mudassiruddin21 commited on
Commit
79b1df7
·
verified ·
1 Parent(s): e13d0f4

Run r1: logs and results

Browse files
runs/r1/engine_logs/20261118-021649-eval-r1.log ADDED
@@ -0,0 +1,285 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ vllm serve /workspace/emc/models/target --served-model-name target --host 127.0.0.1 --port 65351 --dtype bfloat16 --max-model-len 16384 --gpu-memory-utilization 0.9 --max-num-seqs 4 --seed 0 --no-enable-prefix-caching --generation-config vllm --language-model-only --speculative-config {"method": "mtp", "model": "/workspace/emc/models/probe-fp8", "num_speculative_tokens": 3}
2
+
3
+ (APIServer pid=3632) INFO 11-18 02:17:12 [api_utils.py:347]
4
+ (APIServer pid=3632) INFO 11-18 02:17:12 [api_utils.py:347] █ █ █▄ ▄█
5
+ (APIServer pid=3632) INFO 11-18 02:17:12 [api_utils.py:347] ▄▄ ▄█ █ █ █ ▀▄▀ █ version 0.31.0
6
+ (APIServer pid=3632) INFO 11-18 02:17:12 [api_utils.py:347] █▄█▀ █ █ █ █ model /workspace/emc/models/target
7
+ (APIServer pid=3632) INFO 11-18 02:17:12 [api_utils.py:347] ▀▀ ▀▀▀▀▀ ▀▀▀▀▀ ▀ ▀
8
+ (APIServer pid=3632) INFO 11-18 02:17:12 [api_utils.py:347]
9
+ (APIServer pid=3632) INFO 11-18 02:17:12 [api_utils.py:286] non-default args: {'model_tag': '/workspace/emc/models/target', 'host': '127.0.0.1', 'port': 65351, 'model': '/workspace/emc/models/target', 'dtype': 'bfloat16', 'max_model_len': 16384, 'served_model_name': ['target'], 'generation_config': 'vllm', 'gpu_memory_utilization': 0.9, 'enable_prefix_caching': False, 'language_model_only': True, 'max_num_seqs': 4, 'speculative_config': {'method': 'mtp', 'model': '/workspace/emc/models/probe-fp8', 'num_speculative_tokens': 3}}
10
+ (APIServer pid=3632) INFO 11-18 02:17:12 [model.py:699] Resolved architecture: Qwen3_5ForConditionalGeneration
11
+ (APIServer pid=3632) INFO 11-18 02:17:12 [model.py:2109] Using max model len 16384
12
+ (APIServer pid=3632) INFO 11-18 02:17:14 [model.py:917] All limits of multimodal modalities supported by the model are set to 0, running in text-only mode.
13
+ (APIServer pid=3632) INFO 11-18 02:17:14 [model.py:974] Disabled mm_prefix attention mode because multimodal inputs are configuration-disabled. Attention backends without mm_prefix support may now be selected.
14
+ (APIServer pid=3632) INFO 11-18 02:17:14 [model.py:699] Resolved architecture: Qwen3_5MTP
15
+ (APIServer pid=3632) INFO 11-18 02:17:14 [model.py:2109] Using max model len 262144
16
+ (APIServer pid=3632) WARNING 11-18 02:17:14 [speculative.py:1375] Enabling num_speculative_tokens > 1 will run multiple times of forward on same MTP layer,which may result in lower acceptance rate
17
+ (APIServer pid=3632) INFO 11-18 02:17:14 [speculative.py:1663] Overriding draft model max model len from 262144 to 16384
18
+ (APIServer pid=3632) INFO 11-18 02:17:14 [scheduler.py:308] Chunked prefill is enabled with max_num_batched_tokens=8192.
19
+ (APIServer pid=3632) INFO 11-18 02:17:14 [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'])
20
+ (EngineCore pid=3749) INFO 11-18 02:17:36 [core.py:129] Initializing a V1 LLM engine (v0.31.0) with config: model='/workspace/emc/models/target', speculative_config=SpeculativeConfig(method='mtp', model='/workspace/emc/models/probe-fp8', num_spec_tokens=3), 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=16384, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, decode_context_parallel_size=1, dcp_comm_backend=ag_rs, disable_custom_all_reduce=False, quantization=None, 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=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': [8192], '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], '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': 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'], 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)
21
+ (EngineCore pid=3749) INFO 11-18 02:17:38 [parallel_state.py:1825] world_size=1 rank=0 local_rank=0 distributed_init_method=file:///workspace/emc/tmp/vllm_dist_40d9efd76eb641f5b5a1847a604dc464 backend=nccl
22
+ (EngineCore pid=3749) INFO 11-18 02:17:38 [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
23
+ (EngineCore pid=3749) INFO 11-18 02:17:38 [gpu_worker.py:479] Using V2 Model Runner
24
+ (EngineCore pid=3749) INFO 11-18 02:17:40 [model_runner.py:375] Loading model from scratch...
25
+ (EngineCore pid=3749) INFO 11-18 02:17:41 [cuda.py:585] Using backend AttentionBackendEnum.FLASH_ATTN for vit attention
26
+ (EngineCore pid=3749) INFO 11-18 02:17:41 [mm_encoder_attention.py:372] Using AttentionBackendEnum.FLASH_ATTN for MMEncoderAttention.
27
+ (EngineCore pid=3749) INFO 11-18 02:17:41 [qwen_gdn_linear_attn.py:177] Using FlashInfer GDN prefill kernel (requested=auto, head_k_dim=128).
28
+ (EngineCore pid=3749) INFO 11-18 02:17:41 [qwen_gdn_linear_attn.py:533] GDN decode kernel: cuda
29
+ (EngineCore pid=3749) INFO 11-18 02:17:44 [cuda.py:526] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
30
+ (EngineCore pid=3749) INFO 11-18 02:17:44 [flash_attn.py:1167] Using FlashAttention version 2
31
+ (EngineCore pid=3749) INFO 11-18 02:17:45 [weight_utils.py:1063] Filesystem type for checkpoints: OVERLAY. Checkpoint size: 51.75 GiB. Available RAM: 156.11 GiB.
32
+ (EngineCore pid=3749) INFO 11-18 02:17:45 [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.
33
+
34
+ Loading safetensors checkpoint shards: 0% Completed | 0/18 [00:00<?, ?it/s]
35
+
36
+ Loading safetensors checkpoint shards: 6% Completed | 1/18 [00:00<00:08, 1.90it/s]
37
+
38
+ Loading safetensors checkpoint shards: 11% Completed | 2/18 [00:01<00:08, 1.89it/s]
39
+
40
+ Loading safetensors checkpoint shards: 17% Completed | 3/18 [00:01<00:07, 2.04it/s]
41
+
42
+ Loading safetensors checkpoint shards: 22% Completed | 4/18 [00:02<00:07, 1.77it/s]
43
+
44
+ Loading safetensors checkpoint shards: 28% Completed | 5/18 [00:02<00:06, 2.07it/s]
45
+
46
+ Loading safetensors checkpoint shards: 33% Completed | 6/18 [00:03<00:06, 1.84it/s]
