Run r1: logs and results
Browse files- runs/r1/data/gen/chat.jsonl +0 -0
- runs/r1/data/gen/code.jsonl +0 -0
- runs/r1/data/gen/longform.jsonl +0 -0
- runs/r1/data/gen/math.jsonl +0 -0
- runs/r1/logs/generate.log +197 -0
- runs/r1/logs/pipeline.log +1 -0
- runs/r1/logs/pipeline.out +1 -0
- runs/r1/logs/stats.log +1 -0
- runs/r1/logs/supervisor.log +1 -0
- runs/r1/logs/sync.log +1 -0
runs/r1/data/gen/chat.jsonl
ADDED
|
File without changes
|
runs/r1/data/gen/code.jsonl
ADDED
|
File without changes
|
runs/r1/data/gen/longform.jsonl
ADDED
|
File without changes
|
runs/r1/data/gen/math.jsonl
ADDED
|
File without changes
|
runs/r1/logs/generate.log
ADDED
|
@@ -0,0 +1,197 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[02:21:58] chat: 1264 prompts (pool 6788)
|
| 2 |
+
[02:21:58] code: 1223 prompts (pool 9520)
|
| 3 |
+
[02:21:58] math: 1282 prompts (pool 11300)
|
| 4 |
+
[02:21:58] longform: 1056 prompts (pool 12116)
|
| 5 |
+
(MainProcess pid=4633) INFO 11-18 02:22:17 [api_utils.py:286] non-default args: {'tokenizer': '/workspace/emc/models/target', 'dtype': 'bfloat16', 'max_model_len': 2048, 'enable_prefix_caching': False, 'gpu_memory_utilization': 0.9, 'max_num_seqs': 128, 'disable_log_stats': True, 'language_model_only': True, 'generation_config': 'vllm', 'model': '/workspace/emc/models/target'}
|
| 6 |
+
(MainProcess pid=4633) INFO 11-18 02:22:17 [model.py:699] Resolved architecture: Qwen3_5ForConditionalGeneration
|
| 7 |
+
(MainProcess pid=4633) INFO 11-18 02:22:17 [model.py:2109] Using max model len 2048
|
| 8 |
+
(MainProcess pid=4633) WARNING 11-18 02:22:17 [model.py:1069] Model does not support mm_device_do_normalize, forcing mm_device_do_normalize = False.
|
| 9 |
+
(MainProcess pid=4633) INFO 11-18 02:22:20 [model.py:917] All limits of multimodal modalities supported by the model are set to 0, running in text-only mode.
|
| 10 |
+
(MainProcess pid=4633) INFO 11-18 02:22:20 [model.py:974] Disabled mm_prefix attention mode because multimodal inputs are configuration-disabled. Attention backends without mm_prefix support may now be selected.
|
| 11 |
+
(MainProcess pid=4633) INFO 11-18 02:22:20 [scheduler.py:308] Chunked prefill is enabled with max_num_batched_tokens=16384.
|
| 12 |
+
(MainProcess pid=4633) INFO 11-18 02:22:20 [kernel.py:408] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'], gelu_and_mul_sparse=['triton', 'native'])
|
| 13 |
+
(MainProcess pid=4633) WARNING 11-18 02:22:23 [system_utils.py:152] 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
|
| 14 |
+
(EngineCore pid=4744) INFO 11-18 02:22:41 [core.py:129] Initializing a V1 LLM engine (v0.31.0) with config: model='/workspace/emc/models/target', speculative_config=None, tokenizer='/workspace/emc/models/target', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=2048, 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, 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=0, served_model_name=/workspace/emc/models/target, enable_prefix_caching=False, 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::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, 'compile_sizes': [], 'compile_ranges_endpoints': [16384], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.FULL_AND_PIECEWISE: (2, 1)>, 'cudagraph_num_of_warmups': 1, 'cudagraph_capture_sizes': [1, 2, 4, 8, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120, 128, 136, 144, 152, 160, 168, 176, 184, 192, 200, 208, 216, 224, 232, 240, 248, 256], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': False, 'fuse_act_quant': False, 'fuse_attn_quant': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'enable_qk_norm_rope_fusion': False, 'fuse_rope_kvcache_cat_mla': False, 'fuse_act_padding': False, 'fuse_qk_norm_rope_kvcache': False}, 'max_cudagraph_capture_size': 256, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': False, 'static_all_moe_layers': []}, kernel_config=KernelConfig(ir_op_priority=IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'], gelu_and_mul_sparse=['triton', 'native']), enable_flashinfer_autotune=True, enable_cutedsl_warmup=True, enable_jit_warmup=True, moe_backend='auto', sparse_indexer_topk_backend='auto', linear_backend='auto', linear_backend_per_quant=None)
|
| 15 |
+
(EngineCore pid=4744) INFO 11-18 02:22:43 [parallel_state.py:1825] world_size=1 rank=0 local_rank=0 distributed_init_method=file:///workspace/emc/tmp/vllm_dist_07854923a1b84b13a43e6a77cb5eee25 backend=nccl
|
| 16 |
+
(EngineCore pid=4744) INFO 11-18 02:22:43 [parallel_state.py:2269] 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
|
| 17 |
+
(EngineCore pid=4744) INFO 11-18 02:22:43 [gpu_worker.py:479] Using V2 Model Runner
|
| 18 |
+
(EngineCore pid=4744) INFO 11-18 02:22:44 [model_runner.py:375] Loading model from scratch...
