2026-04-16 01:56:57,138 INFO Loaded 6 test events
2026-04-16 01:56:57,139 INFO Found 6 ground truth chains for test events
2026-04-16 01:56:57,471 INFO 
[1/6] Evaluating: 2025-0183-ITA (Italy)
2026-04-16 01:56:57,471 INFO Retrieving similar events for: Flood in Florence, Prato, Pisa, and Livorno (Toscane); Emilia-Romagna, Italy. Date: 2025-03-14. Seve...
2026-04-16 01:57:02,594 INFO No device provided, using cuda:0
2026-04-16 01:57:02,847 INFO Loading SentenceTransformer model from sentence-transformers/all-MiniLM-L6-v2.

Batches:   0%|          | 0/1 [00:00<?, ?it/s]
Batches: 100%|██████████| 1/1 [00:00<00:00,  3.68it/s]
Batches: 100%|██████████| 1/1 [00:00<00:00,  3.67it/s]
2026-04-16 01:57:05,498 INFO Retrieved 5 reference events
2026-04-16 01:57:05,499 INFO Calling LLM for cascade prediction...
2026-04-16 01:57:07,206 INFO Loading model Qwen/Qwen3-8B on GPU(s) [1]...
INFO 04-16 01:57:07 [utils.py:233] non-default args: {'trust_remote_code': True, 'download_dir': '/scratch/prj/cllm/yizhen/cache/hub', 'max_model_len': 8192, 'gpu_memory_utilization': 0.8, 'disable_log_stats': True, 'model': 'Qwen/Qwen3-8B'}
INFO 04-16 01:57:07 [model.py:549] Resolved architecture: Qwen3ForCausalLM
INFO 04-16 01:57:07 [model.py:1678] Using max model len 8192
INFO 04-16 01:57:08 [scheduler.py:238] Chunked prefill is enabled with max_num_batched_tokens=8192.
INFO 04-16 01:57:08 [vllm.py:790] Asynchronous scheduling is enabled.
WARNING 04-16 01:57:09 [system_utils.py:152] We must use the `spawn` multiprocessing start method. Overriding VLLM_WORKER_MULTIPROC_METHOD to 'spawn'. See https://docs.vllm.ai/en/latest/usage/troubleshooting.html#python-multiprocessing for more information. Reasons: CUDA is initialized
(EngineCore pid=1473556) INFO 04-16 01:57:16 [core.py:105] Initializing a V1 LLM engine (v0.19.0) with config: model='Qwen/Qwen3-8B', speculative_config=None, tokenizer='Qwen/Qwen3-8B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=8192, download_dir='/scratch/prj/cllm/yizhen/cache/hub', 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, enforce_eager=False, enable_return_routed_experts=False, kv_cache_dtype=auto, device_config=cuda, structured_outputs_config=StructuredOutputsConfig(backend='auto', disable_any_whitespace=False, disable_additional_properties=False, reasoning_parser='', reasoning_parser_plugin='', enable_in_reasoning=False), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None, kv_cache_metrics=False, kv_cache_metrics_sample=0.01, cudagraph_metrics=False, enable_layerwise_nvtx_tracing=False, enable_mfu_metrics=False, enable_mm_processor_stats=False, enable_logging_iteration_details=False), seed=0, served_model_name=Qwen/Qwen3-8B, enable_prefix_caching=True, enable_chunked_prefill=True, pooler_config=None, compilation_config={'mode': <CompilationMode.VLLM_COMPILE: 3>, 'debug_dump_path': None, 'cache_dir': '', 'compile_cache_save_format': 'binary', 'backend': 'inductor', 'custom_ops': ['none'], 'splitting_ops': ['vllm::unified_attention', 'vllm::unified_attention_with_output', 'vllm::unified_mla_attention', 'vllm::unified_mla_attention_with_output', 'vllm::mamba_mixer2', 'vllm::mamba_mixer', 'vllm::short_conv', 'vllm::linear_attention', 'vllm::plamo2_mamba_mixer', 'vllm::gdn_attention_core', 'vllm::olmo_hybrid_gdn_full_forward', 'vllm::kda_attention', 'vllm::sparse_attn_indexer', 'vllm::rocm_aiter_sparse_attn_indexer', '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_images_per_batch': 0, 'compile_sizes': [], 'compile_ranges_endpoints': [8192], 'inductor_compile_config': {'enable_auto_functionalized_v2': False, 'size_asserts': False, 'alignment_asserts': False, 'scalar_asserts': False, 'combo_kernels': True, 'benchmark_combo_kernel': True}, 'inductor_passes': {}, 'cudagraph_mode': <CUDAGraphMode.FULL_AND_PIECEWISE: (2, 1)>, 'cudagraph_num_of_warmups': 1, 'cudagraph_capture_sizes': [1, 2, 4, 8, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120, 128, 136, 144, 152, 160, 168, 176, 184, 192, 200, 208, 216, 224, 232, 240, 248, 256, 272, 288, 304, 320, 336, 352, 368, 384, 400, 416, 432, 448, 464, 480, 496, 512], '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}, 'max_cudagraph_capture_size': 512, 'dynamic_shapes_config': {'type': <DynamicShapesType.BACKED: 'backed'>, 'evaluate_guards': False, 'assume_32_bit_indexing': False}, 'local_cache_dir': None, 'fast_moe_cold_start': True, 'static_all_moe_layers': []}
