Add OpProf campaign: protocols, results, patches, run evidence (P0-P6)
Workload-conditioned operator profiling on patched vLLM 0.24.0 + Qwen3-30B-A3B/H20. H1b PASS (irregular patterns carry +23-45pp R64 raggedness, 8-45% token-efficiency loss vs rectangular controls); mechanism decomposition kills the padding narrative and finds the arrival-uniformization artifact (-12.9%); cross-version churn surface shows TP2/MNS64 -29.4% across vLLM 0.20->0.24 while the argmax held. Raw Layer-1 JSONL streams (507 MB) stay on disk, git-ignored; footer sidecars and metrics are tracked. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1 @@
|
||||
SERVER taskset -c 20-39 /tmp/wjh-opprof-phase2-dash0-20260711/.venv/bin/vllm serve /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B --host 127.0.0.1 --port 8501 --served-model-name qwen3-30b-a3b-community --max-num-batched-tokens 8192 --max-num-seqs 16 --tensor-parallel-size 1 --shutdown-timeout 120
|
||||
@@ -0,0 +1 @@
|
||||
{"schema":1,"record_type":"footer_checkpoint","stream":"opprof-v1-dp0-pid2635032-1783867760160342792.jsonl","encoded_records":0,"written_records":0,"dropped_records":0,"last_step_index":null,"checkpoint_wall_ns":1783867763197526318,"flush_interval_seconds":1.0,"final":false}
|
||||
@@ -0,0 +1,126 @@
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:36 [api_utils.py:339] █ █ █▄ ▄█
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:36 [api_utils.py:339] ▄▄ ▄█ █ █ █ ▀▄▀ █ version 0.24.1.dev3+g668cfb7e2
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:36 [api_utils.py:339] █▄█▀ █ █ █ █ model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:36 [api_utils.py:339] ▀▀ ▀▀▀▀▀ ▀▀▀▀▀ ▀ ▀
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:36 [api_utils.py:273] non-default args: {'model_tag': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'host': '127.0.0.1', 'port': 8501, 'model': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'served_model_name': ['qwen3-30b-a3b-community'], 'max_num_batched_tokens': 8192, 'max_num_seqs': 16, 'shutdown_timeout': 120}
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:48 [model.py:598] Resolved architecture: Qwen3MoeForCausalLM
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:48 [model.py:1725] Using max model len 40960
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:48 [scheduler.py:252] Chunked prefill is enabled with max_num_batched_tokens=8192.
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:48 [vllm.py:1006] Asynchronous scheduling is enabled.
|
||||
(APIServer pid=2634246) INFO 07-12 14:47:48 [kernel.py:276] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'])
|
||||
(EngineCore pid=2635032) INFO 07-12 14:47:59 [core.py:114] Initializing a V1 LLM engine (v0.24.1.dev3+g668cfb7e2) with config: model='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', speculative_config=None, tokenizer='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=40960, 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, 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, jit_monitor_verbose=False), seed=0, served_model_name=qwen3-30b-a3b-community, 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::linear_attention', 'vllm::plamo2_mamba_mixer', 'vllm::qwen_gdn_attention_core', 'vllm::gdn_attention_core_xpu', 'vllm::olmo_hybrid_gdn_full_forward', 'vllm::kda_attention', 'vllm::sparse_attn_indexer', 'vllm::rocm_aiter_sparse_attn_indexer', 'vllm::deepseek_v4_attention', '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, 'compile_sizes': [], 'compile_ranges_endpoints': [8192], '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': <CUDAGraphMode.FULL_AND_PIECEWISE: (2, 1)>, 'cudagraph_num_of_warmups': 1, 'cudagraph_capture_sizes': [1, 2, 4, 8, 16, 24, 32], '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, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_rope_kvcache_cat_mla': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 32, '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']), enable_flashinfer_autotune=True, moe_backend='auto', linear_backend='auto')
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:01 [parallel_state.py:1588] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://172.27.132.244:56823 backend=nccl
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:01 [parallel_state.py:1923] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank 0, EPLB rank N/A
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:03 [topk_topp_sampler.py:55] Using FlashInfer for top-p & top-k sampling.
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:03 [gpu_model_runner.py:5164] Starting to load model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B...
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:04 [cuda.py:480] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:04 [flash_attn.py:670] Using FlashAttention version 3
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:04 [unquantized.py:247] Using TRITON Unquantized MoE backend out of potential backends: ['TRITON', 'BATCHED_TRITON', 'FlashInfer TRTLLM', 'FlashInfer CUTLASS'].
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:04 [weight_utils.py:849] Filesystem type for checkpoints: FUSE.ALIYUN-ALINAS-EFC. Checkpoint size: 56.87 GiB. Available RAM: 1284.19 GiB.
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:04 [weight_utils.py:872] Auto-prefetch is disabled because the filesystem (FUSE.ALIYUN-ALINAS-EFC) is not a recognized network FS (NFS/Lustre). If you want to force prefetching, start vLLM with --safetensors-load-strategy=prefetch.