|
| 19 |
+
(EngineCore pid=4744) INFO 11-18 02:22:46 [cuda.py:585] Using backend AttentionBackendEnum.FLASH_ATTN for vit attention
|
| 20 |
+
(EngineCore pid=4744) INFO 11-18 02:22:46 [mm_encoder_attention.py:372] Using AttentionBackendEnum.FLASH_ATTN for MMEncoderAttention.
|
| 21 |
+
(EngineCore pid=4744) INFO 11-18 02:22:46 [qwen_gdn_linear_attn.py:177] Using FlashInfer GDN prefill kernel (requested=auto, head_k_dim=128).
|
| 22 |
+
(EngineCore pid=4744) INFO 11-18 02:22:46 [qwen_gdn_linear_attn.py:533] GDN decode kernel: cuda
|
| 23 |
+
(EngineCore pid=4744) INFO 11-18 02:22:49 [cuda.py:526] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
|
| 24 |
+
(EngineCore pid=4744) INFO 11-18 02:22:49 [flash_attn.py:1167] Using FlashAttention version 2
|
| 25 |
+
(EngineCore pid=4744) INFO 11-18 02:22:50 [weight_utils.py:1063] Filesystem type for checkpoints: OVERLAY. Checkpoint size: 51.75 GiB. Available RAM: 156.49 GiB.
|
| 26 |
+
(EngineCore pid=4744) INFO 11-18 02:22:50 [weight_utils.py:1086] Auto-prefetch is disabled because the filesystem (OVERLAY) is not a recognized network FS (NFS/Lustre). If you want to force prefetching, start vLLM with --safetensors-load-strategy=prefetch.
|
| 27 |
+
|
| 28 |
+
Loading safetensors checkpoint shards: 0% Completed | 0/18 [00:00<?, ?it/s]
|
| 29 |
+
|
| 30 |
+
Loading safetensors checkpoint shards: 6% Completed | 1/18 [00:00<00:08, 1.96it/s]
|
| 31 |
+
|
| 32 |
+
Loading safetensors checkpoint shards: 11% Completed | 2/18 [00:01<00:08, 1.93it/s]
|
| 33 |
+
|
| 34 |
+
Loading safetensors checkpoint shards: 17% Completed | 3/18 [00:01<00:06, 2.16it/s]
|
| 35 |
+
|
| 36 |
+
Loading safetensors checkpoint shards: 22% Completed | 4/18 [00:02<00:07, 1.87it/s]
|
| 37 |
+
|
| 38 |
+
Loading safetensors checkpoint shards: 28% Completed | 5/18 [00:02<00:06, 2.15it/s]
|
| 39 |
+
|
| 40 |
+
Loading safetensors checkpoint shards: 33% Completed | 6/18 [00:03<00:06, 1.85it/s]
|
| 41 |
+
|
| 42 |
+
Loading safetensors checkpoint shards: 39% Completed | 7/18 [00:03<00:05, 2.10it/s]
|
| 43 |
+
|
| 44 |
+
Loading safetensors checkpoint shards: 44% Completed | 8/18 [00:04<00:05, 1.82it/s]
|
| 45 |
+
|
| 46 |
+
Loading safetensors checkpoint shards: 50% Completed | 9/18 [00:04<00:04, 2.04it/s]
|
| 47 |
+
|
| 48 |
+
Loading safetensors checkpoint shards: 56% Completed | 10/18 [00:05<00:04, 1.78it/s]
|
| 49 |
+
|
| 50 |
+
Loading safetensors checkpoint shards: 61% Completed | 11/18 [00:05<00:03, 2.00it/s]
|
| 51 |
+
|
| 52 |
+
Loading safetensors checkpoint shards: 67% Completed | 12/18 [00:06<00:03, 1.76it/s]
|
| 53 |
+
|
| 54 |
+
Loading safetensors checkpoint shards: 72% Completed | 13/18 [00:06<00:02, 1.97it/s]
|
| 55 |
+
|
| 56 |
+
Loading safetensors checkpoint shards: 78% Completed | 14/18 [00:07<00:02, 1.75it/s]
|
| 57 |
+
|
| 58 |