47
+
48
+ Loading safetensors checkpoint shards: 39% Completed | 7/18 [00:03<00:05, 2.08it/s]
49
+
50
+ Loading safetensors checkpoint shards: 44% Completed | 8/18 [00:04<00:05, 1.83it/s]
51
+
52
+ Loading safetensors checkpoint shards: 50% Completed | 9/18 [00:04<00:04, 2.05it/s]
53
+
54
+ Loading safetensors checkpoint shards: 56% Completed | 10/18 [00:05<00:04, 1.78it/s]
55
+
56
+ Loading safetensors checkpoint shards: 61% Completed | 11/18 [00:05<00:03, 2.02it/s]
57
+
58
+ Loading safetensors checkpoint shards: 67% Completed | 12/18 [00:06<00:03, 1.78it/s]
59
+
60
+ Loading safetensors checkpoint shards: 72% Completed | 13/18 [00:06<00:02, 1.98it/s]
61
+
62
+ Loading safetensors checkpoint shards: 78% Completed | 14/18 [00:07<00:02, 1.73it/s]
63
+
64
+ Loading safetensors checkpoint shards: 83% Completed | 15/18 [00:07<00:01, 1.91it/s]
65
+
66
+ Loading safetensors checkpoint shards: 89% Completed | 16/18 [00:08<00:01, 1.69it/s]
67
+
68
+ Loading safetensors checkpoint shards: 94% Completed | 17/18 [00:09<00:00, 1.90it/s]
69
+
70
+ Loading safetensors checkpoint shards: 100% Completed | 18/18 [00:09<00:00, 1.98it/s]
71
+
72
+ Loading safetensors checkpoint shards: 100% Completed | 18/18 [00:09<00:00, 1.90it/s]
73
+
74
+ (EngineCore pid=3749) INFO 11-18 02:17:54 [default_loader.py:484] Loading weights took 9.49 seconds
75
+ (EngineCore pid=3749) INFO 11-18 02:17:54 [weight_utils.py:1063] Filesystem type for checkpoints: OVERLAY. Checkpoint size: 0.50 GiB. Available RAM: 156.87 GiB.
76
+
77
+ Loading safetensors checkpoint shards: 0% Completed | 0/1 [00:00<?, ?it/s]
78
+
79
+ Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00, 2.91it/s]
80
+
81
+ Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00, 2.91it/s]
82
+
83
+ (EngineCore pid=3749) INFO 11-18 02:17:54 [default_loader.py:484] Loading weights took 0.35 seconds
84
+ (EngineCore pid=3749) INFO 11-18 02:17:55 [model_runner.py:407] Model loading took 51.02 GiB memory and 15.606564 seconds
85
+ (EngineCore pid=3749) INFO 11-18 02:17:55 [topk_topp_sampler.py:96] FlashInfer top-p/top-k sampling disabled via VLLM_USE_FLASHINFER_SAMPLER=0.
86
+ (EngineCore pid=3749) WARNING 11-18 02:17:55 [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.
87
+ (EngineCore pid=3749) INFO 11-18 02:17:55 [interface.py:989] Setting attention block size to 800 tokens to ensure that attention page size is >= mamba page size.
88
+ (EngineCore pid=3749) INFO 11-18 02:17:55 [interface.py:1013] Padding mamba page size by 0.25% to ensure that mamba page size and attention page size are exactly equal.
89
+ (EngineCore pid=3749) INFO 11-18 02:17:55 [utils.py:343] Using LBNHC KV cache layout.
90
+ (EngineCore pid=3749) INFO 11-18 02:18:05 [backends.py:1090] Using cache directory: /workspace/emc/work/cache/vllm/torch_compile_cache/070dc5d8c5/rank_0_0/backbone for vLLM's torch.compile
91
+ (EngineCore pid=3749) INFO 11-18 02:18:05 [backends.py:1151] Dynamo bytecode transform time: 8.86 s
92
+ (EngineCore pid=3749) INFO 11-18 02:18:23 [backends.py:391] Compiling a graph for compile range (1, 8192) takes 17.48 s
93
+ (EngineCore pid=3749) INFO 11-18 02:18:27 [backends.py:916] collected artifacts: 65 entries, 20 artifacts, 26015801 bytes total
94
+ (EngineCore pid=3749) INFO 11-18 02:18:27 [decorators.py:717] saved AOT compiled function to /workspace/emc/work/cache/vllm/torch_compile_cache/torch_aot_compile/8a0bc292c67e43045b0fd3aa8597dbd95f65c054bf8675a0816e11306b67b354/rank_0_0/model
95
+ (EngineCore pid=3749) INFO 11-18 02:18:27 [monitor.py:53] torch.compile took 30.97 s in total
96
+ (EngineCore pid=3749) INFO 11-18 02:18:43 [monitor.py:81] Initial profiling/warmup run took 15.19 s
97
+ (EngineCore pid=3749) INFO 11-18 02:18:43 [backends.py:1090] Using cache directory: /workspace/emc/work/cache/vllm/torch_compile_cache/070dc5d8c5/rank_0_0/eagle_head for vLLM's torch.compile
98
+ (EngineCore pid=3749) INFO 11-18 02:18:43 [backends.py:1151] Dynamo bytecode transform time: 0.49 s
99
+ (EngineCore pid=3749) INFO 11-18 02:18:45 [backends.py:391] Compiling a graph for compile range (1, 8192) takes 2.12 s
100
+ (EngineCore pid=3749) INFO 11-18 02:18:45 [backends.py:916] collected artifacts: 2 entries, 2 artifacts, 3523155 bytes total
101
+ (EngineCore pid=3749) INFO 11-18 02:18:45 [decorators.py:717] saved AOT compiled function to /workspace/emc/work/cache/vllm/torch_compile_cache/torch_aot_compile/84189ba671cc1bd1f561bbaa20615ae8afd1e53a54fbbd7848261ae7a96d6fbe/rank_0_0/model
102
+ (EngineCore pid=3749) INFO 11-18 02:18:45 [monitor.py:53] torch.compile took 2.82 s in total
103
+ (EngineCore pid=3749) INFO 11-18 02:18:45 [monitor.py:81] Initial profiling/warmup run took 0.00 s
104
+ (EngineCore pid=3749) WARNING 11-18 02:18:47 [kv_cache_utils.py:1574] Add 3 padding layers, may waste at most 6.25% KV cache memory
105
+
106
+ Capturing CUDA graphs (PIECEWISE): 0%| | 0/7 [00:00<?, ?it/s]
107
+ Capturing CUDA graphs (PIECEWISE): 14%|█▍ | 1/7 [00:07<00:46, 7.70s/it]
108
+ Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 2/7 [00:07<00:16, 3.27s/it]
109
+ Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 3/7 [00:08<00:07, 1.85s/it]
110
+ Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 4/7 [00:08<00:03, 1.18s/it]
111
+ Capturing CUDA graphs (PIECEWISE): 71%|███████▏ | 5/7 [00:08<00:01, 1.21it/s]
112
+ Capturing CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:08<00:00, 2.16it/s]
113
+ Capturing CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:08<00:00, 1.23s/it]
114
+
115
+ Capturing CUDA graphs (FULL): 0%| | 0/2 [00:00<?, ?it/s]
116
+ Capturing CUDA graphs (FULL): 100%|██████████| 2/2 [00:00<00:00, 12.49it/s]
117
+ Capturing CUDA graphs (FULL): 100%|██████████| 2/2 [00:00<00:00, 12.47it/s]
118
+ (EngineCore pid=3749) INFO 11-18 02:18:57 [speculator.py:144] Capturing model for speculator...
119
+
120
+ Capturing prefill CUDA graphs (PIECEWISE): 0%| | 0/7 [00:00<?, ?it/s]
121