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 0% Completed | 0/16 [00:00<?, ?it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 6% Completed | 1/16 [00:00<00:14, 1.02it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 12% Completed | 2/16 [00:02<00:14, 1.04s/it]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 19% Completed | 3/16 [00:03<00:13, 1.07s/it]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 25% Completed | 4/16 [00:04<00:13, 1.11s/it]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 31% Completed | 5/16 [00:05<00:11, 1.06s/it]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 38% Completed | 6/16 [00:06<00:10, 1.02s/it]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 44% Completed | 7/16 [00:07<00:08, 1.01it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 50% Completed | 8/16 [00:08<00:07, 1.04it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 56% Completed | 9/16 [00:09<00:06, 1.01it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 62% Completed | 10/16 [00:10<00:05, 1.05it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 69% Completed | 11/16 [00:10<00:04, 1.07it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 75% Completed | 12/16 [00:11<00:03, 1.09it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 81% Completed | 13/16 [00:12<00:02, 1.09it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 88% Completed | 14/16 [00:13<00:01, 1.09it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 94% Completed | 15/16 [00:14<00:00, 1.10it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 100% Completed | 16/16 [00:14<00:00, 1.40it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Loading safetensors checkpoint shards: 100% Completed | 16/16 [00:14<00:00, 1.09it/s]
|
||||
(EngineCore pid=2635032)
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:19 [default_loader.py:430] Loading weights took 14.76 seconds
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:19 [unquantized.py:312] Using MoEPrepareAndFinalizeNoDPEPModular
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:20 [gpu_model_runner.py:5259] Model loading took 56.88 GiB memory and 15.453205 seconds
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:29 [backends.py:1089] Using cache directory: /home/admin/cpfs/wjh/.cache/vllm/torch_compile_cache/f4a50989f8/rank_0_0/backbone for vLLM's torch.compile
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:29 [backends.py:1148] Dynamo bytecode transform time: 9.28 s
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:45 [backends.py:378] Cache the graph of compile range (1, 8192) for later use
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:53 [backends.py:393] Compiling a graph for compile range (1, 8192) takes 22.95 s
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:57 [decorators.py:708] saved AOT compiled function to /home/admin/cpfs/wjh/.cache/vllm/torch_compile_cache/torch_aot_compile/fe5a82dbe929018b7aea7dee05b5cd31a21fe2682aca78ee4cbf2b37c8a086d6/rank_0_0/model
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:57 [monitor.py:53] torch.compile took 37.29 s in total
|
||||
(EngineCore pid=2635032) INFO 07-12 14:48:58 [fused_moe.py:1058] Using configuration from /home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20.json for MoE layer.
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:00 [monitor.py:81] Initial profiling/warmup run took 2.18 s
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:07 [gpu_model_runner.py:6487] Profiling CUDA graph memory: PIECEWISE=7 (largest=32), FULL=5 (largest=16)
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:12 [gpu_model_runner.py:6592] Estimated CUDA graph memory: 0.08 GiB total
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:13 [gpu_worker.py:508] Available KV cache memory: 29.43 GiB
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:13 [gpu_worker.py:523] CUDA graph memory profiling is enabled (default since v0.21.0). The current --gpu-memory-utilization=0.9200 is equivalent to --gpu-memory-utilization=0.9191 without CUDA graph memory profiling. To maintain the same effective KV cache size as before, increase --gpu-memory-utilization to 0.9209. To disable, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0.
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:13 [kv_cache_utils.py:2146] GPU KV cache size: 321,456 tokens
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:13 [kv_cache_utils.py:2147] Maximum concurrency for 40,960 tokens per request: 7.85x
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:13 [deep_gemm.py:175] deep_gemm not found in site-packages, trying vendored vllm.third_party.deep_gemm
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:13 [deep_gemm.py:202] DeepGEMM PDL enabled on vllm.third_party.deep_gemm.
|
||||
(EngineCore pid=2635032) 2026-07-12 14:49:13,542 - INFO - autotuner.py:622 - flashinfer.jit: [Autotuner]: Autotuning process starts ...
|
||||
(EngineCore pid=2635032) 2026-07-12 14:49:13,588 - INFO - autotuner.py:641 - flashinfer.jit: [Autotuner]: Autotuning process ends
|
||||
(EngineCore pid=2635032)
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 0%| | 0/7 [00:00<?, ?it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 14%|█▍ | 1/7 [00:00<00:00, 9.22it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 43%|████▎ | 3/7 [00:00<00:00, 9.82it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 57%|█████▋ | 4/7 [00:00<00:00, 9.78it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 71%|███████▏ | 5/7 [00:02<00:01, 1.59it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 86%|████████▌ | 6/7 [00:03<00:00, 1.30it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 7/7 [00:04<00:00, 1.14it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 7/7 [00:04<00:00, 1.63it/s]
|
||||
(EngineCore pid=2635032)
|
||||
Capturing CUDA graphs (decode, FULL): 0%| | 0/5 [00:00<?, ?it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 40%|████ | 2/5 [00:00<00:00, 10.66it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 80%|████████ | 4/5 [00:00<00:00, 10.93it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 100%|██████████| 5/5 [00:00<00:00, 10.96it/s]
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:19 [gpu_model_runner.py:6660] Graph capturing finished in 6 secs, took 0.10 GiB
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:19 [gpu_worker.py:667] CUDA graph pool memory: 0.1 GiB (actual), 0.08 GiB (estimated), difference: 0.02 GiB (15.7%).
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:19 [jit_monitor.py:60] Kernel JIT monitor activated — Triton JIT compilations during inference will be logged as warnings.
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:19 [core.py:337] init engine (profile, create kv cache, warmup model) took 59.33 s (compilation: 37.29 s)
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:20 [scheduler.py:282] OpProf telemetry enabled: /home/admin/cpfs/wjh/opprof-phase6-dash0-20260712/runs/phase6/cells/tp1_mns16/opprof/opprof-v1-dp0-pid2635032-1783867760160342792.jsonl
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:20 [vllm.py:1006] Asynchronous scheduling is enabled.
|
||||
(EngineCore pid=2635032) INFO 07-12 14:49:20 [kernel.py:276] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'])
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [api_server.py:577] Supported tasks: ['generate']
|
||||
(APIServer pid=2634246) WARNING 07-12 14:49:20 [model.py:1477] Default vLLM sampling parameters have been overridden by the model's `generation_config.json`: `{'temperature': 0.6, 'top_k': 20, 'top_p': 0.95}`. If this is not intended, please relaunch vLLM instance with `--generation-config vllm`.