+
Loading safetensors checkpoint shards: 83% Completed | 15/18 [00:07<00:01, 1.94it/s]
|
| 59 |
+
|
| 60 |
+
Loading safetensors checkpoint shards: 89% Completed | 16/18 [00:08<00:01, 1.73it/s]
|
| 61 |
+
|
| 62 |
+
Loading safetensors checkpoint shards: 94% Completed | 17/18 [00:08<00:00, 1.95it/s]
|
| 63 |
+
|
| 64 |
+
Loading safetensors checkpoint shards: 100% Completed | 18/18 [00:09<00:00, 2.01it/s]
|
| 65 |
+
|
| 66 |
+
Loading safetensors checkpoint shards: 100% Completed | 18/18 [00:09<00:00, 1.93it/s]
|
| 67 |
+
|
| 68 |
+
(EngineCore pid=4744) INFO 11-18 02:22:59 [default_loader.py:484] Loading weights took 9.35 seconds
|
| 69 |
+
(EngineCore pid=4744) INFO 11-18 02:23:00 [model_runner.py:407] Model loading took 50.22 GiB memory and 16.360078 seconds
|
| 70 |
+
(EngineCore pid=4744) INFO 11-18 02:23:00 [topk_topp_sampler.py:96] FlashInfer top-p/top-k sampling disabled via VLLM_USE_FLASHINFER_SAMPLER=0.
|
| 71 |
+
(EngineCore pid=4744) WARNING 11-18 02:23:00 [torch_utils.py:274] OMP_NUM_THREADS=20 is set; leaving Torch threads at 20 for serving. Multi-threaded torch CPU ops during serving can degrade performance through spin-wait contention and cgroup CPU-quota throttling.
|
| 72 |
+
(EngineCore pid=4744) INFO 11-18 02:23:00 [interface.py:989] Setting attention block size to 784 tokens to ensure that attention page size is >= mamba page size.
|
| 73 |
+
(EngineCore pid=4744) INFO 11-18 02:23:00 [interface.py:1013] Padding mamba page size by 0.13% to ensure that mamba page size and attention page size are exactly equal.
|
| 74 |
+
(EngineCore pid=4744) INFO 11-18 02:23:00 [utils.py:343] Using LBNHC KV cache layout.
|
| 75 |
+
(EngineCore pid=4744) INFO 11-18 02:23:10 [backends.py:1090] Using cache directory: /workspace/emc/work/cache/vllm/torch_compile_cache/b2387587fa/rank_0_0/backbone for vLLM's torch.compile
|
| 76 |
+
(EngineCore pid=4744) INFO 11-18 02:23:10 [backends.py:1151] Dynamo bytecode transform time: 9.10 s
|
| 77 |
+
(EngineCore pid=4744) INFO 11-18 02:23:28 [backends.py:391] Compiling a graph for compile range (1, 16384) takes 17.89 s
|
| 78 |
+
(EngineCore pid=4744) INFO 11-18 02:23:33 [backends.py:916] collected artifacts: 65 entries, 20 artifacts, 22721791 bytes total
|
| 79 |
+
(EngineCore pid=4744) INFO 11-18 02:23:33 [decorators.py:717] saved AOT compiled function to /workspace/emc/work/cache/vllm/torch_compile_cache/torch_aot_compile/03b855643ea6994e6b652654ffa3a4dbaf23ff3c1304eff96893a890b9176365/rank_0_0/model
|
| 80 |
+
(EngineCore pid=4744) INFO 11-18 02:23:33 [monitor.py:53] torch.compile took 32.10 s in total
|
| 81 |
+
(EngineCore pid=4744) INFO 11-18 02:23:50 [monitor.py:81] Initial profiling/warmup run took 16.86 s
|
| 82 |
+
|
| 83 |
+
Capturing CUDA graphs (PIECEWISE): 0%| | 0/35 [00:00<?, ?it/s]
|
| 84 |
+