+ Capturing prefill CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:00<00:00, 136.03it/s]
122
+
123
+ Capturing prefill CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
124
+ Capturing prefill CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 150.64it/s]
125
+
126
+ Capturing decode CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
127
+ Capturing decode CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 83.67it/s]
128
+ (EngineCore pid=3749) INFO 11-18 02:18:57 [model_runner.py:1087] Graph capturing finished in 10 secs, took 0.23 GiB
129
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [gpu_worker.py:692] Available KV cache memory: 30.91 GiB
130
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [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=0.8975 without CUDA graph memory profiling. To maintain the same effective KV cache size as before, increase --gpu-memory-utilization to 0.9025. To disable, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0.
131
+ (EngineCore pid=3749) WARNING 11-18 02:18:58 [kv_cache_utils.py:1574] Add 3 padding layers, may waste at most 6.25% KV cache memory
132
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [kv_cache_utils.py:2464] GPU KV cache size: 295,408 tokens, Maximum concurrency for 16,384 tokens per request: 18.03x
133
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [kernel_warmup.py:172] JIT kernel warmup starting.
134
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [kernel_warmup.py:185] JIT kernel warmup finished in 0.00s.
135
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [qwen_triton_warmup.py:294] Warming up Qwen GDN Triton kernels for model_type=qwen3_5_text.
136
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [qwen_vl_triton_warmup.py:98] Warmed M-RoPE Triton kernels.
137
+ (EngineCore pid=3749) INFO 11-18 02:18:58 [mamba_triton_warmup.py:42] Warmed Mamba batch_memcpy_kernel.
138
+ (EngineCore pid=3749) INFO 11-18 02:18:59 [kernel_warmup.py:437] Using FlashInfer autotune cache file: /workspace/emc/work/cache/vllm/flashinfer_autotune_cache/0.7.0.post1/120f/0881ec9c562a209723a6b6cf4fcfccee53df672972c395bfe03858f09816333d/autotune_configs.json
139
+ 2026-11-18 02:18:59,888 - INFO - autotuner.py:3691 - flashinfer.jit: [Autotuner]: Loaded 0 configs from /workspace/emc/work/cache/vllm/flashinfer_autotune_cache/0.7.0.post1/120f/0881ec9c562a209723a6b6cf4fcfccee53df672972c395bfe03858f09816333d/autotune_configs.json
140
+ 2026-11-18 02:18:59,888 - INFO - autotuner.py:1046 - flashinfer.jit: [Autotuner]: Autotuning process starts ...
141
+ (EngineCore pid=3749) INFO 11-18 02:18:59 [kernel_warmup.py:359] Running FlashInfer autotune with 8192 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).
142
+ 2026-11-18 02:18:59,917 - INFO - autotuner.py:1069 - flashinfer.jit: [Autotuner]: Autotuning process ends
143
+ 2026-11-18 02:18:59,919 - INFO - autotuner.py:3554 - flashinfer.jit: [Autotuner]: Saved 0 configs to /workspace/emc/work/cache/vllm/flashinfer_autotune_cache/0.7.0.post1/120f/0881ec9c562a209723a6b6cf4fcfccee53df672972c395bfe03858f09816333d/autotune_configs.json (0 new, 0 from previous config)
144
+
145
+ Capturing CUDA graphs (PIECEWISE): 0%| | 0/7 [00:00<?, ?it/s]
146
+ Capturing CUDA graphs (PIECEWISE): 14%|█▍ | 1/7 [00:00<00:01, 5.57it/s]
147
+ Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 2/7 [00:00<00:00, 5.78it/s]
148
+ Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 3/7 [00:00<00:00, 5.59it/s]
149
+ Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 4/7 [00:00<00:00, 5.46it/s]
150
+ Capturing CUDA graphs (PIECEWISE): 86%|████████▌ | 6/7 [00:00<00:00, 7.34it/s]
151
+ Capturing CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:01<00:00, 6.97it/s]
152
+
153
+ Capturing CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
154
+ Capturing CUDA graphs (FULL): 67%|██████▋ | 2/3 [00:00<00:00, 14.95it/s]
155
+ Capturing CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 15.01it/s]
156
+ (EngineCore pid=3749) INFO 11-18 02:19:20 [speculator.py:144] Capturing model for speculator...
157
+
158
+ Capturing prefill CUDA graphs (PIECEWISE): 0%| | 0/7 [00:00<?, ?it/s]
159
+ Capturing prefill CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:00<00:00, 130.65it/s]
160
+
161
+ Capturing prefill CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
162
+ Capturing prefill CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 159.52it/s]
163
+
164
+ Capturing decode CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
165
+ Capturing decode CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 99.04it/s]
166
+ (EngineCore pid=3749) INFO 11-18 02:19:21 [model_runner.py:1087] Graph capturing finished in 3 secs, took 0.17 GiB
167
+ (EngineCore pid=3749) INFO 11-18 02:19:21 [gpu_worker.py:892] CUDA graph pool memory: 0.17 GiB (actual), 0.24 GiB (estimated), difference: 0.07 GiB (40.7%).
168
+ (EngineCore pid=3749) INFO 11-18 02:19:21 [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 51.37 GiB for consumed memory (weights + non-torch), 3.2 GiB for peak activation, and 0.17 GiB for CUDAGraph memory. Replace gpu_memory_utilization config with `--kv-cache-memory=32850334516` (30.59 GiB) to fit into requested memory, or `--kv-cache-memory=42462617600` (39.55 GiB) to fully utilize gpu memory. Current kv cache memory in use is 30.91 GiB.
169
+ (EngineCore pid=3749) INFO 11-18 02:19:22 [jit_monitor.py:84] Kernel JIT monitor activated; monitored JIT compilations during inference will use mode=warn.
170
+ (EngineCore pid=3749) INFO 11-18 02:19:23 [core.py:373] init engine (profile, create kv cache, warmup model) took 87.67 s (compilation: 33.79 s)
171
+ (EngineCore pid=3749) INFO 11-18 02:19:23 [kv_cache_utils.py:772] kv cache group sizes [16384, 16384, 16384, 800]
172
+ (EngineCore pid=3749) INFO 11-18 02:19:23 [kv_cache_utils.py:773] kv lcm block sizes 409600
173
+ (EngineCore pid=3749) INFO 11-18 02:19:23 [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'])