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [hf.py:548] Detected the chat template content format to be 'string'. You can set `--chat-template-content-format` to override this.
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [api_server.py:581] Starting vLLM server on http://127.0.0.1:8501
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:37] Available routes are:
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /openapi.json, Methods: HEAD, GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /docs, Methods: HEAD, GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /docs/oauth2-redirect, Methods: HEAD, GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /redoc, Methods: HEAD, GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /load, Methods: GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /version, Methods: GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /health, Methods: GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /metrics, Methods: GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /tokenize, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /detokenize, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /v1/models, Methods: GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /ping, Methods: GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /ping, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /invocations, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /v1/chat/completions, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /v1/chat/completions/batch, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /v1/responses, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /v1/responses/{response_id}, Methods: GET
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /v1/responses/{response_id}/cancel, Methods: POST
|
||||
(APIServer pid=2634246) INFO 07-12 14:49:20 [launcher.py:46] Route: /v1/completions, Methods: POST
|
||||
@@ -0,0 +1 @@
|
||||
SERVER taskset -c 40-59 /tmp/wjh-opprof-phase2-dash0-20260711/.venv/bin/vllm serve /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B --host 127.0.0.1 --port 8502 --served-model-name qwen3-30b-a3b-community --max-num-batched-tokens 8192 --max-num-seqs 32 --tensor-parallel-size 1 --shutdown-timeout 120
|
||||
@@ -0,0 +1 @@
|
||||
{"schema":1,"record_type":"footer_checkpoint","stream":"opprof-v1-dp0-pid2635031-1783867762612574727.jsonl","encoded_records":0,"written_records":0,"dropped_records":0,"last_step_index":null,"checkpoint_wall_ns":1783867763616401470,"flush_interval_seconds":1.0,"final":false}
|
||||
@@ -0,0 +1,126 @@
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:36 [api_utils.py:339] █ █ █▄ ▄█
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:36 [api_utils.py:339] ▄▄ ▄█ █ █ █ ▀▄▀ █ version 0.24.1.dev3+g668cfb7e2
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:36 [api_utils.py:339] █▄█▀ █ █ █ █ model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:36 [api_utils.py:339] ▀▀ ▀▀▀▀▀ ▀▀▀▀▀ ▀ ▀
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:36 [api_utils.py:273] non-default args: {'model_tag': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'host': '127.0.0.1', 'port': 8502, 'model': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'served_model_name': ['qwen3-30b-a3b-community'], 'max_num_batched_tokens': 8192, 'max_num_seqs': 32, 'shutdown_timeout': 120}
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:48 [model.py:598] Resolved architecture: Qwen3MoeForCausalLM
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:48 [model.py:1725] Using max model len 40960
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:48 [scheduler.py:252] Chunked prefill is enabled with max_num_batched_tokens=8192.
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:48 [vllm.py:1006] Asynchronous scheduling is enabled.
|
||||
(APIServer pid=2634247) INFO 07-12 14:47:48 [kernel.py:276] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'])
|
||||
(EngineCore pid=2635031) INFO 07-12 14:47:59 [core.py:114] Initializing a V1 LLM engine (v0.24.1.dev3+g668cfb7e2) with config: model='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', speculative_config=None, tokenizer='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=40960, 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, 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, jit_monitor_verbose=False), seed=0, served_model_name=qwen3-30b-a3b-community, 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::linear_attention', 'vllm::plamo2_mamba_mixer', 'vllm::qwen_gdn_attention_core', 'vllm::gdn_attention_core_xpu', 'vllm::olmo_hybrid_gdn_full_forward', 'vllm::kda_attention', 'vllm::sparse_attn_indexer', 'vllm::rocm_aiter_sparse_attn_indexer', 'vllm::deepseek_v4_attention', '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, 'compile_sizes': [], 'compile_ranges_endpoints': [8192], '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': <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], '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, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_rope_kvcache_cat_mla': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 64, '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']), enable_flashinfer_autotune=True, moe_backend='auto', linear_backend='auto')
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:02 [parallel_state.py:1588] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://172.27.132.244:47899 backend=nccl
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:02 [parallel_state.py:1923] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank 0, EPLB rank N/A
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:03 [topk_topp_sampler.py:55] Using FlashInfer for top-p & top-k sampling.
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:03 [gpu_model_runner.py:5164] Starting to load model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B...
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:04 [cuda.py:480] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:04 [flash_attn.py:670] Using FlashAttention version 3
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:04 [unquantized.py:247] Using TRITON Unquantized MoE backend out of potential backends: ['TRITON', 'BATCHED_TRITON', 'FlashInfer TRTLLM', 'FlashInfer CUTLASS'].
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:04 [weight_utils.py:849] Filesystem type for checkpoints: FUSE.ALIYUN-ALINAS-EFC. Checkpoint size: 56.87 GiB. Available RAM: 1284.19 GiB.
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:04 [weight_utils.py:872] Auto-prefetch is disabled because the filesystem (FUSE.ALIYUN-ALINAS-EFC) is not a recognized network FS (NFS/Lustre). If you want to force prefetching, start vLLM with --safetensors-load-strategy=prefetch.