Capturing CUDA graphs (PIECEWISE): 3%|▎ | 1/35 [00:08<04:35, 8.11s/it]
|
| 85 |
+
Capturing CUDA graphs (PIECEWISE): 6%|▌ | 2/35 [00:08<02:06, 3.82s/it]
|
| 86 |
+
Capturing CUDA graphs (PIECEWISE): 9%|▊ | 3/35 [00:09<01:12, 2.28s/it]
|
| 87 |
+
Capturing CUDA graphs (PIECEWISE): 11%|█▏ | 4/35 [00:09<00:47, 1.55s/it]
|
| 88 |
+
Capturing CUDA graphs (PIECEWISE): 14%|█▍ | 5/35 [00:10<00:34, 1.14s/it]
|
| 89 |
+
Capturing CUDA graphs (PIECEWISE): 17%|█▋ | 6/35 [00:10<00:25, 1.13it/s]
|
| 90 |
+
Capturing CUDA graphs (PIECEWISE): 20%|██ | 7/35 [00:10<00:20, 1.39it/s]
|
| 91 |
+
Capturing CUDA graphs (PIECEWISE): 23%|██▎ | 8/35 [00:11<00:16, 1.65it/s]
|
| 92 |
+
Capturing CUDA graphs (PIECEWISE): 26%|██▌ | 9/35 [00:11<00:13, 1.92it/s]
|
| 93 |
+
Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 10/35 [00:12<00:11, 2.18it/s]
|
| 94 |
+
Capturing CUDA graphs (PIECEWISE): 31%|███▏ | 11/35 [00:12<00:09, 2.42it/s]
|
| 95 |
+
Capturing CUDA graphs (PIECEWISE): 34%|███▍ | 12/35 [00:12<00:08, 2.67it/s]
|
| 96 |
+
Capturing CUDA graphs (PIECEWISE): 37%|███▋ | 13/35 [00:12<00:07, 2.92it/s]
|
| 97 |
+
Capturing CUDA graphs (PIECEWISE): 40%|████ | 14/35 [00:13<00:06, 3.20it/s]
|
| 98 |
+
Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 15/35 [00:13<00:05, 3.57it/s]
|
| 99 |
+
Capturing CUDA graphs (PIECEWISE): 46%|████▌ | 16/35 [00:13<00:04, 3.81it/s]
|
| 100 |
+
Capturing CUDA graphs (PIECEWISE): 49%|████▊ | 17/35 [00:14<00:09, 1.98it/s]
|
| 101 |
+
Capturing CUDA graphs (PIECEWISE): 51%|█████▏ | 18/35 [00:14<00:07, 2.22it/s]
|
| 102 |
+
Capturing CUDA graphs (PIECEWISE): 54%|█████▍ | 19/35 [00:15<00:05, 2.90it/s]
|
| 103 |
+
Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 20/35 [00:15<00:04, 3.66it/s]
|
| 104 |
+
Capturing CUDA graphs (PIECEWISE): 60%|██████ | 21/35 [00:15<00:03, 4.48it/s]
|
| 105 |
+
Capturing CUDA graphs (PIECEWISE): 63%|██████▎ | 22/35 [00:15<00:02, 5.15it/s]
|
| 106 |
+
Capturing CUDA graphs (PIECEWISE): 66%|██████▌ | 23/35 [00:15<00:02, 5.62it/s]
|
| 107 |
+
Capturing CUDA graphs (PIECEWISE): 69%|██████▊ | 24/35 [00:15<00:01, 6.05it/s]
|
| 108 |
+
Capturing CUDA graphs (PIECEWISE): 71%|███████▏ | 25/35 [00:15<00:01, 6.44it/s]
|
| 109 |
+
Capturing CUDA graphs (PIECEWISE): 74%|███████▍ | 26/35 [00:15<00:01, 6.93it/s]
|
| 110 |
+
Capturing CUDA graphs (PIECEWISE): 77%|███████▋ | 27/35 [00:16<00:01, 7.25it/s]
|
| 111 |
+
Capturing CUDA graphs (PIECEWISE): 80%|████████ | 28/35 [00:16<00:00, 7.42it/s]
|
| 112 |
+
Capturing CUDA graphs (PIECEWISE): 83%|████████▎ | 29/35 [00:16<00:00, 7.67it/s]
|
| 113 |
+
Capturing CUDA graphs (PIECEWISE): 86%|████████▌ | 30/35 [00:16<00:00, 7.97it/s]
|
| 114 |
+
Capturing CUDA graphs (PIECEWISE): 89%|████████▊ | 31/35 [00:16<00:00, 8.14it/s]
|
| 115 |
+
Capturing CUDA graphs (PIECEWISE): 91%|█████████▏| 32/35 [00:16<00:00, 8.22it/s]
