(APIServer pid=28509) INFO 07-16 17:16:59 [api_utils.py:339] 
(APIServer pid=28509) INFO 07-16 17:16:59 [api_utils.py:339]        █     █     █▄   ▄█
(APIServer pid=28509) INFO 07-16 17:16:59 [api_utils.py:339]  ▄▄ ▄█ █     █     █ ▀▄▀ █  version 0.25.0
(APIServer pid=28509) INFO 07-16 17:16:59 [api_utils.py:339]   █▄█▀ █     █     █     █  model   outputs/healed/grid_math/glean_keep75_s1224/step0150
(APIServer pid=28509) INFO 07-16 17:16:59 [api_utils.py:339]    ▀▀  ▀▀▀▀▀ ▀▀▀▀▀ ▀     ▀
(APIServer pid=28509) INFO 07-16 17:16:59 [api_utils.py:339] 
(APIServer pid=28509) INFO 07-16 17:16:59 [api_utils.py:273] non-default args: {'model_tag': 'outputs/healed/grid_math/glean_keep75_s1224/step0150', 'host': '127.0.0.1', 'port': 8391, 'model': 'outputs/healed/grid_math/glean_keep75_s1224/step0150', 'max_model_len': 2048, 'enforce_eager': True, 'served_model_name': ['student'], 'gpu_memory_utilization': 0.85}
(APIServer pid=28509) INFO 07-16 17:16:59 [model.py:619] Resolved architecture: PrunedOlmoeForCausalLM
(APIServer pid=28509) INFO 07-16 17:16:59 [model.py:1776] Using max model len 2048
(APIServer pid=28509) INFO 07-16 17:16:59 [vllm.py:1042] Asynchronous scheduling is enabled.
(APIServer pid=28509) WARNING 07-16 17:16:59 [vllm.py:1096] Enforce eager set, disabling torch.compile and CUDAGraphs. This is equivalent to setting -cc.mode=none -cc.cudagraph_mode=none
(APIServer pid=28509) WARNING 07-16 17:16:59 [vllm.py:1144] Inductor compilation was disabled by user settings, optimizations settings that are only active during inductor compilation will be ignored.
(APIServer pid=28509) INFO 07-16 17:16:59 [kernel.py:292] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['vllm_c', 'native'], fused_add_rms_norm=['vllm_c', 'native'])
(APIServer pid=28509) INFO 07-16 17:16:59 [vllm.py:1322] Cudagraph is disabled under eager mode
(APIServer pid=28509) INFO 07-16 17:16:59 [compilation.py:312] Enabled custom fusions: norm_quant, act_quant
(EngineCore pid=28706) INFO 07-16 17:17:06 [core.py:114] Initializing a V1 LLM engine (v0.25.0) with config: model='outputs/healed/grid_math/glean_keep75_s1224/step0150', speculative_config=None, tokenizer='outputs/healed/grid_math/glean_keep75_s1224/step0150', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=2048, download_dir=None, load_format=auto, tensor_parallel_size=1, pipeline_parallel_size=1, data_parallel_size=1, decode_context_parallel_size=1, dcp_comm_backend=ag_rs, disable_custom_all_reduce=False, quantization=None, quantization_config=None, enforce_eager=True, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False, jit_monitor_mode='warn', jit_monitor_verbose=False), seed=0, served_model_name=student, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'mode': <CompilationMode.NONE: 0>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['all'], 'ir_enable_torch_wrap': False, 'splitting_ops': [], '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': [2048], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'size_asserts': False, 'alignment_asserts': False, 'scalar_asserts': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.NONE: 0>, 'cudagraph_num_of_warmups': 0, 'cudagraph_capture_sizes': [], 'cudagraph_copy_inputs': False, 'cudagraph_specialize_lora': True, 'use_inductor_graph_partition': False, 'pass_config': {'fuse_norm_quant': True, 'fuse_act_quant': True, 'fuse_attn_quant': False, 'enable_sp': False, 'fuse_gemm_comms': False, 'fuse_allreduce_rms': False, 'fuse_rope_kvcache_cat_mla': False, 'fuse_act_padding': False}, 'max_cudagraph_capture_size': 0, '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=['vllm_c', 'native'], fused_add_rms_norm=['vllm_c', 'native']), enable_flashinfer_autotune=True, enable_cutedsl_warmup=True, moe_backend='auto', linear_backend='auto')
(EngineCore pid=28706) INFO 07-16 17:17:07 [parallel_state.py:1607] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://192.168.0.15:55697 backend=nccl
(EngineCore pid=28706) INFO 07-16 17:17:07 [parallel_state.py:1942] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank 0, EPLB rank N/A
(EngineCore pid=28706) INFO 07-16 17:17:08 [topk_topp_sampler.py:55] Using FlashInfer for top-p & top-k sampling.
