mudassiruddin21 commited on
Commit
ddbbf7f
·
verified ·
1 Parent(s): a3dae7f

Run r1: logs and results

Browse files
runs/r1/engine_logs/20261118-031805-eval-r1.log ADDED
@@ -0,0 +1,363 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ vllm serve /workspace/emc/models/target --served-model-name target --host 127.0.0.1 --port 50451 --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/runs/main/sub_fp8", "num_speculative_tokens": 3}
2
+
3
+ (APIServer pid=11045) INFO 11-18 03:18:23 [api_utils.py:347]
4
+ (APIServer pid=11045) INFO 11-18 03:18:23 [api_utils.py:347] █ █ █▄ ▄█
5
+ (APIServer pid=11045) INFO 11-18 03:18:23 [api_utils.py:347] ▄▄ ▄█ █ █ █ ▀▄▀ █ version 0.31.0
6
+ (APIServer pid=11045) INFO 11-18 03:18:23 [api_utils.py:347] █▄█▀ █ █ █ █ model /workspace/emc/models/target
7
+ (APIServer pid=11045) INFO 11-18 03:18:23 [api_utils.py:347] ▀▀ ▀▀▀▀▀ ▀▀▀▀▀ ▀ ▀
8
+ (APIServer pid=11045) INFO 11-18 03:18:23 [api_utils.py:347]
9
+ (APIServer pid=11045) INFO 11-18 03:18:23 [api_utils.py:286] non-default args: {'model_tag': '/workspace/emc/models/target', 'host': '127.0.0.1', 'port': 50451, '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/runs/main/sub_fp8', 'num_speculative_tokens': 3}}
10
+ (APIServer pid=11045) INFO 11-18 03:18:23 [model.py:699] Resolved architecture: Qwen3_5ForConditionalGeneration
11
+ (APIServer pid=11045) INFO 11-18 03:18:23 [model.py:2109] Using max model len 16384
12
+ (APIServer pid=11045) INFO 11-18 03:18:25 [model.py:917] All limits of multimodal modalities supported by the model are set to 0, running in text-only mode.
13
+ (APIServer pid=11045) INFO 11-18 03:18:25 [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=11045) INFO 11-18 03:18:25 [model.py:699] Resolved architecture: Qwen3_5MTP
15
+ (APIServer pid=11045) INFO 11-18 03:18:25 [model.py:2109] Using max model len 262144
16
+ (APIServer pid=11045) WARNING 11-18 03:18:25 [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=11045) INFO 11-18 03:18:25 [speculative.py:1663] Overriding draft model max model len from 262144 to 16384
18
+ (APIServer pid=11045) INFO 11-18 03:18:25 [scheduler.py:308] Chunked prefill is enabled with max_num_batched_tokens=8192.
19
+ (APIServer pid=11045) INFO 11-18 03:18:25 [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=11144) INFO 11-18 03:18:44 [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/runs/main/sub_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=11144) INFO 11-18 03:18:45 [parallel_state.py:1825] world_size=1 rank=0 local_rank=0 distributed_init_method=file:///workspace/emc/tmp/vllm_dist_d5d4a6608e7d4a4f9adf60e465f65006 backend=nccl
22
+ (EngineCore pid=11144) INFO 11-18 03:18:45 [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=11144) INFO 11-18 03:18:45 [gpu_worker.py:479] Using V2 Model Runner
24
+ (EngineCore pid=11144) INFO 11-18 03:18:47 [model_runner.py:375] Loading model from scratch...
25
+ (EngineCore pid=11144) INFO 11-18 03:18:47 [cuda.py:585] Using backend AttentionBackendEnum.FLASH_ATTN for vit attention
26
+ (EngineCore pid=11144) INFO 11-18 03:18:47 [mm_encoder_attention.py:372] Using AttentionBackendEnum.FLASH_ATTN for MMEncoderAttention.
27
+ (EngineCore pid=11144) INFO 11-18 03:18:47 [qwen_gdn_linear_attn.py:177] Using FlashInfer GDN prefill kernel (requested=auto, head_k_dim=128).
28
+ (EngineCore pid=11144) INFO 11-18 03:18:47 [qwen_gdn_linear_attn.py:533] GDN decode kernel: cuda
29
+ (EngineCore pid=11144) INFO 11-18 03:18:50 [cuda.py:526] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
30
+ (EngineCore pid=11144) INFO 11-18 03:18:50 [flash_attn.py:1167] Using FlashAttention version 2
31
+ (EngineCore pid=11144) INFO 11-18 03:18:51 [weight_utils.py:1063] Filesystem type for checkpoints: OVERLAY. Checkpoint size: 51.75 GiB. Available RAM: 156.12 GiB.
32
+ (EngineCore pid=11144) INFO 11-18 03:18:51 [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<00:07, 2.29it/s]
37
+
38
+ Loading safetensors checkpoint shards: 11% Completed | 2/18 [00:00<00:07, 2.28it/s]
39
+
40
+ Loading safetensors checkpoint shards: 17% Completed | 3/18 [00:01<00:05, 2.52it/s]
41
+
42
+ Loading safetensors checkpoint shards: 22% Completed | 4/18 [00:01<00:06, 2.14it/s]
43
+
44
+ Loading safetensors checkpoint shards: 28% Completed | 5/18 [00:02<00:05, 2.49it/s]
45
+
46
+ Loading safetensors checkpoint shards: 33% Completed | 6/18 [00:02<00:05, 2.22it/s]
47
+
48
+ Loading safetensors checkpoint shards: 39% Completed | 7/18 [00:02<00:04, 2.56it/s]
49
+
50
+ Loading safetensors checkpoint shards: 44% Completed | 8/18 [00:03<00:04, 2.27it/s]
51
+
52
+ Loading safetensors checkpoint shards: 50% Completed | 9/18 [00:03<00:03, 2.55it/s]
53
+
54
+ Loading safetensors checkpoint shards: 56% Completed | 10/18 [00:04<00:03, 2.19it/s]
55
+
56
+ Loading safetensors checkpoint shards: 61% Completed | 11/18 [00:04<00:02, 2.47it/s]
57
+
58
+ Loading safetensors checkpoint shards: 67% Completed | 12/18 [00:05<00:02, 2.16it/s]
59
+
60
+ Loading safetensors checkpoint shards: 72% Completed | 13/18 [00:05<00:02, 2.47it/s]
61
+
62
+ Loading safetensors checkpoint shards: 78% Completed | 14/18 [00:05<00:01, 2.31it/s]
63
+
64
+ Loading safetensors checkpoint shards: 83% Completed | 15/18 [00:06<00:01, 2.56it/s]
65
+
66
+ Loading safetensors checkpoint shards: 89% Completed | 16/18 [00:06<00:00, 2.29it/s]
67
+
68
+ Loading safetensors checkpoint shards: 94% Completed | 17/18 [00:07<00:00, 2.60it/s]
69
+
70
+ Loading safetensors checkpoint shards: 100% Completed | 18/18 [00:07<00:00, 2.67it/s]
71
+
72
+ Loading safetensors checkpoint shards: 100% Completed | 18/18 [00:07<00:00, 2.42it/s]
73
+
74
+ (EngineCore pid=11144) INFO 11-18 03:18:58 [default_loader.py:484] Loading weights took 7.45 seconds
75
+ (EngineCore pid=11144) INFO 11-18 03:18:58 [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, 3.50it/s]
80
+
81
+ Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00, 3.50it/s]
82
+
83
+ (EngineCore pid=11144) INFO 11-18 03:18:58 [default_loader.py:484] Loading weights took 0.29 seconds
84
+ (EngineCore pid=11144) INFO 11-18 03:18:59 [model_runner.py:407] Model loading took 51.02 GiB memory and 12.647580 seconds
85
+ (EngineCore pid=11144) INFO 11-18 03:18:59 [topk_topp_sampler.py:96] FlashInfer top-p/top-k sampling disabled via VLLM_USE_FLASHINFER_SAMPLER=0.
