "chain_tokens_p90": 1875.9999999999998, "chain_tokens_max": 2485, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=44936) INFO 10-31 07:00:57 [loggers.py:335] Engine 000: Avg prompt throughput: 295.2 tokens/s, Avg generation throughput: 1453.9 tokens/s, Running: 16 reqs, Waiting: 0 reqs, GPU KV cache usage: 12.2%, Prefix cache hit rate: 0.0% | |
| progress {"time": "07:01:57", "elapsed_min": 2.0, "submitted": 24, "finished": 8, "in_flight": 16, "gen_tokens": 9524, "tok_per_s": 79, "chain_tokens_median": 1116.0, "chain_tokens_p90": 1875.9999999999998, "chain_tokens_max": 2485, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=44936) INFO 10-31 07:01:57 [loggers.py:335] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 1445.8 tokens/s, Running: 16 reqs, Waiting: 0 reqs, GPU KV cache usage: 21.5%, Prefix cache hit rate: 0.0% | |
| progress {"time": "07:02:57", "elapsed_min": 3.0, "submitted": 24, "finished": 8, "in_flight": 16, "gen_tokens": 9524, "tok_per_s": 53, "chain_tokens_median": 1116.0, "chain_tokens_p90": 1875.9999999999998, "chain_tokens_max": 2485, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=44936) INFO 10-31 07:02:57 [loggers.py:335] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 1207.2 tokens/s, Running: 16 reqs, Waiting: 0 reqs, GPU KV cache usage: 29.1%, Prefix cache hit rate: 0.0% | |
| progress {"time": "07:03:57", "elapsed_min": 4.0, "submitted": 24, "finished": 8, "in_flight": 16, "gen_tokens": 9524, "tok_per_s": 40, "chain_tokens_median": 1116.0, "chain_tokens_p90": 1875.9999999999998, "chain_tokens_max": 2485, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=44936) INFO 10-31 07:03:57 [loggers.py:335] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 1101.8 tokens/s, Running: 16 reqs, Waiting: 0 reqs, GPU KV cache usage: 36.2%, Prefix cache hit rate: 0.0% | |
| progress {"time": "07:04:57", "elapsed_min": 5.0, "submitted": 24, "finished": 9, "in_flight": 15, "gen_tokens": 29839, "tok_per_s": 99, "chain_tokens_median": 1480.0, "chain_tokens_p90": 6051.000000000004, "chain_tokens_max": 20315, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=44936) INFO 10-31 07:04:57 [loggers.py:335] Engine 000: Avg prompt throughput: 340.3 tokens/s, Avg generation throughput: 1111.7 tokens/s, Running: 15 reqs, Waiting: 0 reqs, GPU KV cache usage: 41.1%, Prefix cache hit rate: 1.4% | |
| progress {"time": "07:05:57", "elapsed_min": 6.0, "submitted": 24, "finished": 9, "in_flight": 15, "gen_tokens": 29839, "tok_per_s": 83, "chain_tokens_median": 1480.0, "chain_tokens_p90": 6051.000000000004, "chain_tokens_max": 20315, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=44936) INFO 10-31 07:05:57 [loggers.py:335] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 1027.5 tokens/s, Running: 15 reqs, Waiting: 0 reqs, GPU KV cache usage: 47.2%, Prefix cache hit rate: 1.4% | |
| progress {"time": "07:06:57", "elapsed_min": 7.0, "submitted": 24, "finished": 9, "in_flight": 15, "gen_tokens": 29839, "tok_per_s": 71, "chain_tokens_median": 1480.0, "chain_tokens_p90": 6051.000000000004, "chain_tokens_max": 20315, "parse_rate": 1.0, "looped": 0, "minutes_to_deadline": Infinity} | |
| (MainProcess pid=44936) INFO 10-31 07:06:57 [loggers.py:335] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 861.7 tokens/s, Running: 15 reqs, Waiting: 0 reqs, GPU KV cache usage: 53.0%, Prefix cache hit rate: 1.4% | |
| (MainProcess pid=44936) INFO 10-31 07:07:32 [core_client.py:770] [shutdown] MPClient: start timeout=default | |
| (MainProcess pid=44936) INFO 10-31 07:07:32 [core_client.py:772] [shutdown] MPClient: stopping engine manager | |
| (MainProcess pid=44936) INFO 10-31 07:07:32 [utils.py:628] [shutdown] Process manager: send sigterm to process EngineCore | |
| (EngineCore pid=45494) INFO 10-31 07:07:32 [core.py:1454] [shutdown] EngineCore: trigger received signal=SIGTERM | |
| (EngineCore pid=45494) INFO 10-31 07:07:32 [core.py:1635] [shutdown] EngineCore: start mode=abort timeout=0s | |
| (EngineCore pid=45494) INFO 10-31 07:07:32 [core.py:1666] [shutdown] EngineCore: request processing complete; starting resource teardown | |
| (EngineCore pid=45494) INFO 10-31 07:07:32 [core.py:1499] [shutdown] EngineCore: exiting busy loop | |
| (MainProcess pid=44936) WARNING 10-31 07:07:37 [utils.py:648] [shutdown] Process manager: force killing remaining processes count=1 | |
| (MainProcess pid=44936) WARNING 10-31 07:07:37 [utils.py:653] [shutdown] Process manager: force killing remaining process EngineCore pid 45494 | |
| (MainProcess pid=44936) INFO 10-31 07:07:37 [core_client.py:774] [shutdown] MPClient: engine manager stopped | |
| (MainProcess pid=44936) INFO 10-31 07:07:37 [core_client.py:775] [shutdown] MPClient: cleaning up background resources | |
| (MainProcess pid=44936) INFO 10-31 07:07:37 [core_client.py:777] [shutdown] MPClient: complete | |
| closed chains 9 of 24 | ids-vs-text mean max|dp| 0.0000 max 0.0001 | argmax agree 9/9 | ids-readout = generated letter 9/9 | gen 455s | |
Xet Storage Details
- Size:
- 30.9 kB
- Xet hash:
- ea48249f3fba86bc31883088653a0e49100028a03890af8ff915f301f8ab8f83
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.