(EngineCore pid=1473556) INFO 04-16 01:57:17 [parallel_state.py:1400] world_size=1 rank=0 local_rank=0 distributed_init_method=tcp://10.211.5.142:53629 backend=nccl
(EngineCore pid=1473556) INFO 04-16 01:57:17 [parallel_state.py:1716] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, PCP rank 0, TP rank 0, EP rank N/A, EPLB rank N/A
(EngineCore pid=1473556) INFO 04-16 01:57:18 [gpu_model_runner.py:4735] Starting to load model Qwen/Qwen3-8B...
(EngineCore pid=1473556) INFO 04-16 01:57:18 [cuda.py:334] Using FLASH_ATTN attention backend out of potential backends: ['FLASH_ATTN', 'FLASHINFER', 'TRITON_ATTN', 'FLEX_ATTENTION'].
(EngineCore pid=1473556) INFO 04-16 01:57:18 [flash_attn.py:596] Using FlashAttention version 2
(EngineCore pid=1473556) <frozen importlib._bootstrap_external>:1301: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
(EngineCore pid=1473556) <frozen importlib._bootstrap_external>:1301: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
(EngineCore pid=1473556) 
Loading safetensors checkpoint shards:   0% Completed | 0/5 [00:00<?, ?it/s]
(EngineCore pid=1473556) 
Loading safetensors checkpoint shards:  20% Completed | 1/5 [00:02<00:09,  2.41s/it]
(EngineCore pid=1473556) 
Loading safetensors checkpoint shards:  40% Completed | 2/5 [00:04<00:06,  2.19s/it]
(EngineCore pid=1473556) 
Loading safetensors checkpoint shards:  60% Completed | 3/5 [00:06<00:03,  1.99s/it]
(EngineCore pid=1473556) 
Loading safetensors checkpoint shards:  80% Completed | 4/5 [00:07<00:01,  1.76s/it]
(EngineCore pid=1473556) 
Loading safetensors checkpoint shards: 100% Completed | 5/5 [00:08<00:00,  1.34s/it]
(EngineCore pid=1473556) 
Loading safetensors checkpoint shards: 100% Completed | 5/5 [00:08<00:00,  1.64s/it]
(EngineCore pid=1473556) 
(EngineCore pid=1473556) INFO 04-16 01:57:27 [default_loader.py:384] Loading weights took 8.22 seconds
(EngineCore pid=1473556) INFO 04-16 01:57:28 [gpu_model_runner.py:4820] Model loading took 15.27 GiB memory and 9.576766 seconds
(EngineCore pid=1473556) INFO 04-16 01:57:31 [backends.py:1051] Using cache directory: /users/k25062276/.cache/vllm/torch_compile_cache/3e4edceef4/rank_0_0/backbone for vLLM's torch.compile
(EngineCore pid=1473556) INFO 04-16 01:57:31 [backends.py:1111] Dynamo bytecode transform time: 2.89 s
(EngineCore pid=1473556) INFO 04-16 01:57:32 [backends.py:285] Directly load the compiled graph(s) for compile range (1, 8192) from the cache, took 1.161 s
(EngineCore pid=1473556) INFO 04-16 01:57:32 [decorators.py:303] Directly load AOT compilation from path /users/k25062276/.cache/vllm/torch_compile_cache/torch_aot_compile/609ec08e9dd84001f7688a7e025b973cc84990f46abe61d386de8103fc218ad0/rank_0_0/model
(EngineCore pid=1473556) INFO 04-16 01:57:32 [monitor.py:48] torch.compile took 4.34 s in total
(EngineCore pid=1473556) INFO 04-16 01:57:33 [monitor.py:76] Initial profiling/warmup run took 0.12 s
(EngineCore pid=1473556) INFO 04-16 01:57:33 [kv_cache_utils.py:829] Overriding num_gpu_blocks=0 with num_gpu_blocks_override=512
(EngineCore pid=1473556) INFO 04-16 01:57:33 [gpu_model_runner.py:5876] Profiling CUDA graph memory: PIECEWISE=51 (largest=512), FULL=35 (largest=256)
(EngineCore pid=1473556) INFO 04-16 01:57:34 [gpu_model_runner.py:5955] Estimated CUDA graph memory: 2.35 GiB total
(EngineCore pid=1473556) INFO 04-16 01:57:35 [gpu_worker.py:436] Available KV cache memory: 47.06 GiB