86
+ (EngineCore pid=11144) WARNING 11-18 03:18:59 [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=11144) INFO 11-18 03:18:59 [interface.py:989] Setting attention block size to 800 tokens to ensure that attention page size is >= mamba page size.
88
+ (EngineCore pid=11144) INFO 11-18 03:18:59 [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=11144) INFO 11-18 03:18:59 [utils.py:343] Using LBNHC KV cache layout.
90
+ (EngineCore pid=11144) INFO 11-18 03:19:07 [backends.py:1090] Using cache directory: /workspace/emc/work/cache/vllm/torch_compile_cache/80f2acb952/rank_0_0/backbone for vLLM's torch.compile
91
+ (EngineCore pid=11144) INFO 11-18 03:19:07 [backends.py:1151] Dynamo bytecode transform time: 7.19 s
92
+ (EngineCore pid=11144) INFO 11-18 03:19:21 [backends.py:391] Compiling a graph for compile range (1, 8192) takes 13.88 s
93
+ (EngineCore pid=11144) INFO 11-18 03:19:25 [backends.py:916] collected artifacts: 65 entries, 20 artifacts, 24813950 bytes total
94
+ (EngineCore pid=11144) INFO 11-18 03:19:25 [decorators.py:717] saved AOT compiled function to /workspace/emc/work/cache/vllm/torch_compile_cache/torch_aot_compile/e2165c0e08ff485af35a252d7feae75280342c2fb8b8fc69d15470fbd7b70d13/rank_0_0/model
95
+ (EngineCore pid=11144) INFO 11-18 03:19:25 [monitor.py:53] torch.compile took 25.01 s in total
96
+ (EngineCore pid=11144) INFO 11-18 03:19:38 [monitor.py:81] Initial profiling/warmup run took 13.33 s
97
+ (EngineCore pid=11144) INFO 11-18 03:19:39 [backends.py:1090] Using cache directory: /workspace/emc/work/cache/vllm/torch_compile_cache/80f2acb952/rank_0_0/eagle_head for vLLM's torch.compile
98
+ (EngineCore pid=11144) INFO 11-18 03:19:39 [backends.py:1151] Dynamo bytecode transform time: 0.39 s
99
+ (EngineCore pid=11144) INFO 11-18 03:19:41 [backends.py:391] Compiling a graph for compile range (1, 8192) takes 1.90 s
100
+ (EngineCore pid=11144) INFO 11-18 03:19:41 [backends.py:916] collected artifacts: 2 entries, 2 artifacts, 3535270 bytes total
101
+ (EngineCore pid=11144) INFO 11-18 03:19:41 [decorators.py:717] saved AOT compiled function to /workspace/emc/work/cache/vllm/torch_compile_cache/torch_aot_compile/5c8773d6c0e8425afe6ef3bcf22f3b10d2523541ac3a4d11c24870f570fa5d00/rank_0_0/model
102
+ (EngineCore pid=11144) INFO 11-18 03:19:41 [monitor.py:53] torch.compile took 2.48 s in total
103
+ (EngineCore pid=11144) INFO 11-18 03:19:41 [monitor.py:81] Initial profiling/warmup run took 0.00 s
104
+ (EngineCore pid=11144) WARNING 11-18 03:19:42 [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 [00:06<00:38, 6.40s/it]
108
+ Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 2/7 [00:06<00:13, 2.72s/it]
109
+ Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 3/7 [00:06<00:06, 1.54s/it]
110
+ Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 4/7 [00:06<00:02, 1.01it/s]
111
+ Capturing CUDA graphs (PIECEWISE): 71%|███████▏ | 5/7 [00:07<00:01, 1.43it/s]
112
+ Capturing CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:07<00:00, 2.54it/s]
113
+ Capturing CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:07<00:00, 1.03s/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, 13.83it/s]
117
+ Capturing CUDA graphs (FULL): 100%|██████████| 2/2 [00:00<00:00, 13.81it/s]
118
+ (EngineCore pid=11144) INFO 11-18 03:19:50 [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, 146.98it/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, 173.13it/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, 97.14it/s]
128
+ (EngineCore pid=11144) INFO 11-18 03:19:50 [model_runner.py:1087] Graph capturing finished in 8 secs, took 0.23 GiB
129
+ (EngineCore pid=11144) INFO 11-18 03:19:51 [gpu_worker.py:692] Available KV cache memory: 30.91 GiB
130
+ (EngineCore pid=11144) INFO 11-18 03:19:51 [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=11144) WARNING 11-18 03:19:51 [kv_cache_utils.py:1574] Add 3 padding layers, may waste at most 6.25% KV cache memory
132
+ (EngineCore pid=11144) INFO 11-18 03:19:51 [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=11144) INFO 11-18 03:19:51 [kernel_warmup.py:172] JIT kernel warmup starting.
134
+ (EngineCore pid=11144) INFO 11-18 03:19:51 [kernel_warmup.py:185] JIT kernel warmup finished in 0.00s.
135
+ (EngineCore pid=11144) INFO 11-18 03:19:51 [qwen_triton_warmup.py:294] Warming up Qwen GDN Triton kernels for model_type=qwen3_5_text.