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 0% Completed | 0/16 [00:00<?, ?it/s]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 6% Completed | 1/16 [00:00<00:14, 1.02it/s]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 12% Completed | 2/16 [00:02<00:14, 1.04s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 19% Completed | 3/16 [00:03<00:13, 1.07s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 25% Completed | 4/16 [00:04<00:13, 1.11s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 31% Completed | 5/16 [00:05<00:11, 1.08s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 38% Completed | 6/16 [00:06<00:10, 1.05s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 44% Completed | 7/16 [00:07<00:09, 1.01s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 50% Completed | 8/16 [00:08<00:07, 1.02it/s]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 56% Completed | 9/16 [00:09<00:06, 1.01it/s]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 62% Completed | 10/16 [00:10<00:06, 1.01s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 69% Completed | 11/16 [00:11<00:05, 1.03s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 75% Completed | 12/16 [00:12<00:04, 1.02s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 81% Completed | 13/16 [00:13<00:03, 1.05s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 88% Completed | 14/16 [00:14<00:02, 1.07s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 94% Completed | 15/16 [00:15<00:01, 1.09s/it]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 100% Completed | 16/16 [00:16<00:00, 1.18it/s]
|
||||
(EngineCore pid=2635031)
|
||||
Loading safetensors checkpoint shards: 100% Completed | 16/16 [00:16<00:00, 1.00s/it]
|
||||
(EngineCore pid=2635031)
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:20 [default_loader.py:430] Loading weights took 16.02 seconds
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:20 [unquantized.py:312] Using MoEPrepareAndFinalizeNoDPEPModular
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:21 [gpu_model_runner.py:5259] Model loading took 56.88 GiB memory and 16.709959 seconds
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:31 [backends.py:1089] Using cache directory: /home/admin/cpfs/wjh/.cache/vllm/torch_compile_cache/30ef41b5f5/rank_0_0/backbone for vLLM's torch.compile
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:31 [backends.py:1148] Dynamo bytecode transform time: 9.21 s
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:46 [backends.py:378] Cache the graph of compile range (1, 8192) for later use
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:54 [backends.py:393] Compiling a graph for compile range (1, 8192) takes 23.51 s
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:59 [decorators.py:708] saved AOT compiled function to /home/admin/cpfs/wjh/.cache/vllm/torch_compile_cache/torch_aot_compile/738c624149a63d13eaf115eec4d2189ece948ac500e524d3e06e801d9915352d/rank_0_0/model
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:59 [monitor.py:53] torch.compile took 37.55 s in total
|
||||
(EngineCore pid=2635031) INFO 07-12 14:48:59 [fused_moe.py:1058] Using configuration from /home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20.json for MoE layer.
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:01 [monitor.py:81] Initial profiling/warmup run took 2.33 s
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:08 [gpu_model_runner.py:6487] Profiling CUDA graph memory: PIECEWISE=11 (largest=64), FULL=7 (largest=32)
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:13 [gpu_model_runner.py:6592] Estimated CUDA graph memory: 0.11 GiB total
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:13 [gpu_worker.py:508] Available KV cache memory: 29.4 GiB
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:13 [gpu_worker.py:523] CUDA graph memory profiling is enabled (default since v0.21.0). The current --gpu-memory-utilization=0.9200 is equivalent to --gpu-memory-utilization=0.9188 without CUDA graph memory profiling. To maintain the same effective KV cache size as before, increase --gpu-memory-utilization to 0.9212. To disable, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0.
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:13 [kv_cache_utils.py:2146] GPU KV cache size: 321,136 tokens
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:13 [kv_cache_utils.py:2147] Maximum concurrency for 40,960 tokens per request: 7.84x
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:13 [deep_gemm.py:175] deep_gemm not found in site-packages, trying vendored vllm.third_party.deep_gemm
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:13 [deep_gemm.py:202] DeepGEMM PDL enabled on vllm.third_party.deep_gemm.
|
||||
(EngineCore pid=2635031) 2026-07-12 14:49:13,993 - INFO - autotuner.py:622 - flashinfer.jit: [Autotuner]: Autotuning process starts ...
|
||||
(EngineCore pid=2635031) 2026-07-12 14:49:14,032 - INFO - autotuner.py:641 - flashinfer.jit: [Autotuner]: Autotuning process ends
|
||||
(EngineCore pid=2635031)
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 0%| | 0/11 [00:00<?, ?it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 18%|█▊ | 2/11 [00:00<00:00, 10.96it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 36%|███▋ | 4/11 [00:00<00:00, 10.67it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 55%|█████▍ | 6/11 [00:00<00:00, 10.68it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 73%|███████▎ | 8/11 [00:02<00:01, 2.04it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 82%|████████▏ | 9/11 [00:04<00:01, 1.37it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 91%|█████████ | 10/11 [00:05<00:00, 1.25it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 11/11 [00:06<00:00, 1.17it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 11/11 [00:06<00:00, 1.72it/s]
|
||||
(EngineCore pid=2635031)
|
||||
Capturing CUDA graphs (decode, FULL): 0%| | 0/7 [00:00<?, ?it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 29%|██▊ | 2/7 [00:00<00:00, 12.14it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 57%|█████▋ | 4/7 [00:00<00:00, 12.24it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 86%|████████▌ | 6/7 [00:00<00:00, 12.26it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 100%|██████████| 7/7 [00:00<00:00, 12.33it/s]
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:21 [gpu_model_runner.py:6660] Graph capturing finished in 8 secs, took 0.13 GiB
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:21 [gpu_worker.py:667] CUDA graph pool memory: 0.13 GiB (actual), 0.11 GiB (estimated), difference: 0.02 GiB (12.1%).
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:21 [jit_monitor.py:60] Kernel JIT monitor activated — Triton JIT compilations during inference will be logged as warnings.
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:22 [core.py:337] init engine (profile, create kv cache, warmup model) took 60.59 s (compilation: 37.55 s)
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:22 [scheduler.py:282] OpProf telemetry enabled: /home/admin/cpfs/wjh/opprof-phase6-dash0-20260712/runs/phase6/cells/tp1_mns32/opprof/opprof-v1-dp0-pid2635031-1783867762612574727.jsonl
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:22 [vllm.py:1006] Asynchronous scheduling is enabled.
|
||||
(EngineCore pid=2635031) INFO 07-12 14:49:22 [kernel.py:276] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'])
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:22 [api_server.py:577] Supported tasks: ['generate']
|
||||
(APIServer pid=2634247) WARNING 07-12 14:49:22 [model.py:1477] Default vLLM sampling parameters have been overridden by the model's `generation_config.json`: `{'temperature': 0.6, 'top_k': 20, 'top_p': 0.95}`. If this is not intended, please relaunch vLLM instance with `--generation-config vllm`.