(EngineCore pid=1473556) INFO 04-16 01:57:35 [gpu_worker.py:470] In v0.19, CUDA graph memory profiling will be enabled by default (VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=1), which more accurately accounts for CUDA graph memory during KV cache allocation. To try it now, set VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=1 and increase --gpu-memory-utilization from 0.8000 to 0.8297 to maintain the same effective KV cache size.
(EngineCore pid=1473556) INFO 04-16 01:57:35 [kv_cache_utils.py:1319] GPU KV cache size: 342,704 tokens
(EngineCore pid=1473556) INFO 04-16 01:57:35 [kv_cache_utils.py:1324] Maximum concurrency for 8,192 tokens per request: 41.83x
(EngineCore pid=1473556) 
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):   0%|          | 0/51 [00:00<?, ?it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):   4%|▍         | 2/51 [00:00<00:02, 18.98it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):   8%|▊         | 4/51 [00:00<00:02, 19.26it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  12%|█▏        | 6/51 [00:00<00:02, 19.00it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  16%|█▌        | 8/51 [00:00<00:02, 18.73it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  22%|██▏       | 11/51 [00:00<00:02, 19.83it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  27%|██▋       | 14/51 [00:00<00:01, 20.45it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  33%|███▎      | 17/51 [00:00<00:01, 21.16it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  39%|███▉      | 20/51 [00:00<00:01, 22.30it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  45%|████▌     | 23/51 [00:01<00:01, 23.07it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  51%|█████     | 26/51 [00:01<00:01, 23.70it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  57%|█████▋    | 29/51 [00:01<00:00, 23.54it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  63%|██████▎   | 32/51 [00:01<00:00, 24.12it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  69%|██████▊   | 35/51 [00:01<00:00, 25.05it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  75%|███████▍  | 38/51 [00:01<00:00, 25.74it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  80%|████████  | 41/51 [00:01<00:00, 26.42it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  86%|████████▋ | 44/51 [00:01<00:00, 27.18it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE):  92%|█████████▏| 47/51 [00:01<00:00, 27.81it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 51/51 [00:02<00:00, 27.50it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 51/51 [00:02<00:00, 24.01it/s]
(EngineCore pid=1473556) 
Capturing CUDA graphs (decode, FULL):   0%|          | 0/35 [00:00<?, ?it/s]
Capturing CUDA graphs (decode, FULL):   9%|▊         | 3/35 [00:00<00:01, 27.31it/s]
Capturing CUDA graphs (decode, FULL):  17%|█▋        | 6/35 [00:00<00:01, 25.20it/s]
Capturing CUDA graphs (decode, FULL):  26%|██▌       | 9/35 [00:00<00:00, 26.61it/s]
Capturing CUDA graphs (decode, FULL):  34%|███▍      | 12/35 [00:00<00:00, 27.50it/s]
Capturing CUDA graphs (decode, FULL):  43%|████▎     | 15/35 [00:00<00:00, 28.15it/s]
Capturing CUDA graphs (decode, FULL):  54%|█████▍    | 19/35 [00:00<00:00, 29.64it/s]
Capturing CUDA graphs (decode, FULL):  66%|██████▌   | 23/35 [00:00<00:00, 30.95it/s]
Capturing CUDA graphs (decode, FULL):  77%|███████▋  | 27/35 [00:00<00:00, 32.05it/s]
Capturing CUDA graphs (decode, FULL):  89%|████████▊ | 31/35 [00:01<00:00, 33.11it/s]
Capturing CUDA graphs (decode, FULL): 100%|██████████| 35/35 [00:01<00:00, 34.01it/s]
Capturing CUDA graphs (decode, FULL): 100%|██████████| 35/35 [00:01<00:00, 30.97it/s]
(EngineCore pid=1473556) INFO 04-16 01:57:38 [gpu_model_runner.py:6046] Graph capturing finished in 4 secs, took 2.44 GiB
(EngineCore pid=1473556) INFO 04-16 01:57:38 [gpu_worker.py:597] CUDA graph pool memory: 2.44 GiB (actual), 2.35 GiB (estimated), difference: 0.09 GiB (3.6%).
(EngineCore pid=1473556) INFO 04-16 01:57:38 [core.py:283] init engine (profile, create kv cache, warmup model) took 10.44 seconds
(EngineCore pid=1473556) INFO 04-16 01:57:40 [vllm.py:790] Asynchronous scheduling is enabled.
2026-04-16 01:57:40,070 INFO Model loaded successfully.