136
+ (EngineCore pid=11144) INFO 11-18 03:19:51 [qwen_vl_triton_warmup.py:98] Warmed M-RoPE Triton kernels.
137
+ (EngineCore pid=11144) INFO 11-18 03:19:51 [mamba_triton_warmup.py:42] Warmed Mamba batch_memcpy_kernel.
138
+ (EngineCore pid=11144) INFO 11-18 03:19:52 [kernel_warmup.py:437] Using FlashInfer autotune cache file: /workspace/emc/work/cache/vllm/flashinfer_autotune_cache/0.7.0.post1/120f/9fe4bf85285abb476325874844e98a94f5cac3974a483468ea944541322c9545/autotune_configs.json
139
+ 2026-11-18 03:19:52,514 - INFO - autotuner.py:1046 - flashinfer.jit: [Autotuner]: Autotuning process starts ...
140
+ (EngineCore pid=11144) INFO 11-18 03:19:52 [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).
141
+ 2026-11-18 03:19:52,539 - INFO - autotuner.py:1069 - flashinfer.jit: [Autotuner]: Autotuning process ends
142
+ 2026-11-18 03:19:52,708 - INFO - autotuner.py:3554 - flashinfer.jit: [Autotuner]: Saved 0 configs to /workspace/emc/work/cache/vllm/flashinfer_autotune_cache/0.7.0.post1/120f/9fe4bf85285abb476325874844e98a94f5cac3974a483468ea944541322c9545/autotune_configs.json (0 new, 0 from previous config)
143
+
144
+ Capturing CUDA graphs (PIECEWISE): 0%| | 0/7 [00:00<?, ?it/s]
145
+ Capturing CUDA graphs (PIECEWISE): 14%|█▍ | 1/7 [00:00<00:00, 6.74it/s]
146
+ Capturing CUDA graphs (PIECEWISE): 29%|██▊ | 2/7 [00:00<00:00, 6.78it/s]
147
+ Capturing CUDA graphs (PIECEWISE): 43%|████▎ | 3/7 [00:00<00:00, 6.80it/s]
148
+ Capturing CUDA graphs (PIECEWISE): 57%|█████▋ | 4/7 [00:00<00:00, 6.88it/s]
149
+ Capturing CUDA graphs (PIECEWISE): 86%|████████▌ | 6/7 [00:00<00:00, 8.56it/s]
150
+ Capturing CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:00<00:00, 8.19it/s]
151
+
152
+ Capturing CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
153
+ Capturing CUDA graphs (FULL): 67%|██████▋ | 2/3 [00:00<00:00, 15.40it/s]
154
+ Capturing CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 15.26it/s]
155
+ (EngineCore pid=11144) INFO 11-18 03:19:55 [speculator.py:144] Capturing model for speculator...
156
+
157
+ Capturing prefill CUDA graphs (PIECEWISE): 0%| | 0/7 [00:00<?, ?it/s]
158
+ Capturing prefill CUDA graphs (PIECEWISE): 100%|██████████| 7/7 [00:00<00:00, 149.58it/s]
159
+
160
+ Capturing prefill CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
161
+ Capturing prefill CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 172.73it/s]
162
+
163
+ Capturing decode CUDA graphs (FULL): 0%| | 0/3 [00:00<?, ?it/s]
164
+ Capturing decode CUDA graphs (FULL): 100%|██████████| 3/3 [00:00<00:00, 110.00it/s]
165
+ (EngineCore pid=11144) INFO 11-18 03:19:56 [model_runner.py:1087] Graph capturing finished in 2 secs, took 0.17 GiB
166
+ (EngineCore pid=11144) INFO 11-18 03:19:56 [gpu_worker.py:892] CUDA graph pool memory: 0.17 GiB (actual), 0.24 GiB (estimated), difference: 0.07 GiB (40.7%).
167
+ (EngineCore pid=11144) INFO 11-18 03:19:56 [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.
168
+ (EngineCore pid=11144) INFO 11-18 03:19:57 [jit_monitor.py:84] Kernel JIT monitor activated; monitored JIT compilations during inference will use mode=warn.
169
+ (EngineCore pid=11144) INFO 11-18 03:19:58 [core.py:373] init engine (profile, create kv cache, warmup model) took 58.43 s (compilation: 27.49 s)
170
+ (EngineCore pid=11144) INFO 11-18 03:19:58 [kv_cache_utils.py:772] kv cache group sizes [16384, 16384, 16384, 800]
171
+ (EngineCore pid=11144) INFO 11-18 03:19:58 [kv_cache_utils.py:773] kv lcm block sizes 409600
172
+ (EngineCore pid=11144) INFO 11-18 03:19:58 [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'])
173
+ (APIServer pid=11045) INFO 11-18 03:19:58 [entry.py:132] Supported tasks: ['generate']
174
+ (APIServer pid=11045) INFO 11-18 03:19:58 [factories.py:76] Scale-out endpoints are disabled. Set --enable-scale-out to enable them.
175
+ (APIServer pid=11045) INFO 11-18 03:19:59 [hf.py:646] Detected the chat template content format to be 'openai'. You can set `--chat-template-content-format` to override this.
176
+ (APIServer pid=11045) WARNING 11-18 03:19:59 [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.
177
+ (APIServer pid=11045) WARNING 11-18 03:19:59 [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=11045) INFO 11-18 03:19:59 [entry.py:136] Starting vLLM server on http://127.0.0.1:50451
179
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:60] Available routes are:
180
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /openapi.json, Methods: GET, HEAD
181
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /docs, Methods: GET, HEAD
182
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /docs/oauth2-redirect, Methods: GET, HEAD
183
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /redoc, Methods: GET, HEAD
184
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /load, Methods: GET
185
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /version, Methods: GET
186
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /kv_event_sources, Methods: GET
187
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /health, Methods: GET
188
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /metrics, Methods: GET
189
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /tokenize, Methods: POST
190
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /detokenize, Methods: POST
191
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/models, Methods: GET
192
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /ping, Methods: GET
193
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /ping, Methods: POST
194
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /invocations, Methods: POST
195
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/chat/completions, Methods: POST
196
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/chat/completions/batch, Methods: POST
197
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/responses, Methods: POST
198
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/responses/{response_id}, Methods: GET
199
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/responses/{response_id}/cancel, Methods: POST
200
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/completions, Methods: POST
201
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/messages, Methods: POST
202
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /v1/messages/count_tokens, Methods: POST
203
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /generative_scoring, Methods: POST
204
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /scale_elastic_ep, Methods: POST
205
+ (APIServer pid=11045) INFO 11-18 03:19:59 [launcher.py:69] Route: /is_scaling_elastic_ep, Methods: POST
206
+ INFO: Started server process [11045]
207
+ INFO: Waiting for application startup.