(EngineCore pid=28706) INFO 07-16 17:17:08 [gpu_model_runner.py:5209] Starting to load model outputs/healed/grid_math/glean_keep75_s1224/step0150...
(EngineCore pid=28706) INFO 07-16 17:17:08 [cuda.py:476] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
(EngineCore pid=28706) INFO 07-16 17:17:08 [flash_attn.py:718] Using FlashAttention version 2
(EngineCore pid=28706) /home/henry/.cache/glean/megablocks-variable-93a1479bc15b/megablocks/grouped_gemm_util.py:10: UserWarning: Grouped GEMM not available.
(EngineCore pid=28706)   warnings.warn('Grouped GEMM not available.')
(EngineCore pid=28706) INFO 07-16 17:17:08 [weight_utils.py:849] Filesystem type for checkpoints: EXT4. Checkpoint size: 9.89 GiB. Available RAM: 98.08 GiB.
(EngineCore pid=28706) INFO 07-16 17:17:08 [weight_utils.py:872] Auto-prefetch is disabled because the filesystem (EXT4) is not a recognized network FS (NFS/Lustre). If you want to force prefetching, start vLLM with --safetensors-load-strategy=prefetch.
(EngineCore pid=28706) 
Loading safetensors checkpoint shards:   0% Completed | 0/3 [00:00<?, ?it/s]
(EngineCore pid=28706) 
Loading safetensors checkpoint shards:  33% Completed | 1/3 [00:03<00:06,  3.15s/it]
(EngineCore pid=28706) 
Loading safetensors checkpoint shards:  67% Completed | 2/3 [00:06<00:03,  3.54s/it]
(EngineCore pid=28706) 
Loading safetensors checkpoint shards: 100% Completed | 3/3 [00:07<00:00,  2.15s/it]
(EngineCore pid=28706) 
Loading safetensors checkpoint shards: 100% Completed | 3/3 [00:07<00:00,  2.49s/it]
(EngineCore pid=28706) 
(EngineCore pid=28706) INFO 07-16 17:17:16 [default_loader.py:430] Loading weights took 7.47 seconds
(EngineCore pid=28706) INFO 07-16 17:17:16 [gpu_model_runner.py:5306] Model loading took 9.89 GiB memory and 7.657210 seconds
(EngineCore pid=28706) INFO 07-16 17:17:18 [gpu_worker.py:538] Available KV cache memory: 9.8 GiB
(EngineCore pid=28706) INFO 07-16 17:17:18 [kv_cache_utils.py:2146] GPU KV cache size: 80,256 tokens
(EngineCore pid=28706) INFO 07-16 17:17:18 [kv_cache_utils.py:2147] Maximum concurrency for 2,048 tokens per request: 39.19x
(EngineCore pid=28706) INFO 07-16 17:17:18 [cutedsl_warmup.py:97] Skipping CuTeDSL warmup because no compile units were requested.
(EngineCore pid=28706) INFO 07-16 17:17:18 [jit_monitor.py:73] Kernel JIT monitor activated; monitored JIT compilations during inference will use mode=warn.
(EngineCore pid=28706) INFO 07-16 17:17:18 [core.py:344] init engine (profile, create kv cache, warmup model) took 2.07 s
(EngineCore pid=28706) INFO 07-16 17:17:18 [vllm.py:1042] Asynchronous scheduling is enabled.