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [hf.py:548] Detected the chat template content format to be 'string'. You can set `--chat-template-content-format` to override this.
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [api_server.py:581] Starting vLLM server on http://127.0.0.1:8502
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:37] Available routes are:
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /openapi.json, Methods: HEAD, GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /docs, Methods: HEAD, GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /docs/oauth2-redirect, Methods: HEAD, GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /redoc, Methods: HEAD, GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /load, Methods: GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /version, Methods: GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /health, Methods: GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /metrics, Methods: GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /tokenize, Methods: POST
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /detokenize, Methods: POST
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /v1/models, Methods: GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /ping, Methods: GET
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /ping, Methods: POST
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /invocations, Methods: POST
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /v1/chat/completions, Methods: POST
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /v1/chat/completions/batch, Methods: POST
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /v1/responses, Methods: POST
|
||||
(APIServer pid=2634247) INFO 07-12 14:49:23 [launcher.py:46] Route: /v1/responses/{response_id}, Methods: GET
|
||||
@@ -0,0 +1 @@
|
||||
SERVER taskset -c 60-79 /tmp/wjh-opprof-phase2-dash0-20260711/.venv/bin/vllm serve /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B --host 127.0.0.1 --port 8503 --served-model-name qwen3-30b-a3b-community --max-num-batched-tokens 8192 --max-num-seqs 64 --tensor-parallel-size 1 --shutdown-timeout 120
|
||||
@@ -0,0 +1 @@
|
||||
{"schema":1,"record_type":"footer_checkpoint","stream":"opprof-v1-dp0-pid2635034-1783867715629257355.jsonl","encoded_records":0,"written_records":0,"dropped_records":0,"last_step_index":null,"checkpoint_wall_ns":1783867764591942536,"flush_interval_seconds":1.0,"final":false}
|
||||
@@ -0,0 +1,125 @@
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:36 [api_utils.py:339] █ █ █▄ ▄█
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:36 [api_utils.py:339] ▄▄ ▄█ █ █ █ ▀▄▀ █ version 0.24.1.dev3+g668cfb7e2
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:36 [api_utils.py:339] █▄█▀ █ █ █ █ model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:36 [api_utils.py:339] ▀▀ ▀▀▀▀▀ ▀▀▀▀▀ ▀ ▀
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:36 [api_utils.py:273] non-default args: {'model_tag': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'host': '127.0.0.1', 'port': 8503, 'model': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'served_model_name': ['qwen3-30b-a3b-community'], 'max_num_batched_tokens': 8192, 'max_num_seqs': 64, 'shutdown_timeout': 120}
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:48 [model.py:598] Resolved architecture: Qwen3MoeForCausalLM
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:48 [model.py:1725] Using max model len 40960
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:48 [scheduler.py:252] Chunked prefill is enabled with max_num_batched_tokens=8192.
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:48 [vllm.py:1006] Asynchronous scheduling is enabled.
|
||||
(APIServer pid=2634248) INFO 07-12 14:47:48 [kernel.py:276] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'])
|
||||
(EngineCore pid=2635034) INFO 07-12 14:47:59 [core.py:114] Initializing a V1 LLM engine (v0.24.1.dev3+g668cfb7e2) with config: model='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', speculative_config=None, tokenizer='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=40960, 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, 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, jit_monitor_verbose=False), seed=0, served_model_name=qwen3-30b-a3b-community, 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::linear_attention', 'vllm::plamo2_mamba_mixer', 'vllm::qwen_gdn_attention_core', 'vllm::gdn_attention_core_xpu', 'vllm::olmo_hybrid_gdn_full_forward', 'vllm::kda_attention', 'vllm::sparse_attn_indexer', 'vllm::rocm_aiter_sparse_attn_indexer', 'vllm::deepseek_v4_attention', '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, 'compile_sizes': [], 'compile_ranges_endpoints': [8192], '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': <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], '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, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_rope_kvcache_cat_mla': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 128, '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']), enable_flashinfer_autotune=True, moe_backend='auto', linear_backend='auto')
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:02 [parallel_state.py:1588] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://172.27.132.244:40669 backend=nccl
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:02 [parallel_state.py:1923] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank 0, EPLB rank N/A
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:03 [topk_topp_sampler.py:55] Using FlashInfer for top-p & top-k sampling.
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:03 [gpu_model_runner.py:5164] Starting to load model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B...
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:04 [cuda.py:480] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:04 [flash_attn.py:670] Using FlashAttention version 3
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:04 [unquantized.py:247] Using TRITON Unquantized MoE backend out of potential backends: ['TRITON', 'BATCHED_TRITON', 'FlashInfer TRTLLM', 'FlashInfer CUTLASS'].
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:04 [weight_utils.py:849] Filesystem type for checkpoints: FUSE.ALIYUN-ALINAS-EFC. Checkpoint size: 56.87 GiB. Available RAM: 1284.19 GiB.
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:04 [weight_utils.py:872] Auto-prefetch is disabled because the filesystem (FUSE.ALIYUN-ALINAS-EFC) is not a recognized network FS (NFS/Lustre). If you want to force prefetching, start vLLM with --safetensors-load-strategy=prefetch.