208
+ INFO: Application startup complete.
209
+ INFO: 127.0.0.1:50358 - "GET /health HTTP/1.1" 200 OK
210
+ INFO: 127.0.0.1:31858 - "POST /v1/completions HTTP/1.1" 200 OK
211
+ (EngineCore pid=11144) WARNING 11-18 03:20:01 [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.
212
+ (EngineCore pid=11144) WARNING 11-18 03:20:01 [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.
213
+ (EngineCore pid=11144) WARNING 11-18 03:20:01 [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.
214
+ INFO: 127.0.0.1:27433 - "POST /v1/completions HTTP/1.1" 200 OK
215
+ (APIServer pid=11045) INFO 11-18 03:20:09 [loggers.py:320] Engine 000: Avg prompt throughput: 16.4 tokens/s, Avg generation throughput: 41.4 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
216
+ (APIServer pid=11045) INFO 11-18 03:20:09 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.22, Accepted throughput: 28.62 tokens/s, Drafted throughput: 38.69 tokens/s, Accepted: 324 tokens, Drafted: 438 tokens, Per-position acceptance rate: 0.877, 0.753, 0.589, Avg Draft acceptance rate: 74.0%
217
+ INFO: 127.0.0.1:18361 - "POST /v1/completions HTTP/1.1" 200 OK
218
+ INFO: 127.0.0.1:47558 - "POST /v1/completions HTTP/1.1" 200 OK
219
+ INFO: 127.0.0.1:25218 - "POST /v1/completions HTTP/1.1" 200 OK
220
+ (APIServer pid=11045) INFO 11-18 03:20:19 [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%
221
+ (APIServer pid=11045) INFO 11-18 03:20:19 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.49, Accepted throughput: 44.30 tokens/s, Drafted throughput: 53.39 tokens/s, Accepted: 443 tokens, Drafted: 534 tokens, Per-position acceptance rate: 0.904, 0.831, 0.753, Avg Draft acceptance rate: 83.0%
222
+ INFO: 127.0.0.1:23904 - "POST /v1/completions HTTP/1.1" 200 OK
223
+ INFO: 127.0.0.1:60397 - "POST /v1/completions HTTP/1.1" 200 OK
224
+ (APIServer pid=11045) INFO 11-18 03:20:29 [loggers.py:320] Engine 000: Avg prompt throughput: 14.0 tokens/s, Avg generation throughput: 51.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
225
+ (APIServer pid=11045) INFO 11-18 03:20:29 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.89, Accepted throughput: 33.80 tokens/s, Drafted throughput: 53.69 tokens/s, Accepted: 338 tokens, Drafted: 537 tokens, Per-position acceptance rate: 0.855, 0.592, 0.441, Avg Draft acceptance rate: 62.9%
226
+ (APIServer pid=11045) INFO 11-18 03:20:39 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 47.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
227
+ (APIServer pid=11045) INFO 11-18 03:20:39 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.55, Accepted throughput: 28.60 tokens/s, Drafted throughput: 55.20 tokens/s, Accepted: 286 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.772, 0.478, 0.304, Avg Draft acceptance rate: 51.8%
228
+ INFO: 127.0.0.1:34858 - "POST /v1/completions HTTP/1.1" 200 OK
229
+ (APIServer pid=11045) INFO 11-18 03:20:49 [loggers.py:320] Engine 000: Avg prompt throughput: 7.8 tokens/s, Avg generation throughput: 42.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
230
+ (APIServer pid=11045) INFO 11-18 03:20:49 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.32, Accepted throughput: 24.10 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 241 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.681, 0.385, 0.258, Avg Draft acceptance rate: 44.1%
231
+ (APIServer pid=11045) INFO 11-18 03:20:59 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 41.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
232
+ (APIServer pid=11045) INFO 11-18 03:20:59 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.26, Accepted throughput: 23.10 tokens/s, Drafted throughput: 55.19 tokens/s, Accepted: 231 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.647, 0.380, 0.228, Avg Draft acceptance rate: 41.8%
233
+ (APIServer pid=11045) INFO 11-18 03:21:09 [loggers.py:320] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 45.8 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.4%
234
+ (APIServer pid=11045) INFO 11-18 03:21:09 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.49, Accepted throughput: 27.40 tokens/s, Drafted throughput: 55.19 tokens/s, Accepted: 274 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.734, 0.462, 0.293, Avg Draft acceptance rate: 49.6%
235
+ INFO: 127.0.0.1:42084 - "GET /metrics HTTP/1.1" 200 OK
236
+ INFO: 127.0.0.1:44766 - "POST /v1/completions HTTP/1.1" 200 OK
237
+ (APIServer pid=11045) INFO 11-18 03:21:19 [loggers.py:320] Engine 000: Avg prompt throughput: 7.2 tokens/s, Avg generation throughput: 48.1 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
238
+ (APIServer pid=11045) INFO 11-18 03:21:19 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.64, Accepted throughput: 29.80 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 298 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.780, 0.527, 0.330, Avg Draft acceptance rate: 54.6%
239
+ INFO: 127.0.0.1:26872 - "POST /v1/completions HTTP/1.1" 200 OK
240
+ (APIServer pid=11045) INFO 11-18 03:21:29 [loggers.py:320] Engine 000: Avg prompt throughput: 3.9 tokens/s, Avg generation throughput: 60.9 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
241
+ (APIServer pid=11045) INFO 11-18 03:21:29 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.34, Accepted throughput: 42.59 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 426 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.879, 0.780, 0.681, Avg Draft acceptance rate: 78.0%
242
+ INFO: 127.0.0.1:18881 - "POST /v1/completions HTTP/1.1" 200 OK
243
+ INFO: 127.0.0.1:35407 - "POST /v1/completions HTTP/1.1" 200 OK
244
+ (APIServer pid=11045) INFO 11-18 03:21:39 [loggers.py:320] Engine 000: Avg prompt throughput: 11.9 tokens/s, Avg generation throughput: 62.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