174
+ (APIServer pid=3632) INFO 11-18 02:19:23 [entry.py:132] Supported tasks: ['generate']
175
+ (APIServer pid=3632) INFO 11-18 02:19:23 [factories.py:76] Scale-out endpoints are disabled. Set --enable-scale-out to enable them.
176
+ (APIServer pid=3632) INFO 11-18 02:19:24 [hf.py:646] Detected the chat template content format to be 'openai'. You can set `--chat-template-content-format` to override this.
177
+ (APIServer pid=3632) WARNING 11-18 02:19:24 [serving.py:221] The chat template requires system-first ordering, so inline system messages in /v1/messages requests (e.g. host Code's per-turn reminders) are merged into the leading system prompt. Each new one changes the prompt prefix, so the rest of the conversation misses the prefix cache. Pass a --chat-template that accepts non-leading system messages to avoid this.
178
+ (APIServer pid=3632) WARNING 11-18 02:19:24 [serving.py:221] The chat template requires system-first ordering, so inline system messages in /v1/messages requests (e.g. host Code's per-turn reminders) are merged into the leading system prompt. Each new one changes the prompt prefix, so the rest of the conversation misses the prefix cache. Pass a --chat-template that accepts non-leading system messages to avoid this.
179
+ (APIServer pid=3632) INFO 11-18 02:19:24 [entry.py:136] Starting vLLM server on http://127.0.0.1:65351
180
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:60] Available routes are:
181
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /openapi.json, Methods: GET, HEAD
182
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /docs, Methods: GET, HEAD
183
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /docs/oauth2-redirect, Methods: GET, HEAD
184
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /redoc, Methods: GET, HEAD
185
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /load, Methods: GET
186
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /version, Methods: GET
187
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /kv_event_sources, Methods: GET
188
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /health, Methods: GET
189
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /metrics, Methods: GET
190
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /tokenize, Methods: POST
191
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /detokenize, Methods: POST
192
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/models, Methods: GET
193
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /ping, Methods: GET
194
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /ping, Methods: POST
195
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /invocations, Methods: POST
196
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/chat/completions, Methods: POST
197
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/chat/completions/batch, Methods: POST
198
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/responses, Methods: POST
199
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/responses/{response_id}, Methods: GET
200
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/responses/{response_id}/cancel, Methods: POST
201
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/completions, Methods: POST
202
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/messages, Methods: POST
203
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /v1/messages/count_tokens, Methods: POST
204
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /generative_scoring, Methods: POST
205
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /scale_elastic_ep, Methods: POST
206
+ (APIServer pid=3632) INFO 11-18 02:19:24 [launcher.py:69] Route: /is_scaling_elastic_ep, Methods: POST
207
+ INFO: Started server process [3632]
208
+ INFO: Waiting for application startup.
209
+ INFO: Application startup complete.
210
+ INFO: 127.0.0.1:45834 - "GET /health HTTP/1.1" 200 OK
211
+ INFO: 127.0.0.1:21361 - "POST /v1/completions HTTP/1.1" 200 OK
212
+ (EngineCore pid=3749) WARNING 11-18 02:19:26 [jit_monitor.py:140] Triton kernel JIT compilation during inference: _compute_local_logits_stats_kernel. This causes a latency spike; consider extending warmup to cover this shape/config.
213
+ (EngineCore pid=3749) WARNING 11-18 02:19:26 [jit_monitor.py:140] Triton kernel JIT compilation during inference: _rejection_kernel. This causes a latency spike; consider extending warmup to cover this shape/config.
214
+ (EngineCore pid=3749) WARNING 11-18 02:19:26 [jit_monitor.py:140] Triton kernel JIT compilation during inference: _resample_kernel. This causes a latency spike; consider extending warmup to cover this shape/config.
215
+ INFO: 127.0.0.1:37318 - "POST /v1/completions HTTP/1.1" 200 OK
216
+ (APIServer pid=3632) INFO 11-18 02:19:35 [loggers.py:320] Engine 000: Avg prompt throughput: 15.9 tokens/s, Avg generation throughput: 40.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
217
+ (APIServer pid=3632) INFO 11-18 02:19:35 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.13, Accepted throughput: 27.75 tokens/s, Drafted throughput: 39.06 tokens/s, Accepted: 324 tokens, Drafted: 456 tokens, Per-position acceptance rate: 0.816, 0.724, 0.592, Avg Draft acceptance rate: 71.1%
218
+ INFO: 127.0.0.1:22446 - "POST /v1/completions HTTP/1.1" 200 OK
219
+ INFO: 127.0.0.1:17355 - "POST /v1/completions HTTP/1.1" 200 OK
220
+ INFO: 127.0.0.1:26393 - "POST /v1/completions HTTP/1.1" 200 OK
221
+ (APIServer pid=3632) INFO 11-18 02:19:45 [loggers.py:320] Engine 000: Avg prompt throughput: 42.7 tokens/s, Avg generation throughput: 62.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
222
+ (APIServer pid=3632) INFO 11-18 02:19:45 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.55, Accepted throughput: 44.79 tokens/s, Drafted throughput: 52.79 tokens/s, Accepted: 448 tokens, Drafted: 528 tokens, Per-position acceptance rate: 0.943, 0.830, 0.773, Avg Draft acceptance rate: 84.8%