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 0% Completed | 0/16 [00:00<?, ?it/s]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 6% Completed | 1/16 [00:00<00:14, 1.02it/s]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 12% Completed | 2/16 [00:02<00:14, 1.04s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 19% Completed | 3/16 [00:03<00:13, 1.07s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 25% Completed | 4/16 [00:04<00:13, 1.11s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 31% Completed | 5/16 [00:05<00:11, 1.08s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 38% Completed | 6/16 [00:06<00:10, 1.03s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 44% Completed | 7/16 [00:07<00:08, 1.01it/s]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 50% Completed | 8/16 [00:08<00:07, 1.05it/s]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 56% Completed | 9/16 [00:09<00:07, 1.01s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 62% Completed | 10/16 [00:10<00:06, 1.02s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 69% Completed | 11/16 [00:11<00:05, 1.04s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 75% Completed | 12/16 [00:12<00:04, 1.03s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 81% Completed | 13/16 [00:13<00:03, 1.05s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 88% Completed | 14/16 [00:14<00:02, 1.08s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 94% Completed | 15/16 [00:15<00:01, 1.09s/it]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 100% Completed | 16/16 [00:16<00:00, 1.17it/s]
|
||||
(EngineCore pid=2635034)
|
||||
Loading safetensors checkpoint shards: 100% Completed | 16/16 [00:16<00:00, 1.00s/it]
|
||||
(EngineCore pid=2635034)
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:20 [default_loader.py:430] Loading weights took 16.02 seconds
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:20 [unquantized.py:312] Using MoEPrepareAndFinalizeNoDPEPModular
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:21 [gpu_model_runner.py:5259] Model loading took 56.88 GiB memory and 16.709414 seconds
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:25 [backends.py:1089] Using cache directory: /home/admin/cpfs/wjh/.cache/vllm/torch_compile_cache/65b50dadd2/rank_0_0/backbone for vLLM's torch.compile
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:25 [backends.py:1148] Dynamo bytecode transform time: 3.29 s
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:27 [backends.py:292] Directly load the compiled graph(s) for compile range (1, 8192) from the cache, took 2.209 s
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:27 [decorators.py:311] Directly load AOT compilation from path /home/admin/cpfs/wjh/.cache/vllm/torch_compile_cache/torch_aot_compile/ab53f49ec98f407fe24fecb037cb59739264f283939c70f41986d8369686d472/rank_0_0/model
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:27 [monitor.py:53] torch.compile took 5.90 s in total
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:27 [fused_moe.py:1058] Using configuration from /home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20.json for MoE layer.
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:27 [monitor.py:81] Initial profiling/warmup run took 0.23 s
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:28 [gpu_model_runner.py:6487] Profiling CUDA graph memory: PIECEWISE=19 (largest=128), FULL=11 (largest=64)
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:30 [gpu_model_runner.py:6592] Estimated CUDA graph memory: 0.19 GiB total
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:30 [gpu_worker.py:508] Available KV cache memory: 29.33 GiB
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:30 [gpu_worker.py:523] CUDA graph memory profiling is enabled (default since v0.21.0). The current --gpu-memory-utilization=0.9200 is equivalent to --gpu-memory-utilization=0.9180 without CUDA graph memory profiling. To maintain the same effective KV cache size as before, increase --gpu-memory-utilization to 0.9220. To disable, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0.
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:30 [kv_cache_utils.py:2146] GPU KV cache size: 320,304 tokens
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:30 [kv_cache_utils.py:2147] Maximum concurrency for 40,960 tokens per request: 7.82x
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:30 [deep_gemm.py:175] deep_gemm not found in site-packages, trying vendored vllm.third_party.deep_gemm
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:30 [deep_gemm.py:202] DeepGEMM PDL enabled on vllm.third_party.deep_gemm.
|
||||
(EngineCore pid=2635034) 2026-07-12 14:48:30,760 - INFO - autotuner.py:622 - flashinfer.jit: [Autotuner]: Autotuning process starts ...
|
||||
(EngineCore pid=2635034) 2026-07-12 14:48:30,810 - INFO - autotuner.py:641 - flashinfer.jit: [Autotuner]: Autotuning process ends
|
||||
(EngineCore pid=2635034)
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 0%| | 0/19 [00:00<?, ?it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 5%|▌ | 1/19 [00:00<00:01, 9.79it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 16%|█▌ | 3/19 [00:00<00:01, 9.68it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 21%|██ | 4/19 [00:00<00:01, 9.69it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 26%|██▋ | 5/19 [00:00<00:01, 9.66it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 32%|███▏ | 6/19 [00:00<00:01, 9.59it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 37%|███▋ | 7/19 [00:00<00:01, 9.43it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 42%|████▏ | 8/19 [00:00<00:01, 9.39it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 47%|████▋ | 9/19 [00:00<00:01, 9.49it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 53%|█████▎ | 10/19 [00:01<00:00, 9.57it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 58%|█████▊ | 11/19 [00:01<00:00, 9.53it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 63%|██████▎ | 12/19 [00:01<00:00, 8.45it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 68%|██████▊ | 13/19 [00:01<00:00, 8.75it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 74%|███████▎ | 14/19 [00:01<00:00, 7.97it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 79%|███████▉ | 15/19 [00:01<00:00, 7.43it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 84%|████████▍ | 16/19 [00:01<00:00, 7.13it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 89%|████████▉ | 17/19 [00:02<00:00, 6.84it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 95%|█████████▍| 18/19 [00:02<00:00, 6.84it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 19/19 [00:02<00:00, 6.74it/s]
|
||||
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 19/19 [00:02<00:00, 8.17it/s]
|
||||
(EngineCore pid=2635034)
|
||||
Capturing CUDA graphs (decode, FULL): 0%| | 0/11 [00:00<?, ?it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 18%|█▊ | 2/11 [00:00<00:00, 10.68it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 36%|███▋ | 4/11 [00:00<00:00, 11.04it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 55%|█████▍ | 6/11 [00:00<00:00, 10.88it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 73%|███████▎ | 8/11 [00:00<00:00, 10.85it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 91%|█████████ | 10/11 [00:00<00:00, 11.11it/s]
|
||||
Capturing CUDA graphs (decode, FULL): 100%|██████████| 11/11 [00:00<00:00, 11.10it/s]
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:34 [gpu_model_runner.py:6660] Graph capturing finished in 4 secs, took 0.24 GiB
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:34 [gpu_worker.py:667] CUDA graph pool memory: 0.24 GiB (actual), 0.19 GiB (estimated), difference: 0.05 GiB (19.8%).