245
+ (APIServer pid=11045) INFO 11-18 03:21:39 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.44, Accepted throughput: 44.40 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 444 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.934, 0.802, 0.703, Avg Draft acceptance rate: 81.3%
246
+ INFO: 127.0.0.1:28967 - "POST /v1/completions HTTP/1.1" 200 OK
247
+ (APIServer pid=11045) INFO 11-18 03:21:49 [loggers.py:320] Engine 000: Avg prompt throughput: 10.3 tokens/s, Avg generation throughput: 58.9 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
248
+ (APIServer pid=11045) INFO 11-18 03:21:49 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.23, Accepted throughput: 40.79 tokens/s, Drafted throughput: 54.89 tokens/s, Accepted: 408 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.896, 0.732, 0.601, Avg Draft acceptance rate: 74.3%
249
+ INFO: 127.0.0.1:43479 - "POST /v1/completions HTTP/1.1" 200 OK
250
+ (APIServer pid=11045) INFO 11-18 03:21:59 [loggers.py:320] Engine 000: Avg prompt throughput: 6.4 tokens/s, Avg generation throughput: 53.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
251
+ (APIServer pid=11045) INFO 11-18 03:21:59 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.93, Accepted throughput: 35.30 tokens/s, Drafted throughput: 54.89 tokens/s, Accepted: 353 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.820, 0.639, 0.470, Avg Draft acceptance rate: 64.3%
252
+ INFO: 127.0.0.1:54542 - "POST /v1/completions HTTP/1.1" 200 OK
253
+ (APIServer pid=11045) INFO 11-18 03:22:09 [loggers.py:320] Engine 000: Avg prompt throughput: 2.6 tokens/s, Avg generation throughput: 53.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
254
+ (APIServer pid=11045) INFO 11-18 03:22:09 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.93, Accepted throughput: 35.40 tokens/s, Drafted throughput: 54.90 tokens/s, Accepted: 354 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.852, 0.650, 0.432, Avg Draft acceptance rate: 64.5%
255
+ INFO: 127.0.0.1:19182 - "POST /v1/completions HTTP/1.1" 200 OK
256
+ INFO: 127.0.0.1:16681 - "POST /v1/completions HTTP/1.1" 200 OK
257
+ (APIServer pid=11045) INFO 11-18 03:22:19 [loggers.py:320] Engine 000: Avg prompt throughput: 26.7 tokens/s, Avg generation throughput: 53.1 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
258
+ (APIServer pid=11045) INFO 11-18 03:22:19 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.96, Accepted throughput: 35.30 tokens/s, Drafted throughput: 53.99 tokens/s, Accepted: 353 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.878, 0.639, 0.444, Avg Draft acceptance rate: 65.4%
259
+ INFO: 127.0.0.1:45535 - "POST /v1/completions HTTP/1.1" 200 OK
260
+ (APIServer pid=11045) INFO 11-18 03:22:29 [loggers.py:320] Engine 000: Avg prompt throughput: 5.4 tokens/s, Avg generation throughput: 55.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
261
+ (APIServer pid=11045) INFO 11-18 03:22:29 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.01, Accepted throughput: 36.70 tokens/s, Drafted throughput: 54.89 tokens/s, Accepted: 367 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.874, 0.645, 0.486, Avg Draft acceptance rate: 66.8%
262
+ INFO: 127.0.0.1:18906 - "POST /v1/completions HTTP/1.1" 200 OK
263
+ INFO: 127.0.0.1:63462 - "POST /v1/completions HTTP/1.1" 200 OK
264
+ (APIServer pid=11045) INFO 11-18 03:22:39 [loggers.py:320] Engine 000: Avg prompt throughput: 26.5 tokens/s, Avg generation throughput: 60.6 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
265
+ (APIServer pid=11045) INFO 11-18 03:22:39 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.36, Accepted throughput: 42.69 tokens/s, Drafted throughput: 54.29 tokens/s, Accepted: 427 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.923, 0.796, 0.641, Avg Draft acceptance rate: 78.6%
266
+ INFO: 127.0.0.1:31714 - "POST /v1/completions HTTP/1.1" 200 OK
267
+ (APIServer pid=11045) INFO 11-18 03:22:49 [loggers.py:320] Engine 000: Avg prompt throughput: 3.4 tokens/s, Avg generation throughput: 57.6 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
268
+ (APIServer pid=11045) INFO 11-18 03:22:49 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.16, Accepted throughput: 39.39 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 394 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.912, 0.709, 0.544, Avg Draft acceptance rate: 72.2%
269
+ INFO: 127.0.0.1:49064 - "POST /v1/completions HTTP/1.1" 200 OK
270
+ (APIServer pid=11045) INFO 11-18 03:22:59 [loggers.py:320] Engine 000: Avg prompt throughput: 4.4 tokens/s, Avg generation throughput: 43.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
271
+ (APIServer pid=11045) INFO 11-18 03:22:59 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.36, Accepted throughput: 25.00 tokens/s, Drafted throughput: 55.20 tokens/s, Accepted: 250 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.766, 0.391, 0.201, Avg Draft acceptance rate: 45.3%
272
+ INFO: 127.0.0.1:25910 - "POST /v1/completions HTTP/1.1" 200 OK
273
+ INFO: 127.0.0.1:46689 - "POST /v1/completions HTTP/1.1" 200 OK
274
+ (APIServer pid=11045) INFO 11-18 03:23:09 [loggers.py:320] Engine 000: Avg prompt throughput: 13.5 tokens/s, Avg generation throughput: 52.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
275
+ (APIServer pid=11045) INFO 11-18 03:23:09 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.90, Accepted throughput: 34.40 tokens/s, Drafted throughput: 54.30 tokens/s, Accepted: 344 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.840, 0.619, 0.442, Avg Draft acceptance rate: 63.4%
276
+ INFO: 127.0.0.1:31120 - "POST /v1/completions HTTP/1.1" 200 OK
277