223
+ INFO: 127.0.0.1:49546 - "POST /v1/completions HTTP/1.1" 200 OK
224
+ INFO: 127.0.0.1:59863 - "POST /v1/completions HTTP/1.1" 200 OK
225
+ (APIServer pid=3632) INFO 11-18 02:19:55 [loggers.py:320] Engine 000: Avg prompt throughput: 14.0 tokens/s, Avg generation throughput: 52.6 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
226
+ (APIServer pid=3632) INFO 11-18 02:19:55 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.92, Accepted throughput: 34.60 tokens/s, Drafted throughput: 53.99 tokens/s, Accepted: 346 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.844, 0.628, 0.450, Avg Draft acceptance rate: 64.1%
227
+ (APIServer pid=3632) INFO 11-18 02:20:05 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 44.1 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
228
+ (APIServer pid=3632) INFO 11-18 02:20:05 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.41, Accepted throughput: 25.80 tokens/s, Drafted throughput: 54.89 tokens/s, Accepted: 258 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.716, 0.443, 0.251, Avg Draft acceptance rate: 47.0%
229
+ INFO: 127.0.0.1:18628 - "POST /v1/completions HTTP/1.1" 200 OK
230
+ (APIServer pid=3632) INFO 11-18 02:20:15 [loggers.py:320] Engine 000: Avg prompt throughput: 7.8 tokens/s, Avg generation throughput: 41.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
231
+ (APIServer pid=3632) INFO 11-18 02:20:15 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.29, Accepted throughput: 23.40 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 234 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.648, 0.396, 0.242, Avg Draft acceptance rate: 42.9%
232
+ (APIServer pid=3632) INFO 11-18 02:20:25 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 39.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
233
+ (APIServer pid=3632) INFO 11-18 02:20:25 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.13, Accepted throughput: 20.79 tokens/s, Drafted throughput: 55.19 tokens/s, Accepted: 208 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.625, 0.348, 0.158, Avg Draft acceptance rate: 37.7%
234
+ INFO: 127.0.0.1:25412 - "GET /metrics HTTP/1.1" 200 OK
235
+ INFO: 127.0.0.1:63783 - "POST /v1/completions HTTP/1.1" 200 OK
236
+ (APIServer pid=3632) INFO 11-18 02:20:35 [loggers.py:320] Engine 000: Avg prompt throughput: 7.2 tokens/s, Avg generation throughput: 41.8 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
237
+ (APIServer pid=3632) INFO 11-18 02:20:35 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.30, Accepted throughput: 23.60 tokens/s, Drafted throughput: 54.29 tokens/s, Accepted: 236 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.669, 0.398, 0.238, Avg Draft acceptance rate: 43.5%
238
+ INFO: 127.0.0.1:36562 - "POST /v1/completions HTTP/1.1" 200 OK
239
+ (APIServer pid=3632) INFO 11-18 02:20:45 [loggers.py:320] Engine 000: Avg prompt throughput: 3.9 tokens/s, Avg generation throughput: 50.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
240
+ (APIServer pid=3632) INFO 11-18 02:20:45 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.76, Accepted throughput: 32.00 tokens/s, Drafted throughput: 54.60 tokens/s, Accepted: 320 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.769, 0.588, 0.401, Avg Draft acceptance rate: 58.6%
241
+ INFO: 127.0.0.1:60466 - "GET /metrics HTTP/1.1" 200 OK
242
+ INFO: 127.0.0.1:40298 - "GET /metrics HTTP/1.1" 200 OK
243
+ INFO: 127.0.0.1:22141 - "POST /v1/completions HTTP/1.1" 200 OK
244
+ INFO: 127.0.0.1:37227 - "POST /v1/completions HTTP/1.1" 200 OK
245
+ (APIServer pid=3632) INFO 11-18 02:20:55 [loggers.py:320] Engine 000: Avg prompt throughput: 32.5 tokens/s, Avg generation throughput: 64.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
246
+ (APIServer pid=3632) INFO 11-18 02:20:55 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.56, Accepted throughput: 46.09 tokens/s, Drafted throughput: 53.99 tokens/s, Accepted: 461 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.944, 0.839, 0.778, Avg Draft acceptance rate: 85.4%
247
+ INFO: 127.0.0.1:29378 - "GET /metrics HTTP/1.1" 200 OK
248
+ INFO: 127.0.0.1:33520 - "GET /metrics HTTP/1.1" 200 OK
249
+ INFO: 127.0.0.1:64838 - "POST /v1/completions HTTP/1.1" 200 OK
250
+ (APIServer pid=3632) INFO 11-18 02:21:05 [loggers.py:320] Engine 000: Avg prompt throughput: 6.7 tokens/s, Avg generation throughput: 67.4 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
251
+ (APIServer pid=3632) INFO 11-18 02:21:05 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.70, Accepted throughput: 49.09 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 491 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.956, 0.901, 0.841, Avg Draft acceptance rate: 89.9%
252
+ INFO: 127.0.0.1:61760 - "POST /v1/completions HTTP/1.1" 200 OK
253
+ INFO: 127.0.0.1:61357 - "GET /metrics HTTP/1.1" 200 OK
254
+ INFO: 127.0.0.1:47543 - "GET /metrics HTTP/1.1" 200 OK
255
+ INFO: 127.0.0.1:26441 - "POST /v1/completions HTTP/1.1" 200 OK
256
+ (APIServer pid=3632) INFO 11-18 02:21:15 [loggers.py:320] Engine 000: Avg prompt throughput: 16.3 tokens/s, Avg generation throughput: 61.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
257
+ (APIServer pid=3632) INFO 11-18 02:21:15 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.40, Accepted throughput: 43.40 tokens/s, Drafted throughput: 54.30 tokens/s, Accepted: 434 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.928, 0.779, 0.691, Avg Draft acceptance rate: 79.9%
258
+ (APIServer pid=3632) INFO 11-18 02:21:25 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 46.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
259
+ (APIServer pid=3632) INFO 11-18 02:21:25 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.50, Accepted throughput: 27.59 tokens/s, Drafted throughput: 55.19 tokens/s, Accepted: 276 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.739, 0.473, 0.288, Avg Draft acceptance rate: 50.0%