|
| 116 |
+
Capturing CUDA graphs (PIECEWISE): 94%|█████████▍| 33/35 [00:16<00:00, 8.11it/s]
|
| 117 |
+
Capturing CUDA graphs (PIECEWISE): 97%|█████████▋| 34/35 [00:16<00:00, 8.14it/s]
|
| 118 |
+
Capturing CUDA graphs (PIECEWISE): 100%|██████████| 35/35 [00:17<00:00, 2.63it/s]
|
| 119 |
+
Capturing CUDA graphs (PIECEWISE): 100%|██████████| 35/35 [00:17<00:00, 1.96it/s]
|
| 120 |
+
|
| 121 |
+
Capturing CUDA graphs (FULL): 0%| | 0/2 [00:00<?, ?it/s]
|
| 122 |
+
Capturing CUDA graphs (FULL): 100%|██████████| 2/2 [00:00<00:00, 10.49it/s]
|
| 123 |
+
Capturing CUDA graphs (FULL): 100%|████████��█| 2/2 [00:00<00:00, 10.46it/s]
|
| 124 |
+
(EngineCore pid=4744) INFO 11-18 02:24:11 [model_runner.py:1087] Graph capturing finished in 19 secs, took 0.41 GiB
|
| 125 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [gpu_worker.py:692] Available KV cache memory: 31.65 GiB
|
| 126 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [gpu_worker.py:707] CUDA graph memory profiling is enabled (default since v0.21.0). The current --gpu-memory-utilization=0.9000 is equivalent to --gpu-memory-utilization=0.8953 without CUDA graph memory profiling. To maintain the same effective KV cache size as before, increase --gpu-memory-utilization to 0.9047. To disable, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0.
|
| 127 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [kv_cache_utils.py:2464] GPU KV cache size: 225,621 tokens, Maximum concurrency for 2,048 tokens per request: 110.17x
|
| 128 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [kernel_warmup.py:172] JIT kernel warmup starting.
|
| 129 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [kernel_warmup.py:185] JIT kernel warmup finished in 0.00s.
|
| 130 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [qwen_triton_warmup.py:294] Warming up Qwen GDN Triton kernels for model_type=qwen3_5_text.
|
| 131 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [qwen_vl_triton_warmup.py:98] Warmed M-RoPE Triton kernels.
|
| 132 |
+
(EngineCore pid=4744) INFO 11-18 02:24:12 [mamba_triton_warmup.py:42] Warmed Mamba batch_memcpy_kernel.
|
| 133 |
+
(EngineCore pid=4744) INFO 11-18 02:24:13 [kernel_warmup.py:437] Using FlashInfer autotune cache file: /workspace/emc/work/cache/vllm/flashinfer_autotune_cache/0.7.0.post1/120f/0b29de631bfd00f218269fab6c34967e167d577c74fde57f2b4f93715b35db4f/autotune_configs.json
|
| 134 |
+
2026-11-18 02:24:13,880 - INFO - autotuner.py:1046 - flashinfer.jit: [Autotuner]: Autotuning process starts ...
|
| 135 |
+
(EngineCore pid=4744) INFO 11-18 02:24:13 [kernel_warmup.py:359] Running FlashInfer autotune with 16384 tokens and token buckets (1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 768, 1024, 1280, 1536, 1792, 2048, 2560, 3072, 3584, 4096, 8192, 16384).