+ (APIServer pid=11045) INFO 11-18 03:23:19 [loggers.py:320] Engine 000: Avg prompt throughput: 8.3 tokens/s, Avg generation throughput: 57.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
278
+ (APIServer pid=11045) INFO 11-18 03:23:19 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.11, Accepted throughput: 38.90 tokens/s, Drafted throughput: 55.20 tokens/s, Accepted: 389 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.870, 0.701, 0.543, Avg Draft acceptance rate: 70.5%
279
+ INFO: 127.0.0.1:62869 - "POST /v1/completions HTTP/1.1" 200 OK
280
+ (APIServer pid=11045) INFO 11-18 03:23:29 [loggers.py:320] Engine 000: Avg prompt throughput: 3.4 tokens/s, Avg generation throughput: 50.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
281
+ (APIServer pid=11045) INFO 11-18 03:23:29 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.76, Accepted throughput: 32.00 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 320 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.802, 0.549, 0.407, Avg Draft acceptance rate: 58.6%
282
+ INFO: 127.0.0.1:48961 - "POST /v1/completions HTTP/1.1" 200 OK
283
+ (APIServer pid=11045) INFO 11-18 03:23:39 [loggers.py:320] Engine 000: Avg prompt throughput: 4.8 tokens/s, Avg generation throughput: 52.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
284
+ (APIServer pid=11045) INFO 11-18 03:23:39 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.85, Accepted throughput: 33.90 tokens/s, Drafted throughput: 54.90 tokens/s, Accepted: 339 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.814, 0.612, 0.426, Avg Draft acceptance rate: 61.7%
285
+ INFO: 127.0.0.1:53045 - "POST /v1/completions HTTP/1.1" 200 OK
286
+ INFO: 127.0.0.1:57398 - "POST /v1/completions HTTP/1.1" 200 OK
287
+ INFO: 127.0.0.1:23206 - "POST /v1/completions HTTP/1.1" 200 OK
288
+ (APIServer pid=11045) INFO 11-18 03:23:49 [loggers.py:320] Engine 000: Avg prompt throughput: 92.1 tokens/s, Avg generation throughput: 56.4 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
289
+ (APIServer pid=11045) INFO 11-18 03:23:49 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.19, Accepted throughput: 38.80 tokens/s, Drafted throughput: 53.10 tokens/s, Accepted: 388 tokens, Drafted: 531 tokens, Per-position acceptance rate: 0.881, 0.718, 0.593, Avg Draft acceptance rate: 73.1%
290
+ INFO: 127.0.0.1:41930 - "POST /v1/completions HTTP/1.1" 200 OK
291
+ INFO: 127.0.0.1:20447 - "POST /v1/completions HTTP/1.1" 200 OK
292
+ INFO: 127.0.0.1:56129 - "POST /v1/completions HTTP/1.1" 200 OK
293
+ (APIServer pid=11045) INFO 11-18 03:23:59 [loggers.py:320] Engine 000: Avg prompt throughput: 22.4 tokens/s, Avg generation throughput: 64.3 tokens/s, Running: 0 reqs, Waiting: 0 reqs, GPU KV cache usage: 0.0%
294
+ (APIServer pid=11045) INFO 11-18 03:23:59 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.57, Accepted throughput: 46.30 tokens/s, Drafted throughput: 54.00 tokens/s, Accepted: 463 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.961, 0.844, 0.767, Avg Draft acceptance rate: 85.7%
295
+ INFO: 127.0.0.1:24927 - "POST /v1/completions HTTP/1.1" 200 OK
296
+ (APIServer pid=11045) INFO 11-18 03:24:09 [loggers.py:320] Engine 000: Avg prompt throughput: 8.3 tokens/s, Avg generation throughput: 54.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
297
+ (APIServer pid=11045) INFO 11-18 03:24:09 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.97, Accepted throughput: 35.90 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 359 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.841, 0.648, 0.484, Avg Draft acceptance rate: 65.8%
298
+ INFO: 127.0.0.1:28765 - "POST /v1/completions HTTP/1.1" 200 OK
299
+ (APIServer pid=11045) INFO 11-18 03:24:19 [loggers.py:320] Engine 000: Avg prompt throughput: 8.8 tokens/s, Avg generation throughput: 55.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
300
+ (APIServer pid=11045) INFO 11-18 03:24:19 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.04, Accepted throughput: 37.10 tokens/s, Drafted throughput: 54.60 tokens/s, Accepted: 371 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.868, 0.648, 0.522, Avg Draft acceptance rate: 67.9%
301
+ INFO: 127.0.0.1:49945 - "POST /v1/completions HTTP/1.1" 200 OK
302
+ INFO: 127.0.0.1:55101 - "POST /v1/completions HTTP/1.1" 200 OK
303
+ (APIServer pid=11045) INFO 11-18 03:24:29 [loggers.py:320] Engine 000: Avg prompt throughput: 11.9 tokens/s, Avg generation throughput: 58.9 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
304
+ (APIServer pid=11045) INFO 11-18 03:24:29 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.25, Accepted throughput: 40.80 tokens/s, Drafted throughput: 54.29 tokens/s, Accepted: 408 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.917, 0.746, 0.591, Avg Draft acceptance rate: 75.1%
305
+ INFO: 127.0.0.1:22464 - "POST /v1/completions HTTP/1.1" 200 OK
306
+ INFO: 127.0.0.1:61890 - "POST /v1/completions HTTP/1.1" 200 OK
307
+ (APIServer pid=11045) INFO 11-18 03:24:39 [loggers.py:320] Engine 000: Avg prompt throughput: 43.7 tokens/s, Avg generation throughput: 57.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
308
+ (APIServer pid=11045) INFO 11-18 03:24:39 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.18, Accepted throughput: 39.30 tokens/s, Drafted throughput: 53.99 tokens/s, Accepted: 393 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.867, 0.717, 0.600, Avg Draft acceptance rate: 72.8%
309
+ INFO: 127.0.0.1:39253 - "POST /v1/completions HTTP/1.1" 200 OK
310
+ (APIServer pid=11045) INFO 11-18 03:24:49 [loggers.py:320] Engine 000: Avg prompt throughput: 5.1 tokens/s, Avg generation throughput: 64.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
311