260
+ INFO: 127.0.0.1:62694 - "POST /v1/completions HTTP/1.1" 200 OK
261
+ (APIServer pid=3632) INFO 11-18 02:21:35 [loggers.py:320] Engine 000: Avg prompt throughput: 3.8 tokens/s, Avg generation throughput: 43.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
262
+ (APIServer pid=3632) INFO 11-18 02:21:35 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.39, Accepted throughput: 25.10 tokens/s, Drafted throughput: 54.29 tokens/s, Accepted: 251 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.696, 0.448, 0.243, Avg Draft acceptance rate: 46.2%
263
+ (APIServer pid=3632) INFO 11-18 02:21:45 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 41.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
264
+ (APIServer pid=3632) INFO 11-18 02:21:45 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.23, Accepted throughput: 22.70 tokens/s, Drafted throughput: 55.50 tokens/s, Accepted: 227 tokens, Drafted: 555 tokens, Per-position acceptance rate: 0.665, 0.362, 0.200, Avg Draft acceptance rate: 40.9%
265
+ (APIServer pid=3632) INFO 11-18 02:21:55 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 45.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.4%
266
+ (APIServer pid=3632) INFO 11-18 02:21:55 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.50, Accepted throughput: 27.39 tokens/s, Drafted throughput: 54.89 tokens/s, Accepted: 274 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.738, 0.503, 0.257, Avg Draft acceptance rate: 49.9%
267
+ INFO: 127.0.0.1:43781 - "GET /metrics HTTP/1.1" 200 OK
268
+ (APIServer pid=3632) INFO 11-18 02:21:55 [launcher.py:128] [shutdown] API server: shutdown triggered
269
+ (APIServer pid=3632) INFO 11-18 02:21:55 [launcher.py:144] [shutdown] API server: stopping engine client mode=abort timeout=0s
270
+ (APIServer pid=3632) INFO 11-18 02:21:55 [core_client.py:762] [shutdown] MPClient: start timeout=0s
271
+ (APIServer pid=3632) INFO 11-18 02:21:55 [core_client.py:764] [shutdown] MPClient: stopping engine manager
272
+ (APIServer pid=3632) INFO 11-18 02:21:55 [utils.py:638] [shutdown] Process manager: send sigterm to process EngineCore
273
+ (APIServer pid=3632) WARNING 11-18 02:21:55 [utils.py:658] [shutdown] Process manager: force killing remaining processes count=1
274
+ (APIServer pid=3632) WARNING 11-18 02:21:55 [utils.py:663] [shutdown] Process manager: force killing remaining process EngineCore pid 3749
275
+ (APIServer pid=3632) INFO 11-18 02:21:55 [core_client.py:766] [shutdown] MPClient: engine manager stopped
276
+ (APIServer pid=3632) INFO 11-18 02:21:55 [core_client.py:767] [shutdown] MPClient: cleaning up background resources
277
+ (APIServer pid=3632) INFO 11-18 02:21:55 [core_client.py:769] [shutdown] MPClient: complete
278
+ (APIServer pid=3632) INFO 11-18 02:21:55 [launcher.py:153] [shutdown] API server: engine client stopped
279
+ (APIServer pid=3632) INFO 11-18 02:21:55 [launcher.py:156] [shutdown] API server: signalling HTTP server shutdown
280
+ (APIServer pid=3632) INFO 11-18 02:21:55 [launcher.py:178] [shutdown] API server: shutting down FastAPI HTTP server
281
+ INFO: Shutting down
282
+ INFO: Waiting for application shutdown.
283
+ INFO: Application shutdown complete.
284
+ /workspace/emc/uv-python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/multiprocessing/resource_tracker.py:279: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
285
+ warnings.warn('resource_tracker: There appear to be %d '
runs/r1/kit_results/20261118-022156-results.json ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "method": "mtp",
3
+ "prompts": 8,
4
+ "repeats": 1,
5
+ "acceptance_length": 2.867478025693036,
6
+ "acceptance_by_domain": {
7
+ "chat": 3.062874251497006,
8
+ "code": 3.7514450867052025,
9
+ "math": 3.6769230769230767,
10
+ "longform": 2.383526383526384
11
+ },
12
+ "tpot_median_s": 0.015302492555498957,
13
+ "ttft_median_s": 0.109490264499982,
14
+ "memory_bytes": 607177728,
15
+ "weight_bytes": 540068864,
16
+ "kv_cache_bytes": 67108864,
17
+ "decode_total_s": 81.19256583799984,
18
+ "num_speculative_tokens": 3,
19
+ "max_model_len": 16384,
20
+ "engine": {
21
+ "vllm_version": "0.31.0",
22
+ "vllm_version_pinned": "0.31.0",
23
+ "image": "pip:vllm==0.31.0",
24
+ "matches_pin": true
25
+ },
26
+ "per_prompt": [
27
+ {
28
+ "id": "chat-0000",
29
+ "domain": "chat",
30
+ "tokens": 512,
31
+ "ttft_s": 0.10654,
32
+ "total_s": 10.24148,
33
+ "tpot_s": 0.019834
34
+ },
35
+ {
36
+ "id": "chat-0001",
37
+ "domain": "chat",
38
+ "tokens": 512,
39
+ "ttft_s": 0.11244,
40
+ "total_s": 8.07832,
41
+ "tpot_s": 0.015589
42
+ },
43
+ {
44
+ "id": "code-0000",
45
+ "domain": "code",
46
+ "tokens": 139,
47
+ "ttft_s": 0.10167,
48
+ "total_s": 2.00082,
49
+ "tpot_s": 0.013762
50
+ },
51
+ {
52
+ "id": "code-0001",
53
+ "domain": "code",
54
+ "tokens": 512,
55
+ "ttft_s": 0.1213,
56
+ "total_s": 7.61486,
57
+ "tpot_s": 0.014665
58
+ },
59
+ {
60
+ "id": "math-0000",
61
+ "domain": "math",
62
+ "tokens": 384,
63
+ "ttft_s": 0.09996,
64
+ "total_s": 5.67863,
65
+ "tpot_s": 0.014566
66
+ },
67
+ {
68
+ "id": "math-0001",
69
+ "domain": "math",
70
+ "tokens": 333,
71
+ "ttft_s": 0.11644,
72
+ "total_s": 5.10181,
73
+ "tpot_s": 0.015016
74
+ },
75
+ {
76
+ "id": "longform-0000",
77
+ "domain": "longform",
78
+ "tokens": 943,
79
+ "ttft_s": 0.10042,
80
+ "total_s": 21.33993,
81
+ "tpot_s": 0.022547
82
+ },
83
+ {
84
+ "id": "longform-0001",
85
+ "domain": "longform",
86
+ "tokens": 907,
87
+ "ttft_s": 0.14278,
88
+ "total_s": 21.1367,
89
+ "tpot_s": 0.023172
90
+ }
91
+ ],
92
+ "score": {
93
+ "score": 110.58,
94
+ "ranked": true,
95
+ "s_acceptance": 1.0361,
96
+ "s_memory": 1.5,
97
+ "s_tpot": 0.8511
98
+ }
99
+ }
runs/r1/logs/pipeline.log CHANGED
@@ -66,3 +66,11 @@
66
  raise RuntimeError(
67
  RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}
68
  [2026-11-18 02:16:15] stopping: smoke_base is required
 