|
| 136 |
+
2026-11-18 02:24:13,906 - INFO - autotuner.py:1069 - flashinfer.jit: [Autotuner]: Autotuning process ends
|
| 137 |
+
2026-11-18 02:24:14,092 - INFO - autotuner.py:3554 - flashinfer.jit: [Autotuner]: Saved 0 configs to /workspace/emc/work/cache/vllm/flashinfer_autotune_cache/0.7.0.post1/120f/0b29de631bfd00f218269fab6c34967e167d577c74fde57f2b4f93715b35db4f/autotune_configs.json (0 new, 0 from previous config)
|
| 138 |
+
|
| 139 |
+
Capturing CUDA graphs (PIECEWISE): 0%| | 0/35 [00:00<?, ?it/s]
|
| 140 |
+
Capturing CUDA graphs (PIECEWISE): 3%|▎ | 1/35 [00:00<00:17, 1.94it/s]
|
| 141 |
+
Capturing CUDA graphs (PIECEWISE): 6%|▌ | 2/35 [00:00<00:16, 2.06it/s]
|
| 142 |
+
Capturing CUDA graphs (PIECEWISE): 9%|▊ | 3/35 [00:01<00:14, 2.13it/s]
|
| 143 |
+
Capturing CUDA graphs (PIECEWISE): 11%|█▏ | 4/35 [00:01<00:13, 2.24it/s]
|
| 144 |
+
Capturing CUDA graphs (PIECEWISE): 14%|█▍ | 5/35 [00:02<00:12, 2.34it/s]
|
| 145 |
+
Capturing CUDA graphs (PIECEWISE): 17%|█▋ | 6/35 [00:02<00:11, 2.43it/s]
|
| 146 |
+
Capturing CUDA graphs (PIECEWISE): 20%|██ | 7/35 [00:02<00:11, 2.54it/s]
|
| 147 |
+
Capturing CUDA graphs (PIECEWISE): 23%|██▎ | 8/35 [00:03<00:10, 2.65it/s]
|
| 148 |
+
Capturing CUDA graphs (PIECEWISE): 26%|██▌ | 9/35 [00:03<00:09, 2.77it/s]
|
| 149 |
+
Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 10/35 [00:03<00:08, 2.88it/s]
|
| 150 |
+
Capturing CUDA graphs (PIECEWISE): 31%|███▏ | 11/35 [00:04<00:07, 3.01it/s]
|
| 151 |
+
Capturing CUDA graphs (PIECEWISE): 34%|███▍ | 12/35 [00:04<00:07, 3.13it/s]
|
| 152 |
+
Capturing CUDA graphs (PIECEWISE): 37%|███▋ | 13/35 [00:04<00:06, 3.32it/s]
|
| 153 |
+
Capturing CUDA graphs (PIECEWISE): 40%|████ | 14/35 [00:05<00:05, 3.62it/s]
|
| 154 |
+
Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 15/35 [00:05<00:05, 3.88it/s]
|
| 155 |
+
Capturing CUDA graphs (PIECEWISE): 46%|████▌ | 16/35 [00:05<00:04, 4.16it/s]
|
| 156 |
+
Capturing CUDA graphs (PIECEWISE): 49%|████▊ | 17/35 [00:05<00:03, 4.70it/s]
|
| 157 |
+
Capturing CUDA graphs (PIECEWISE): 51%|█████▏ | 18/35 [00:05<00:03, 5.21it/s]
|
| 158 |
+
Capturing CUDA graphs (PIECEWISE): 54%|█████▍ | 19/35 [00:05<00:02, 5.58it/s]
|
| 159 |
+
Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 20/35 [00:06<00:02, 5.91it/s]
|
| 160 |
+
Capturing CUDA graphs (PIECEWISE): 60%|██████ | 21/35 [00:06<00:02, 6.25it/s]
|
| 161 |
+
Capturing CUDA graphs (PIECEWISE): 63%|██████▎ | 22/35 [00:06<00:02, 6.39it/s]
|
| 162 |
+
Capturing CUDA graphs (PIECEWISE): 66%|██████▌ | 23/35 [00:06<00:01, 6.46it/s]
|
| 163 |
+
Capturing CUDA graphs (PIECEWISE): 69%|██████▊ | 24/35 [00:06<00:01, 6.46it/s]
|
| 164 |
+
Capturing CUDA graphs (PIECEWISE): 71%|███████▏ | 25/35 [00:06<00:01, 6.62it/s]
|
| 165 |
+
Capturing CUDA graphs (PIECEWISE): 74%|███████▍ | 26/35 [00:06<00:01, 6.76it/s]
|
| 166 |
+