Rendering prompts:   0%|          | 0/1 [00:00<?, ?it/s]
Rendering prompts: 100%|██████████| 1/1 [00:00<00:00, 40.65it/s]

Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:24<00:00, 24.06s/it, est. speed input: 231.41 toks/s, output: 80.31 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:24<00:00, 24.06s/it, est. speed input: 231.41 toks/s, output: 80.31 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:24<00:00, 24.06s/it, est. speed input: 231.41 toks/s, output: 80.31 toks/s]
2026-04-16 01:58:04,171 INFO   Predicted 10 cascade nodes
2026-04-16 01:58:04,171 INFO   Domains: ['economy/business', 'environment/pollution', 'health/disease', 'health/hospital', 'infrastructure/communication', 'infrastructure/power', 'infrastructure/transport', 'social/displacement', 'social/housing']
2026-04-16 01:58:04,171 INFO   Domain F1: 1.00
2026-04-16 01:58:04,171 INFO 
[2/6] Evaluating: 2025-0230-BIH (Bosnia and Herzegovina)
2026-04-16 01:58:04,171 INFO Retrieving similar events for: Flood in Prijedor and Sanski (Slavonia region), Bosnia and Herzegovina. Date: 2025-03-26. Severity: ...

Batches:   0%|          | 0/1 [00:00<?, ?it/s]
Batches: 100%|██████████| 1/1 [00:00<00:00, 56.61it/s]
2026-04-16 01:58:04,215 INFO Retrieved 5 reference events
2026-04-16 01:58:04,215 INFO Calling LLM for cascade prediction...