+ (APIServer pid=11045) INFO 11-18 03:24:49 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.51, Accepted throughput: 45.90 tokens/s, Drafted throughput: 54.89 tokens/s, Accepted: 459 tokens, Drafted: 549 tokens, Per-position acceptance rate: 0.940, 0.842, 0.727, Avg Draft acceptance rate: 83.6%
312
+ INFO: 127.0.0.1:60415 - "POST /v1/completions HTTP/1.1" 200 OK
313
+ INFO: 127.0.0.1:39665 - "POST /v1/completions HTTP/1.1" 200 OK
314
+ (APIServer pid=11045) INFO 11-18 03:24:59 [loggers.py:320] Engine 000: Avg prompt throughput: 10.5 tokens/s, Avg generation throughput: 62.9 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
315
+ (APIServer pid=11045) INFO 11-18 03:24:59 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.46, Accepted throughput: 44.79 tokens/s, Drafted throughput: 54.59 tokens/s, Accepted: 448 tokens, Drafted: 546 tokens, Per-position acceptance rate: 0.918, 0.819, 0.725, Avg Draft acceptance rate: 82.1%
316
+ INFO: 127.0.0.1:39839 - "POST /v1/completions HTTP/1.1" 200 OK
317
+ (APIServer pid=11045) INFO 11-18 03:25:09 [loggers.py:320] Engine 000: Avg prompt throughput: 3.4 tokens/s, Avg generation throughput: 48.3 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
318
+ (APIServer pid=11045) INFO 11-18 03:25:09 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.64, Accepted throughput: 30.10 tokens/s, Drafted throughput: 55.20 tokens/s, Accepted: 301 tokens, Drafted: 552 tokens, Per-position acceptance rate: 0.766, 0.543, 0.326, Avg Draft acceptance rate: 54.5%
319
+ INFO: 127.0.0.1:27721 - "POST /v1/completions HTTP/1.1" 200 OK
320
+ INFO: 127.0.0.1:18976 - "POST /v1/completions HTTP/1.1" 200 OK
321
+ (APIServer pid=11045) INFO 11-18 03:25:19 [loggers.py:320] Engine 000: Avg prompt throughput: 9.0 tokens/s, Avg generation throughput: 47.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
322
+ (APIServer pid=11045) INFO 11-18 03:25:19 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 2.64, Accepted throughput: 29.69 tokens/s, Drafted throughput: 54.29 tokens/s, Accepted: 297 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.724, 0.541, 0.376, Avg Draft acceptance rate: 54.7%
323
+ INFO: 127.0.0.1:55210 - "POST /v1/completions HTTP/1.1" 200 OK
324
+ INFO: 127.0.0.1:29626 - "POST /v1/completions HTTP/1.1" 200 OK
325
+ (APIServer pid=11045) INFO 11-18 03:25:29 [loggers.py:320] Engine 000: Avg prompt throughput: 12.1 tokens/s, Avg generation throughput: 58.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
326
+ (APIServer pid=11045) INFO 11-18 03:25:29 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.23, Accepted throughput: 40.39 tokens/s, Drafted throughput: 54.29 tokens/s, Accepted: 404 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.906, 0.746, 0.580, Avg Draft acceptance rate: 74.4%
327
+ INFO: 127.0.0.1:16577 - "POST /v1/completions HTTP/1.1" 200 OK
328
+ INFO: 127.0.0.1:59656 - "GET /metrics HTTP/1.1" 200 OK
329
+ INFO: 127.0.0.1:19004 - "GET /metrics HTTP/1.1" 200 OK
330
+ INFO: 127.0.0.1:46174 - "POST /v1/completions HTTP/1.1" 200 OK
331
+ (APIServer pid=11045) INFO 11-18 03:25:39 [loggers.py:320] Engine 000: Avg prompt throughput: 17.0 tokens/s, Avg generation throughput: 62.6 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
332
+ (APIServer pid=11045) INFO 11-18 03:25:39 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.48, Accepted throughput: 44.80 tokens/s, Drafted throughput: 54.30 tokens/s, Accepted: 448 tokens, Drafted: 543 tokens, Per-position acceptance rate: 0.906, 0.834, 0.735, Avg Draft acceptance rate: 82.5%
333
+ INFO: 127.0.0.1:37598 - "POST /v1/completions HTTP/1.1" 200 OK
334
+ INFO: 127.0.0.1:62190 - "POST /v1/completions HTTP/1.1" 200 OK
335
+ (APIServer pid=11045) INFO 11-18 03:25:49 [loggers.py:320] Engine 000: Avg prompt throughput: 39.0 tokens/s, Avg generation throughput: 68.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
336
+ (APIServer pid=11045) INFO 11-18 03:25:49 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.82, Accepted throughput: 50.70 tokens/s, Drafted throughput: 54.00 tokens/s, Accepted: 507 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.972, 0.944, 0.900, Avg Draft acceptance rate: 93.9%
337
+ INFO: 127.0.0.1:24380 - "POST /v1/completions HTTP/1.1" 200 OK
338
+ INFO: 127.0.0.1:29738 - "POST /v1/completions HTTP/1.1" 200 OK
339
+ (APIServer pid=11045) INFO 11-18 03:25:59 [loggers.py:320] Engine 000: Avg prompt throughput: 50.9 tokens/s, Avg generation throughput: 70.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
340
+ (APIServer pid=11045) INFO 11-18 03:25:59 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.93, Accepted throughput: 52.70 tokens/s, Drafted throughput: 54.00 tokens/s, Accepted: 527 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.994, 0.978, 0.956, Avg Draft acceptance rate: 97.6%
341
+ INFO: 127.0.0.1:18286 - "POST /v1/completions HTTP/1.1" 200 OK
342
+ INFO: 127.0.0.1:36057 - "POST /v1/completions HTTP/1.1" 200 OK
343
+ (APIServer pid=11045) INFO 11-18 03:26:09 [loggers.py:320] Engine 000: Avg prompt throughput: 33.3 tokens/s, Avg generation throughput: 67.9 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
344
+ (APIServer pid=11045) INFO 11-18 03:26:09 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.78, Accepted throughput: 49.99 tokens/s, Drafted throughput: 53.99 tokens/s, Accepted: 500 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.978, 0.933, 0.867, Avg Draft acceptance rate: 92.6%
345
+ INFO: 127.0.0.1:17433 - "POST /v1/completions HTTP/1.1" 200 OK
346
+ INFO: 127.0.0.1:41086 - "POST /v1/completions HTTP/1.1" 200 OK
347
+ INFO: 127.0.0.1:17700 - "POST /v1/completions HTTP/1.1" 200 OK
348
+ INFO: 127.0.0.1:51949 - "POST /v1/completions HTTP/1.1" 200 OK
349