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:34 [jit_monitor.py:60] Kernel JIT monitor activated — Triton JIT compilations during inference will be logged as warnings.
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:35 [core.py:337] init engine (profile, create kv cache, warmup model) took 13.60 s (compilation: 5.90 s)
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:35 [scheduler.py:282] OpProf telemetry enabled: /home/admin/cpfs/wjh/opprof-phase6-dash0-20260712/runs/phase6/cells/tp1_mns64/opprof/opprof-v1-dp0-pid2635034-1783867715629257355.jsonl
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:35 [vllm.py:1006] Asynchronous scheduling is enabled.
|
||||
(EngineCore pid=2635034) INFO 07-12 14:48:35 [kernel.py:276] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'])
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:35 [api_server.py:577] Supported tasks: ['generate']
|
||||
(APIServer pid=2634248) WARNING 07-12 14:48:35 [model.py:1477] Default vLLM sampling parameters have been overridden by the model's `generation_config.json`: `{'temperature': 0.6, 'top_k': 20, 'top_p': 0.95}`. If this is not intended, please relaunch vLLM instance with `--generation-config vllm`.
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [hf.py:548] Detected the chat template content format to be 'string'. You can set `--chat-template-content-format` to override this.
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [api_server.py:581] Starting vLLM server on http://127.0.0.1:8503
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [launcher.py:37] Available routes are:
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [launcher.py:46] Route: /openapi.json, Methods: GET, HEAD
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [launcher.py:46] Route: /docs, Methods: GET, HEAD
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [launcher.py:46] Route: /docs/oauth2-redirect, Methods: GET, HEAD
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [launcher.py:46] Route: /redoc, Methods: GET, HEAD
|
||||
(APIServer pid=2634248) INFO 07-12 14:48:36 [launcher.py:46] Route: /load, Methods: GET
|
||||
@@ -0,0 +1 @@
|
||||
SERVER taskset -c 0-19 /tmp/wjh-opprof-phase2-dash0-20260711/.venv/bin/vllm serve /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B --host 127.0.0.1 --port 8500 --served-model-name qwen3-30b-a3b-community --max-num-batched-tokens 8192 --max-num-seqs 8 --tensor-parallel-size 1 --shutdown-timeout 120
|
||||
@@ -0,0 +1,90 @@
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:36 [api_utils.py:339] █ █ █▄ ▄█
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:36 [api_utils.py:339] ▄▄ ▄█ █ █ █ ▀▄▀ █ version 0.24.1.dev3+g668cfb7e2
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:36 [api_utils.py:339] █▄█▀ █ █ █ █ model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:36 [api_utils.py:339] ▀▀ ▀▀▀▀▀ ▀▀▀▀▀ ▀ ▀
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:36 [api_utils.py:339]
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:36 [api_utils.py:273] non-default args: {'model_tag': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'host': '127.0.0.1', 'port': 8500, 'model': '/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', 'served_model_name': ['qwen3-30b-a3b-community'], 'max_num_batched_tokens': 8192, 'max_num_seqs': 8, 'shutdown_timeout': 120}
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:48 [model.py:598] Resolved architecture: Qwen3MoeForCausalLM
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:48 [model.py:1725] Using max model len 40960
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:48 [scheduler.py:252] Chunked prefill is enabled with max_num_batched_tokens=8192.
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:48 [vllm.py:1006] Asynchronous scheduling is enabled.
|
||||
(APIServer pid=2634245) INFO 07-12 14:47:48 [kernel.py:276] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'])
|
||||
(EngineCore pid=2635013) INFO 07-12 14:47:59 [core.py:114] Initializing a V1 LLM engine (v0.24.1.dev3+g668cfb7e2) with config: model='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', speculative_config=None, tokenizer='/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=40960, 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, 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, jit_monitor_verbose=False), seed=0, served_model_name=qwen3-30b-a3b-community, 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::linear_attention', 'vllm::plamo2_mamba_mixer', 'vllm::qwen_gdn_attention_core', 'vllm::gdn_attention_core_xpu', 'vllm::olmo_hybrid_gdn_full_forward', 'vllm::kda_attention', 'vllm::sparse_attn_indexer', 'vllm::rocm_aiter_sparse_attn_indexer', 'vllm::deepseek_v4_attention', '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, 'compile_sizes': [], 'compile_ranges_endpoints': [8192], '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': <CUDAGraphMode.FULL_AND_PIECEWISE: (2, 1)>, 'cudagraph_num_of_warmups': 1, 'cudagraph_capture_sizes': [1, 2, 4, 8, 16], '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, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_rope_kvcache_cat_mla': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 16, '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']), enable_flashinfer_autotune=True, moe_backend='auto', linear_backend='auto')
|
||||
(EngineCore pid=2635013) INFO 07-12 14:48:02 [parallel_state.py:1588] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://172.27.132.244:57917 backend=nccl
|
||||
(EngineCore pid=2635013) INFO 07-12 14:48:02 [parallel_state.py:1923] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank 0, EPLB rank N/A
|
||||
(EngineCore pid=2635013) INFO 07-12 14:48:03 [topk_topp_sampler.py:55] Using FlashInfer for top-p & top-k sampling.
|
||||
(EngineCore pid=2635013) INFO 07-12 14:48:03 [gpu_model_runner.py:5164] Starting to load model /home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B...