Rendering prompts:   0%|          | 0/1 [00:00<?, ?it/s]
Rendering prompts: 100%|██████████| 1/1 [00:00<00:00, 45.47it/s]

Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:19<00:00, 19.63s/it, est. speed input: 278.40 toks/s, output: 80.79 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:19<00:00, 19.63s/it, est. speed input: 278.40 toks/s, output: 80.79 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:19<00:00, 19.63s/it, est. speed input: 278.40 toks/s, output: 80.79 toks/s]
2026-04-16 01:58:23,869 INFO   Predicted 7 cascade nodes
2026-04-16 01:58:23,869 INFO   Domains: ['health/disease', 'health/hospital', 'infrastructure/communication', 'infrastructure/power', 'infrastructure/transport', 'social/displacement']
2026-04-16 01:58:23,869 INFO   Domain F1: 0.80
2026-04-16 01:58:23,869 INFO 
[3/6] Evaluating: 2025-0230-HRV (Croatia)
2026-04-16 01:58:23,870 INFO Retrieving similar events for: Flood in Dubrovnik city area, Croatia. Date: 2025-03-26. Severity: medium. Flood in Dubrovnik city a...

Batches:   0%|          | 0/1 [00:00<?, ?it/s]
Batches: 100%|██████████| 1/1 [00:00<00:00, 66.27it/s]
2026-04-16 01:58:23,971 INFO Retrieved 5 reference events
2026-04-16 01:58:23,971 INFO Calling LLM for cascade prediction...

Rendering prompts:   0%|          | 0/1 [00:00<?, ?it/s]
Rendering prompts: 100%|██████████| 1/1 [00:00<00:00, 36.18it/s]

Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:22<00:00, 22.73s/it, est. speed input: 237.65 toks/s, output: 81.01 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:22<00:00, 22.73s/it, est. speed input: 237.65 toks/s, output: 81.01 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:22<00:00, 22.73s/it, est. speed input: 237.65 toks/s, output: 81.01 toks/s]
2026-04-16 01:58:46,730 INFO   Predicted 10 cascade nodes
2026-04-16 01:58:46,730 INFO   Domains: ['economy/business', 'environment/pollution', 'health/disease', 'health/emergency_services', 'health/hospital', 'infrastructure/communication', 'infrastructure/power', 'infrastructure/transport', 'social/displacement']
2026-04-16 01:58:46,730 INFO   Domain F1: 1.00
2026-04-16 01:58:46,731 INFO 
[4/6] Evaluating: 2025-0632-ROU (Romania)
2026-04-16 01:58:46,731 INFO Retrieving similar events for: Flood in Suceava, Neamt counties, Romania. Date: 2025-07-28. Severity: medium. Flood in Suceava, Nea...

Batches:   0%|          | 0/1 [00:00<?, ?it/s]
Batches: 100%|██████████| 1/1 [00:00<00:00, 87.98it/s]
2026-04-16 01:58:46,779 INFO Retrieved 5 reference events
2026-04-16 01:58:46,780 INFO Calling LLM for cascade prediction...

Rendering prompts:   0%|          | 0/1 [00:00<?, ?it/s]
Rendering prompts: 100%|██████████| 1/1 [00:00<00:00, 36.85it/s]

Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:19<00:00, 19.44s/it, est. speed input: 281.16 toks/s, output: 80.67 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:19<00:00, 19.44s/it, est. speed input: 281.16 toks/s, output: 80.67 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:19<00:00, 19.44s/it, est. speed input: 281.16 toks/s, output: 80.67 toks/s]
2026-04-16 01:59:06,248 INFO   Predicted 8 cascade nodes
2026-04-16 01:59:06,248 INFO   Domains: ['economy/business', 'health/disease', 'health/hospital', 'infrastructure/communication', 'infrastructure/power', 'infrastructure/transport', 'social/displacement']
2026-04-16 01:59:06,248 INFO   Domain F1: 0.93
2026-04-16 01:59:06,248 INFO 
[5/6] Evaluating: 2025-0848-UKR (Ukraine)
2026-04-16 01:59:06,248 INFO Retrieving similar events for: Flood in Odesa city (Odesa Oblast), Ukraine. Date: 2025-09-30. Severity: medium. Flood in Odesa city...