Capturing CUDA graphs (PIECEWISE): 77%|███████▋ | 27/35 [00:07<00:01, 6.85it/s]
|
| 167 |
+
Capturing CUDA graphs (PIECEWISE): 80%|████████ | 28/35 [00:07<00:01, 6.92it/s]
|
| 168 |
+
Capturing CUDA graphs (PIECEWISE): 83%|████████▎ | 29/35 [00:07<00:00, 6.84it/s]
|
| 169 |
+
Capturing CUDA graphs (PIECEWISE): 86%|████████▌ | 30/35 [00:07<00:00, 6.85it/s]
|
| 170 |
+
Capturing CUDA graphs (PIECEWISE): 89%|████████▊ | 31/35 [00:07<00:00, 6.76it/s]
|
| 171 |
+
Capturing CUDA graphs (PIECEWISE): 91%|█████████▏| 32/35 [00:07<00:00, 6.74it/s]
|
| 172 |
+
Capturing CUDA graphs (PIECEWISE): 94%|█████████▍| 33/35 [00:07<00:00, 6.68it/s]
|
| 173 |
+
Capturing CUDA graphs (PIECEWISE): 97%|█████████▋| 34/35 [00:08<00:00, 6.53it/s]
|
| 174 |
+
Capturing CUDA graphs (PIECEWISE): 100%|██████████| 35/35 [00:08<00:00, 6.68it/s]
|
| 175 |
+
Capturing CUDA graphs (PIECEWISE): 100%|██████████| 35/35 [00:08<00:00, 4.25it/s]
|
| 176 |
+
|
| 177 |
+
Capturing CUDA graphs (FULL): 0%| | 0/19 [00:00<?, ?it/s]
|
| 178 |
+
Capturing CUDA graphs (FULL): 11%|█ | 2/19 [00:00<00:01, 10.92it/s]
|
| 179 |
+
Capturing CUDA graphs (FULL): 21%|██ | 4/19 [00:00<00:01, 10.92it/s]
|
| 180 |
+
Capturing CUDA graphs (FULL): 32%|███▏ | 6/19 [00:00<00:01, 10.90it/s]
|
| 181 |
+
Capturing CUDA graphs (FULL): 42%|████▏ | 8/19 [00:00<00:01, 10.99it/s]
|
| 182 |
+
Capturing CUDA graphs (FULL): 53%|█████▎ | 10/19 [00:00<00:00, 11.07it/s]
|
| 183 |
+
Capturing CUDA graphs (FULL): 63%|██████▎ | 12/19 [00:01<00:00, 11.24it/s]
|
| 184 |
+
Capturing CUDA graphs (FULL): 74%|███████▎ | 14/19 [00:01<00:00, 10.95it/s]
|
| 185 |
+
Capturing CUDA graphs (FULL): 84%|████████▍ | 16/19 [00:01<00:00, 11.23it/s]
|
| 186 |
+
Capturing CUDA graphs (FULL): 95%|█████████▍| 18/19 [00:01<00:00, 11.41it/s]
|
| 187 |
+
Capturing CUDA graphs (FULL): 100%|██████████| 19/19 [00:01<00:00, 11.27it/s]
|
| 188 |
+
(EngineCore pid=4744) INFO 11-18 02:24:31 [model_runner.py:1087] Graph capturing finished in 11 secs, took 0.19 GiB
|
| 189 |
+
(EngineCore pid=4744) INFO 11-18 02:24:31 [gpu_worker.py:892] CUDA graph pool memory: 0.19 GiB (actual), 0.45 GiB (estimated), difference: 0.25 GiB (130.3%).
|
| 190 |
+
(EngineCore pid=4744) INFO 11-18 02:24:31 [gpu_worker.py:955] Free memory on device (94.43/94.97 GiB) on startup. Desired GPU memory utilization is (0.9, 85.48 GiB). Actual usage is 50.62 GiB for consumed memory (weights + non-torch), 3.21 GiB for peak activation, and 0.19 GiB for CUDAGraph memory. Replace gpu_memory_utilization config with `--kv-cache-memory=33615142708` (31.31 GiB) to fit into requested memory, or `--kv-cache-memory=43227425792` (40.26 GiB) to fully utilize gpu memory. Current kv cache memory in use is 31.65 GiB.
|
| 191 |
+
(EngineCore pid=4744) INFO 11-18 02:24:33 [jit_monitor.py:84] Kernel JIT monitor activated; monitored JIT compilations during inference will use mode=warn.