+ (APIServer pid=11045) INFO 11-18 03:26:19 [loggers.py:320] Engine 000: Avg prompt throughput: 60.2 tokens/s, Avg generation throughput: 65.7 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
350
+ (APIServer pid=11045) INFO 11-18 03:26:19 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.75, Accepted throughput: 48.40 tokens/s, Drafted throughput: 52.80 tokens/s, Accepted: 484 tokens, Drafted: 528 tokens, Per-position acceptance rate: 0.960, 0.915, 0.875, Avg Draft acceptance rate: 91.7%
351
+ INFO: 127.0.0.1:19946 - "POST /v1/completions HTTP/1.1" 200 OK
352
+ INFO: 127.0.0.1:65173 - "POST /v1/completions HTTP/1.1" 200 OK
353
+ (APIServer pid=11045) INFO 11-18 03:26:29 [loggers.py:320] Engine 000: Avg prompt throughput: 55.2 tokens/s, Avg generation throughput: 70.5 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
354
+ (APIServer pid=11045) INFO 11-18 03:26:29 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.92, Accepted throughput: 52.50 tokens/s, Drafted throughput: 54.00 tokens/s, Accepted: 525 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.983, 0.972, 0.961, Avg Draft acceptance rate: 97.2%
355
+ INFO: 127.0.0.1:30568 - "POST /v1/completions HTTP/1.1" 200 OK
356
+ INFO: 127.0.0.1:56340 - "POST /v1/completions HTTP/1.1" 200 OK
357
+ (APIServer pid=11045) INFO 11-18 03:26:39 [loggers.py:320] Engine 000: Avg prompt throughput: 26.9 tokens/s, Avg generation throughput: 66.8 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
358
+ (APIServer pid=11045) INFO 11-18 03:26:39 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.71, Accepted throughput: 48.80 tokens/s, Drafted throughput: 54.00 tokens/s, Accepted: 488 tokens, Drafted: 540 tokens, Per-position acceptance rate: 0.956, 0.917, 0.839, Avg Draft acceptance rate: 90.4%
359
+ INFO: 127.0.0.1:32598 - "POST /v1/completions HTTP/1.1" 200 OK
360
+ INFO: 127.0.0.1:38921 - "POST /v1/completions HTTP/1.1" 200 OK
361
+ (APIServer pid=11045) INFO 11-18 03:26:49 [loggers.py:320] Engine 000: Avg prompt throughput: 61.6 tokens/s, Avg generation throughput: 70.0 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.2%
362
+ (APIServer pid=11045) INFO 11-18 03:26:49 [metrics.py:120] SpecDecoding metrics: Mean acceptance length: 3.92, Accepted throughput: 52.19 tokens/s, Drafted throughput: 53.69 tokens/s, Accepted: 522 tokens, Drafted: 537 tokens, Per-position acceptance rate: 1.000, 0.989, 0.927, Avg Draft acceptance rate: 97.2%
363
+ INFO: 127.0.0.1:55465 - "POST /v1/completions HTTP/1.1" 200 OK
runs/r1/logs/eval_main.log ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ method: mtp
2
+ parameters: 424.7 M
3
+ weights: 0.540 GB
4
+ KV cache @ full context: 0.067 GB
5
+ memory metric: 0.607 GB
6
+ PASS
7
+ [03:18:05] repeat 1/1: fresh vLLM process
8
+ [03:18:05] starting vLLM (log: /workspace/emc/work/logs/20261118-031805-eval-r1.log)
9
+ [03:20:01] vLLM ready after 116s
10
+ [03:25:39] chat: 40 prompts done
runs/r1/logs/pipeline.log CHANGED
@@ -80,3 +80,4 @@
80
  [2026-11-18 03:03:03] done extract (874s)
81
  [2026-11-18 03:03:03] start train
82
  [2026-11-18 03:17:54] done train (891s)
 
 
80
  [2026-11-18 03:03:03] done extract (874s)
81
  [2026-11-18 03:03:03] start train
82
  [2026-11-18 03:17:54] done train (891s)
83
+ [2026-11-18 03:17:56] start eval_main
runs/r1/logs/pipeline.out CHANGED
@@ -80,3 +80,4 @@
80
  [2026-11-18 03:03:03] done extract (874s)
81
  [2026-11-18 03:03:03] start train
82
  [2026-11-18 03:17:54] done train (891s)
 
 
80
  [2026-11-18 03:03:03] done extract (874s)
81
  [2026-11-18 03:03:03] start train
82
  [2026-11-18 03:17:54] done train (891s)
83
+ [2026-11-18 03:17:56] start eval_main
runs/r1/logs/stats.log CHANGED
@@ -8,3 +8,4 @@
8
  [2026-11-18 03:16:51] ram 2.0/172 GB used (cache 0.1, shmem 0.0) | cgroup 2.0/172 GB | disk 91 GB used | gpu 65.1/95.6 GiB, 100%, 442 W | load 0.0 | largest python 8.2 GB, python 0.1 GB, python 0.1 GB
9
  [2026-11-18 03:17:54] ram 0.2/172 GB used (cache 0.0, shmem 0.0) | cgroup 0.2/172 GB | disk 93 GB used | gpu 0.0/95.6 GiB, 98%, 88 W | load 0.0 | largest python 0.1 GB, python 0.1 GB, python 0.0 GB
10
  [2026-11-18 03:17:55] ram 0.2/172 GB used (cache 0.0, shmem 0.0) | cgroup 0.2/172 GB | disk 93 GB used | gpu 0.0/95.6 GiB, 0%, 83 W | load 0.0 | largest python 0.1 GB, python 0.1 GB, python 0.0 GB
 
 
8
  [2026-11-18 03:16:51] ram 2.0/172 GB used (cache 0.1, shmem 0.0) | cgroup 2.0/172 GB | disk 91 GB used | gpu 65.1/95.6 GiB, 100%, 442 W | load 0.0 | largest python 8.2 GB, python 0.1 GB, python 0.1 GB
9
  [2026-11-18 03:17:54] ram 0.2/172 GB used (cache 0.0, shmem 0.0) | cgroup 0.2/172 GB | disk 93 GB used | gpu 0.0/95.6 GiB, 98%, 88 W | load 0.0 | largest python 0.1 GB, python 0.1 GB, python 0.0 GB
10
  [2026-11-18 03:17:55] ram 0.2/172 GB used (cache 0.0, shmem 0.0) | cgroup 0.2/172 GB | disk 93 GB used | gpu 0.0/95.6 GiB, 0%, 83 W | load 0.0 | largest python 0.1 GB, python 0.1 GB, python 0.0 GB
11
+ [2026-11-18 03:26:52] ram 4.6/172 GB used (cache 0.3, shmem 0.0) | cgroup 4.6/172 GB | disk 93 GB used | gpu 83.1/95.6 GiB, 100%, 290 W | load 0.0 | largest VLLM::EngineCor 4.3 GB, vllm 2.5 GB, python 1.2 GB
runs/r1/logs/sync.log CHANGED
@@ -18,3 +18,6 @@ uploaded logs/results to mudassiruddin21/EMC:runs/r1
18
  uploaded logs/results to mudassiruddin21/EMC:runs/r1
19
  uploaded logs/results to mudassiruddin21/EMC:runs/r1
20
  uploaded logs/results to mudassiruddin21/EMC:runs/r1
 
 
 
 
18
  uploaded logs/results to mudassiruddin21/EMC:runs/r1
19
  uploaded logs/results to mudassiruddin21/EMC:runs/r1
20
  uploaded logs/results to mudassiruddin21/EMC:runs/r1
21
+ uploaded logs/results to mudassiruddin21/EMC:runs/r1
22
+ uploaded sub_bf16 to mudassiruddin21/EMC:submissions/r1/main-bf16
23
+ uploaded sub_fp8 to mudassiruddin21/EMC:submissions/r1/main-fp8