Batches:   0%|          | 0/1 [00:00<?, ?it/s]
Batches: 100%|██████████| 1/1 [00:00<00:00, 82.11it/s]
2026-04-16 01:59:06,285 INFO Retrieved 5 reference events
2026-04-16 01:59:06,285 INFO Calling LLM for cascade prediction...

Rendering prompts:   0%|          | 0/1 [00:00<?, ?it/s]
Rendering prompts: 100%|██████████| 1/1 [00:00<00:00, 35.73it/s]

Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:18<00:00, 18.37s/it, est. speed input: 297.84 toks/s, output: 80.50 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:18<00:00, 18.37s/it, est. speed input: 297.84 toks/s, output: 80.50 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:18<00:00, 18.37s/it, est. speed input: 297.84 toks/s, output: 80.50 toks/s]
2026-04-16 01:59:24,688 INFO   Predicted 7 cascade nodes
2026-04-16 01:59:24,689 INFO   Domains: ['health/disease', 'health/hospital', 'infrastructure/communication', 'infrastructure/power', 'infrastructure/transport', 'social/displacement']
2026-04-16 01:59:24,689 INFO   Domain F1: 0.80
2026-04-16 01:59:24,689 INFO 
[6/6] Evaluating: 2025-0852-BGR (Bulgaria)
2026-04-16 01:59:24,689 INFO Retrieving similar events for: Flood in Elenite (Burgas province), Bulgaria. Date: 2025-10-03. Severity: medium. Flood in Elenite (...

Batches:   0%|          | 0/1 [00:00<?, ?it/s]
Batches: 100%|██████████| 1/1 [00:00<00:00, 85.51it/s]
2026-04-16 01:59:24,788 INFO Retrieved 5 reference events
2026-04-16 01:59:24,788 INFO Calling LLM for cascade prediction...

Rendering prompts:   0%|          | 0/1 [00:00<?, ?it/s]
Rendering prompts: 100%|██████████| 1/1 [00:00<00:00, 32.74it/s]

Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:24<00:00, 24.36s/it, est. speed input: 229.40 toks/s, output: 80.72 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:24<00:00, 24.36s/it, est. speed input: 229.40 toks/s, output: 80.72 toks/s]
Processed prompts: 100%|██████████| 1/1 [00:24<00:00, 24.36s/it, est. speed input: 229.40 toks/s, output: 80.72 toks/s]
2026-04-16 01:59:49,178 INFO   Predicted 10 cascade nodes
2026-04-16 01:59:49,178 INFO   Domains: ['economy/business', 'environment/pollution', 'health/disease', 'health/hospital', 'infrastructure/communication', 'infrastructure/power', 'infrastructure/transport', 'social/displacement', 'social/housing']
2026-04-16 01:59:49,178 INFO   Domain F1: 0.89
(EngineCore pid=1473556) INFO 04-16 01:59:49 [core.py:1210] Shutdown initiated (timeout=0)
(EngineCore pid=1473556) INFO 04-16 01:59:49 [core.py:1233] Shutdown complete
[rank0]:[W416 01:59:49.709939690 ProcessGroupNCCL.cpp:1553] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())

============================================================
EVALUATION SUMMARY
============================================================
Total test events: 6
Events with ground truth: 6
Avg predicted nodes per event: 8.7
Avg reference similarity: 0.60
Avg prediction confidence: 0.84

Domain matching (vs ground truth):
  Precision: 0.98
  Recall:    0.85
  F1:        0.90

Detailed report saved to: data/evaluation_report.json