|
||||
(APIServer pid=2634245) Traceback (most recent call last):
|
||||
(APIServer pid=2634245) File "/tmp/wjh-opprof-phase2-dash0-20260711/.venv/bin/vllm", line 8, in <module>
|
||||
(APIServer pid=2634245) sys.exit(main())
|
||||
(APIServer pid=2634245) ^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/entrypoints/cli/main.py", line 95, in main
|
||||
(APIServer pid=2634245) args.dispatch_function(args)
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/entrypoints/cli/serve.py", line 148, in cmd
|
||||
(APIServer pid=2634245) uvloop.run(run_server(args))
|
||||
(APIServer pid=2634245) File "/tmp/wjh-opprof-phase2-dash0-20260711/.venv/lib/python3.12/site-packages/uvloop/__init__.py", line 96, in run
|
||||
(APIServer pid=2634245) return __asyncio.run(
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/usr/lib/python3.12/asyncio/runners.py", line 194, in run
|
||||
(APIServer pid=2634245) return runner.run(main)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/usr/lib/python3.12/asyncio/runners.py", line 118, in run
|
||||
(APIServer pid=2634245) return self._loop.run_until_complete(task)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "uvloop/loop.pyx", line 1512, in uvloop.loop.Loop.run_until_complete
|
||||
(APIServer pid=2634245) File "uvloop/loop.pyx", line 1505, in uvloop.loop.Loop.run_until_complete
|
||||
(APIServer pid=2634245) File "uvloop/loop.pyx", line 1379, in uvloop.loop.Loop.run_forever
|
||||
(APIServer pid=2634245) File "uvloop/loop.pyx", line 557, in uvloop.loop.Loop._run
|
||||
(APIServer pid=2634245) File "uvloop/loop.pyx", line 476, in uvloop.loop.Loop._on_idle
|
||||
(APIServer pid=2634245) File "uvloop/cbhandles.pyx", line 83, in uvloop.loop.Handle._run
|
||||
(APIServer pid=2634245) File "uvloop/cbhandles.pyx", line 61, in uvloop.loop.Handle._run
|
||||
(APIServer pid=2634245) File "/tmp/wjh-opprof-phase2-dash0-20260711/.venv/lib/python3.12/site-packages/uvloop/__init__.py", line 48, in wrapper
|
||||
(APIServer pid=2634245) return await main
|
||||
(APIServer pid=2634245) ^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/entrypoints/openai/api_server.py", line 663, in run_server
|
||||
(APIServer pid=2634245) await run_server_worker(listen_address, sock, args, **uvicorn_kwargs)
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/entrypoints/openai/api_server.py", line 677, in run_server_worker
|
||||
(APIServer pid=2634245) async with build_async_engine_client(
|
||||
(APIServer pid=2634245) File "/usr/lib/python3.12/contextlib.py", line 210, in __aenter__
|
||||
(APIServer pid=2634245) return await anext(self.gen)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/entrypoints/openai/api_server.py", line 99, in build_async_engine_client
|
||||
(APIServer pid=2634245) async with build_async_engine_client_from_engine_args(
|
||||
(APIServer pid=2634245) File "/usr/lib/python3.12/contextlib.py", line 210, in __aenter__
|
||||
(APIServer pid=2634245) return await anext(self.gen)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/entrypoints/openai/api_server.py", line 135, in build_async_engine_client_from_engine_args
|
||||
(APIServer pid=2634245) async_llm = AsyncLLM.from_vllm_config(
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/v1/engine/async_llm.py", line 217, in from_vllm_config
|
||||
(APIServer pid=2634245) return cls(
|
||||
(APIServer pid=2634245) ^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/v1/engine/async_llm.py", line 146, in __init__
|
||||
(APIServer pid=2634245) self.engine_core = EngineCoreClient.make_async_mp_client(
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/tracing/otel.py", line 178, in sync_wrapper
|
||||
(APIServer pid=2634245) return func(*args, **kwargs)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/v1/engine/core_client.py", line 132, in make_async_mp_client
|
||||
(APIServer pid=2634245) return AsyncMPClient(*client_args)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/tracing/otel.py", line 178, in sync_wrapper
|
||||
(APIServer pid=2634245) return func(*args, **kwargs)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/v1/engine/core_client.py", line 963, in __init__
|
||||
(APIServer pid=2634245) super().__init__(
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/v1/engine/core_client.py", line 619, in __init__
|
||||
(APIServer pid=2634245) if not sync_input_socket.poll(
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/tmp/wjh-opprof-phase2-dash0-20260711/.venv/lib/python3.12/site-packages/zmq/sugar/socket.py", line 1062, in poll
|
||||
(APIServer pid=2634245) evts = dict(p.poll(timeout))
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "/tmp/wjh-opprof-phase2-dash0-20260711/.venv/lib/python3.12/site-packages/zmq/sugar/poll.py", line 106, in poll
|
||||
(APIServer pid=2634245) return zmq_poll(self.sockets, timeout=timeout)
|
||||
(APIServer pid=2634245) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
(APIServer pid=2634245) File "zmq/backend/cython/_zmq.py", line 1680, in zmq.backend.cython._zmq.zmq_poll
|
||||
(APIServer pid=2634245) File "zmq/backend/cython/_zmq.py", line 179, in zmq.backend.cython._zmq._check_rc
|
||||
(APIServer pid=2634245) File "/home/admin/cpfs/wjh/opprof-phase2-dash0-20260711/vllm-v0.24.0/vllm/entrypoints/openai/api_server.py", line 658, in _interrupt_init
|
||||
(APIServer pid=2634245) raise KeyboardInterrupt("terminated")
|
||||
(APIServer pid=2634245) KeyboardInterrupt: terminated
|
||||
Reference in New Issue
Block a user