(EngineCore pid=28706) WARNING 07-16 17:17:18 [vllm.py:1096] Enforce eager set, disabling torch.compile and CUDAGraphs. This is equivalent to setting -cc.mode=none -cc.cudagraph_mode=none
(EngineCore pid=28706) WARNING 07-16 17:17:18 [vllm.py:1144] Inductor compilation was disabled by user settings, optimizations settings that are only active during inductor compilation will be ignored.
(EngineCore pid=28706) INFO 07-16 17:17:18 [kernel.py:292] Final IR op priority after setting platform defaults: IrOpPriorityConfig(rms_norm=['vllm_c', 'native'], fused_add_rms_norm=['vllm_c', 'native'])
(EngineCore pid=28706) INFO 07-16 17:17:18 [vllm.py:1322] Cudagraph is disabled under eager mode
(EngineCore pid=28706) INFO 07-16 17:17:18 [compilation.py:312] Enabled custom fusions: norm_quant, act_quant
(APIServer pid=28509) INFO 07-16 17:17:18 [api_server.py:612] Supported tasks: ['generate']
(APIServer pid=28509) WARNING 07-16 17:17:18 [__init__.py:36] SECURITY WARNING: Development endpoints are enabled! This should NOT be used in production!
(APIServer pid=28509) INFO 07-16 17:17:19 [hf.py:548] Detected the chat template content format to be 'string'. You can set `--chat-template-content-format` to override this.
(APIServer pid=28509) INFO 07-16 17:17:19 [api_server.py:616] Starting vLLM server on http://127.0.0.1:8391
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:37] Available routes are:
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /openapi.json, Methods: GET, HEAD
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /docs, Methods: GET, HEAD
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /docs/oauth2-redirect, Methods: GET, HEAD
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /redoc, Methods: GET, HEAD
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /load, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /version, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /health, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /metrics, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /tokenize, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /detokenize, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/models, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /ping, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /ping, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /invocations, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /reset_prefix_cache, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /reset_mm_cache, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /reset_encoder_cache, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /pause, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /resume, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /is_paused, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /init_weight_transfer_engine, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /start_weight_update, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /update_weights, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /finish_weight_update, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /get_world_size, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /collective_rpc, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /server_info, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /sleep, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /wake_up, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /is_sleeping, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/chat/completions, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/chat/completions/batch, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/responses, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/responses/{response_id}, Methods: GET
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/responses/{response_id}/cancel, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/completions, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/messages, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/messages/count_tokens, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /generative_scoring, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /scale_elastic_ep, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /is_scaling_elastic_ep, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/chat/completions/render, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/completions/render, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/chat/completions/derender, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /v1/completions/derender, Methods: POST
(APIServer pid=28509) INFO 07-16 17:17:19 [launcher.py:46] Route: /inference/v1/generate, Methods: POST
(APIServer pid=28509) INFO:     Started server process [28509]
(APIServer pid=28509) INFO:     Waiting for application startup.
(APIServer pid=28509) INFO:     Application startup complete.
(APIServer pid=28509) INFO:     127.0.0.1:53614 - "GET /health HTTP/1.1" 200 OK
(EngineCore pid=28706) WARNING 07-16 17:17:20 [jit_monitor.py:129] Triton kernel JIT compilation during inference: _build_route_rows. This causes a latency spike; consider extending warmup to cover this shape/config.