 
 
 
 
 
 
 
 
66
  raise RuntimeError(
67
  RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}
68
  [2026-11-18 02:16:15] stopping: smoke_base is required
69
+ [2026-11-18 02:16:40] pipeline start: run r1, code eb139866e50f
70
+ [2026-11-18 02:16:40] skip env (done)
71
+ [2026-11-18 02:16:40] skip download (done)
72
+ [2026-11-18 02:16:40] skip baseline (done)
73
+ [2026-11-18 02:16:40] skip checkenv (done)
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)
runs/r1/logs/pipeline.out CHANGED
@@ -66,3 +66,11 @@
66
  raise RuntimeError(
67
  RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}
68
  [2026-11-18 02:16:15] stopping: smoke_base is required
 
 
 
 
 
 
 
 
 
66
  raise RuntimeError(
67
  RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}
68
  [2026-11-18 02:16:15] stopping: smoke_base is required
69
+ [2026-11-18 02:16:40] pipeline start: run r1, code eb139866e50f
70
+ [2026-11-18 02:16:40] skip env (done)
71
+ [2026-11-18 02:16:40] skip download (done)
72
+ [2026-11-18 02:16:40] skip baseline (done)
73
+ [2026-11-18 02:16:40] skip checkenv (done)
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)
runs/r1/logs/smoke_fp8.log CHANGED
@@ -4,64 +4,29 @@
4
  KV cache @ full context: 0.067 GB
5
  memory metric: 0.607 GB
6
  PASS
7
- [02:08:45] repeat 1/1: fresh vLLM process
8
- [02:08:45] starting vLLM (log: /workspace/emc/work/logs/20261118-020845-eval-r1.log)
9
- Traceback (most recent call last):
10
- File "/workspace/emc/repo/scripts/kit.py", line 47, in <module>
11
- cli.main()
12
- File "/workspace/emc/kit/emc2/__main__.py", line 131, in main
13
- sys.exit(fn(args))
14
- ^^^^^^^^
15
- File "/workspace/emc/kit/emc2/__main__.py", line 78, in cmd_eval
16
- evaluate.run(c, t, meta, spec, info, args.limit_per_domain, args.repeats, args.write_baseline,
17
- File "/workspace/emc/kit/emc2/evaluate.py", line 80, in run
18
- with VllmServer(contract, target, spec=spec_cfg, name=f"eval-r{rep}") as srv:
19
- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
20
- File "/workspace/emc/kit/emc2/server.py", line 102, in __enter__
21
- self.start()
22
- File "/workspace/emc/kit/emc2/server.py", line 69, in start
23
- raise RuntimeError(f"vLLM exited with code {self.proc.returncode} during "
24
- RuntimeError: vLLM exited with code 1 during startup:
25
- ^^^^^^^^^^^^^^^^^^^^^^^^^^
26
- File "/workspace/emc/uv-python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/contextlib.py", line 210, in __aenter__
27
- return await anext(self.gen)
28
- ^^^^^^^^^^^^^^^^^^^^^
29
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/entrypoints/launchers/api_server/entry.py", line 58, in build_async_engine_client
30
- async with build_async_engine_client_from_engine_args(
31
- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
32
- File "/workspace/emc/uv-python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/contextlib.py", line 210, in __aenter__
33
- return await anext(self.gen)
34
- ^^^^^^^^^^^^^^^^^^^^^
35
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/entrypoints/launchers/api_server/entry.py", line 92, in build_async_engine_client_from_engine_args
36
- async_llm = AsyncLLM.from_vllm_config(
37
- ^^^^^^^^^^^^^^^^^^^^^^^^^^
38
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 248, in from_vllm_config
39
- return cls(
40
- ^^^^
41
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/v1/engine/async_llm.py", line 180, in __init__
42
- self.engine_core = EngineCoreClient.make_async_mp_client(
43
- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
44
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/tracing/otel.py", line 175, in sync_wrapper
45
- return func(*args, **kwargs)
46
- ^^^^^^^^^^^^^^^^^^^^^
47
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 167, in make_async_mp_client
48
- return AsyncMPClient(
49
- ^^^^^^^^^^^^^^
50
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/tracing/otel.py", line 175, in sync_wrapper
51
- return func(*args, **kwargs)
52
- ^^^^^^^^^^^^^^^^^^^^^
53
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 1100, in __init__
54
- super().__init__(
55
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/v1/engine/core_client.py", line 674, in __init__
56
- with launch_core_engines(
57
- ^^^^^^^^^^^^^^^^^^^^
58
- File "/workspace/emc/uv-python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/contextlib.py", line 144, in __exit__
59
- next(self.gen)
60
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 1268, in launch_core_engines
61
- wait_for_engine_startup(
62
- File "/workspace/emc/venv/lib/python3.12/site-packages/vllm/v1/engine/utils.py", line 1348, in wait_for_engine_startup
63
- raise RuntimeError(
64
- RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}
65
- Traceback (most recent call last):
66
- File "<stdin>", line 2, in <module>
67
- FileNotFoundError: [Errno 2] No such file or directory: '/workspace/emc/results/smoke_fp8.json'
 