|
| 192 |
+
(EngineCore pid=4744) INFO 11-18 02:24:33 [core.py:373] init engine (profile, create kv cache, warmup model) took 93.22 s (compilation: 32.10 s)
|
| 193 |
+
(EngineCore pid=4744) INFO 11-18 02:24:33 [kv_cache_utils.py:772] kv cache group sizes [2048, 2048, 2048, 784]
|
| 194 |
+
(EngineCore pid=4744) INFO 11-18 02:24:33 [kv_cache_utils.py:773] kv lcm block sizes 100352
|
| 195 |
+
(EngineCore pid=4744) INFO 11-18 02:24:33 [kernel.py:408] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['native'], fused_add_rms_norm=['native'], gelu_and_mul_sparse=['triton', 'native'])
|
| 196 |
+
(MainProcess pid=4633) INFO 11-18 02:24:34 [hf.py:646] Detected the chat template content format to be 'openai'. You can set `--chat-template-content-format` to override this.
|
| 197 |
+
[02:24:34] engine ready in 140s
|
runs/r1/logs/pipeline.log
CHANGED
|
@@ -74,3 +74,4 @@
|
|
| 74 |
[2026-11-18 02:16:40] skip probe (done)
|
| 75 |
[2026-11-18 02:16:40] start smoke_fp8
|
| 76 |
[2026-11-18 02:21:57] done smoke_fp8 (317s)
|
|
|
|
|
|
| 74 |
[2026-11-18 02:16:40] skip probe (done)
|
| 75 |
[2026-11-18 02:16:40] start smoke_fp8
|
| 76 |
[2026-11-18 02:21:57] done smoke_fp8 (317s)
|
| 77 |
+
[2026-11-18 02:21:58] start generate
|
runs/r1/logs/pipeline.out
CHANGED
|
@@ -74,3 +74,4 @@
|
|
| 74 |
[2026-11-18 02:16:40] skip probe (done)
|
| 75 |
[2026-11-18 02:16:40] start smoke_fp8
|
| 76 |
[2026-11-18 02:21:57] done smoke_fp8 (317s)
|
|
|
|
|
|
| 74 |
[2026-11-18 02:16:40] skip probe (done)
|
| 75 |
[2026-11-18 02:16:40] start smoke_fp8
|
| 76 |
[2026-11-18 02:21:57] done smoke_fp8 (317s)
|
| 77 |
+
[2026-11-18 02:21:58] start generate
|
runs/r1/logs/stats.log
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
[2026-11-18 02:26:41] ram 4.6/172 GB used (cache 0.3, shmem 0.0) | cgroup 4.6/172 GB | disk 66 GB used | gpu 85.7/95.6 GiB, 100%, 350 W | load 0.0 | largest VLLM::EngineCor 4.7 GB, python 2.6 GB, python 0.1 GB
|
runs/r1/logs/supervisor.log
CHANGED
|
@@ -29,3 +29,4 @@ Installed 15 packages in 98ms
|
|
| 29 |
[2026-11-18 02:15:39] code updated to eb139866e50f
|
| 30 |
[2026-11-18 02:16:39] pipeline exited with code 1
|
| 31 |
[2026-11-18 02:16:40] pipeline started (pid 3579) at eb139866e50f
|
|
|
|
|
|
| 29 |
[2026-11-18 02:15:39] code updated to eb139866e50f
|
| 30 |
[2026-11-18 02:16:39] pipeline exited with code 1
|
| 31 |
[2026-11-18 02:16:40] pipeline started (pid 3579) at eb139866e50f
|
| 32 |
+
[2026-11-18 02:22:40] code updated to 37d0ed276569
|
runs/r1/logs/sync.log
CHANGED
|
@@ -8,3 +8,4 @@ uploaded logs/results to mudassiruddin21/EMC:runs/r1
|
|
| 8 |
uploaded logs/results to mudassiruddin21/EMC:runs/r1
|
| 9 |
uploaded logs/results to mudassiruddin21/EMC:runs/r1
|
| 10 |
uploaded logs/results to mudassiruddin21/EMC:runs/r1
|
|
|
|
|
|
| 8 |
uploaded logs/results to mudassiruddin21/EMC:runs/r1
|
| 9 |
uploaded logs/results to mudassiruddin21/EMC:runs/r1
|
| 10 |
uploaded logs/results to mudassiruddin21/EMC:runs/r1
|
| 11 |
+
uploaded logs/results to mudassiruddin21/EMC:runs/r1
|