(APIServer pid=28509) INFO 07-16 17:17:29 [loggers.py:273] Engine 000: Avg prompt throughput: 2550.5 tokens/s, Avg generation throughput: 1970.3 tokens/s, Running: 254 reqs, Waiting: 986 reqs, GPU KV cache usage: 49.4%, Prefix cache hit rate: 93.2%
(APIServer pid=28509) INFO 07-16 17:17:39 [loggers.py:273] Engine 000: Avg prompt throughput: 1806.8 tokens/s, Avg generation throughput: 2511.3 tokens/s, Running: 255 reqs, Waiting: 759 reqs, GPU KV cache usage: 51.4%, Prefix cache hit rate: 93.3%
(APIServer pid=28509) INFO 07-16 17:17:49 [loggers.py:273] Engine 000: Avg prompt throughput: 1821.4 tokens/s, Avg generation throughput: 2485.1 tokens/s, Running: 252 reqs, Waiting: 525 reqs, GPU KV cache usage: 49.4%, Prefix cache hit rate: 93.3%
(APIServer pid=28509) INFO 07-16 17:17:59 [loggers.py:273] Engine 000: Avg prompt throughput: 1795.2 tokens/s, Avg generation throughput: 2511.9 tokens/s, Running: 255 reqs, Waiting: 305 reqs, GPU KV cache usage: 52.9%, Prefix cache hit rate: 93.4%
(APIServer pid=28509) INFO 07-16 17:18:09 [loggers.py:273] Engine 000: Avg prompt throughput: 1723.0 tokens/s, Avg generation throughput: 2538.1 tokens/s, Running: 254 reqs, Waiting: 87 reqs, GPU KV cache usage: 52.5%, Prefix cache hit rate: 93.4%
(APIServer pid=28509) INFO 07-16 17:18:19 [loggers.py:273] Engine 000: Avg prompt throughput: 722.7 tokens/s, Avg generation throughput: 2337.9 tokens/s, Running: 86 reqs, Waiting: 0 reqs, GPU KV cache usage: 26.9%, Prefix cache hit rate: 93.4%
(APIServer pid=28509) INFO 07-16 17:18:29 [loggers.py:273] Engine 000: Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 466.2 tokens/s, Running: 1 reqs, Waiting: 0 reqs, GPU KV cache usage: 2.1%, Prefix cache hit rate: 93.4%
(APIServer pid=28509) INFO:     127.0.0.1:53626 - "POST /v1/completions HTTP/1.1" 200 OK
(EngineCore pid=28706) INFO 07-16 17:18:30 [core.py:1214] [shutdown] EngineCore: trigger received signal=SIGTERM
(APIServer pid=28509) INFO 07-16 17:18:30 [launcher.py:100] [shutdown] API server: shutdown triggered
(EngineCore pid=28706) INFO 07-16 17:18:30 [core.py:1333] [shutdown] EngineCore: start mode=abort timeout=0s
(APIServer pid=28509) INFO 07-16 17:18:30 [launcher.py:116] [shutdown] API server: stopping engine client mode=abort timeout=0s
(EngineCore pid=28706) INFO 07-16 17:18:30 [core.py:1364] [shutdown] EngineCore: request processing complete; starting resource teardown
(EngineCore pid=28706) INFO 07-16 17:18:30 [core.py:1227] [shutdown] EngineCore: exiting busy loop
(APIServer pid=28509) INFO 07-16 17:18:30 [core_client.py:655] [shutdown] MPClient: start timeout=0s
(APIServer pid=28509) INFO 07-16 17:18:30 [core_client.py:657] [shutdown] MPClient: stopping engine manager
(APIServer pid=28509) WARNING 07-16 17:18:30 [utils.py:626] [shutdown] Process manager: force killing remaining processes count=1
(APIServer pid=28509) INFO 07-16 17:18:30 [core_client.py:659] [shutdown] MPClient: engine manager stopped
(APIServer pid=28509) INFO 07-16 17:18:30 [core_client.py:660] [shutdown] MPClient: cleaning up background resources
(APIServer pid=28509) INFO 07-16 17:18:30 [core_client.py:662] [shutdown] MPClient: complete
(APIServer pid=28509) INFO 07-16 17:18:30 [launcher.py:125] [shutdown] API server: engine client stopped
(APIServer pid=28509) INFO 07-16 17:18:30 [launcher.py:128] [shutdown] API server: signalling HTTP server shutdown
(APIServer pid=28509) INFO 07-16 17:18:30 [launcher.py:149] [shutdown] API server: shutting down FastAPI HTTP server
(APIServer pid=28509) INFO:     Shutting down
(APIServer pid=28509) INFO:     Waiting for application shutdown.
(APIServer pid=28509) INFO:     Application shutdown complete.
/home/henry/.local/share/uv/python/cpython-3.12.12-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
  warnings.warn('resource_tracker: There appear to be %d '