4
  KV cache @ full context: 0.067 GB
5
  memory metric: 0.607 GB
6
  PASS
7
+ [02:16:49] repeat 1/1: fresh vLLM process
8
+ [02:16:49] starting vLLM (log: /workspace/emc/work/logs/20261118-021649-eval-r1.log)
9
+ [02:19:25] vLLM ready after 156s
10
+ [02:20:53] chat: 2 prompts done
11
+ [02:21:02] code: 2 prompts done
12
+ [02:21:13] math: 2 prompts done
13
+ [02:21:55] longform: 2 prompts done
14
+
15
+ ================ emc2 eval summary ================
16
+ engine vLLM 0.31.0
17
+ method mtp (K = 3, context 16384)
18
+ prompts x repeats 8 x 1
19
+ acceptance length 2.867
20
+ chat 3.063
21
+ code 3.751
22
+ math 3.677
23
+ longform 2.384
24
+ TPOT (median) 15.30 ms per output token (this GPU)
25
+ TTFT (median, info) 109.5 ms (this GPU; not scored)
26
+ memory metric 0.607 GB (weights 0.540 + KV 0.067)
27
+ score 110.6 (baseline = 100)
28
+ components acceptance 1.036, memory 1.500, tpot 0.851
29
+ full results /workspace/emc/work/results/20261118-022156-results.json
30
+ ===================================================
31
+
32
+ FP8 probe acceptance 2.867 vs published baseline 2.802 (ratio 1.036)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
runs/r1/logs/supervisor.log CHANGED
@@ -28,3 +28,4 @@ Installed 15 packages in 98ms
28
  [2026-11-18 02:03:36] pipeline started (pid 163) at 7eba65de078a
29
  [2026-11-18 02:15:39] code updated to eb139866e50f
30
  [2026-11-18 02:16:39] pipeline exited with code 1
 
 
28
  [2026-11-18 02:03:36] pipeline started (pid 163) at 7eba65de078a
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
runs/r1/logs/sync.log CHANGED
@@ -7,3 +7,4 @@ uploaded logs/results to mudassiruddin21/EMC:runs/r1
7
  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
 
 
7
  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
runs/r1/results/smoke_fp8.json ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "method": "mtp",
3
+ "prompts": 8,
4
+ "repeats": 1,
5
+ "acceptance_length": 2.867478025693036,
6
+ "acceptance_by_domain": {
7
+ "chat": 3.062874251497006,
8
+ "code": 3.7514450867052025,
9
+ "math": 3.6769230769230767,
10
+ "longform": 2.383526383526384
11
+ },
12
+ "tpot_median_s": 0.015302492555498957,
13
+ "ttft_median_s": 0.109490264499982,
14
+ "memory_bytes": 607177728,
15
+ "weight_bytes": 540068864,
16
+ "kv_cache_bytes": 67108864,
17
+ "decode_total_s": 81.19256583799984,
18
+ "num_speculative_tokens": 3,
19
+ "max_model_len": 16384,
20
+ "engine": {
21
+ "vllm_version": "0.31.0",
22
+ "vllm_version_pinned": "0.31.0",
23
+ "image": "pip:vllm==0.31.0",
24
+ "matches_pin": true
25
+ },
26
+ "per_prompt": [
27
+ {
28
+ "id": "chat-0000",
29
+ "domain": "chat",
30
+ "tokens": 512,
31
+ "ttft_s": 0.10654,
32
+ "total_s": 10.24148,
33
+ "tpot_s": 0.019834
34
+ },
35
+ {
36
+ "id": "chat-0001",
37
+ "domain": "chat",
38
+ "tokens": 512,
39
+ "ttft_s": 0.11244,
40
+ "total_s": 8.07832,
41
+ "tpot_s": 0.015589
42
+ },
43
+ {
44
+ "id": "code-0000",
45
+ "domain": "code",
46
+ "tokens": 139,
47
+ "ttft_s": 0.10167,
48
+ "total_s": 2.00082,
49
+ "tpot_s": 0.013762
50
+ },
51
+ {
52
+ "id": "code-0001",
53
+ "domain": "code",
54
+ "tokens": 512,
55
+ "ttft_s": 0.1213,
56
+ "total_s": 7.61486,
57
+ "tpot_s": 0.014665
58
+ },
59
+ {
60
+ "id": "math-0000",
61
+ "domain": "math",
62
+ "tokens": 384,
63
+ "ttft_s": 0.09996,
64
+ "total_s": 5.67863,
65
+ "tpot_s": 0.014566
66
+ },
67
+ {
68
+ "id": "math-0001",
69
+ "domain": "math",
70
+ "tokens": 333,
71
+ "ttft_s": 0.11644,
72
+ "total_s": 5.10181,
73
+ "tpot_s": 0.015016
74
+ },
75
+ {
76
+ "id": "longform-0000",
77
+ "domain": "longform",
78
+ "tokens": 943,
79
+ "ttft_s": 0.10042,
80
+ "total_s": 21.33993,
81
+ "tpot_s": 0.022547
82
+ },
83
+ {
84
+ "id": "longform-0001",
85
+ "domain": "longform",
86
+ "tokens": 907,
87
+ "ttft_s": 0.14278,
88
+ "total_s": 21.1367,
89
+ "tpot_s": 0.023172
90
+ }
91
+ ],
92
+ "score": {
93
+ "score": 110.58,
94
+ "ranked": true,
95
+ "s_acceptance": 1.0361,
96
+ "s_memory": 1.5,
97
+ "s_tpot": 0.8511
98
+ }
99
+ }
runs/r1/state/fp8_ok ADDED
@@ -0,0 +1 @@
 
 
1
+ 1.0361
runs/r1/state/pipeline_commit CHANGED
@@ -1 +1 @@
1
- 7eba65de078a94e6c1068bf645bc92b107c44200
 
1
+ eb139866e50f49213c820e26516acb5588897e87
runs/r1/state/smoke_fp8.done ADDED
@@ -0,0 +1 @@
 
 
1
+ 317