(vllm) root@ubuntu:/data/app_workspace/minerU# time docker exec -it mineru-api mineru -p /app/input -o /app/output --source modelscope
2026-03-26 18:25:45.780470798 [W:onnxruntime:Default, device_discovery.cc:325 DiscoverDevicesForPlatform] GPU device discovery failed: device_discovery.cc:92 ReadFileContents Failed to open file: "/sys/class/drm/card0/device/vendor"
2026-03-26 18:25:47.240 | INFO     | mineru.utils.engine_utils:get_vlm_engine:32 - Using vllm-engine as the inference engine for VLM.
Downloading Model from https://www.modelscope.cn to directory: /root/.cache/modelscope/models/OpenDataLab/MinerU2.5-2509-1.2B
2026-03-26 18:25:50,225 - modelscope - INFO - Target directory already exists, skipping creation.
2026-03-26 18:25:50.226 | INFO     | mineru.backend.vlm.utils:enable_custom_logits_processors:55 - compute_capability: 9.0 >= 8.0 and vllm version: 0.10.1.1 >= 0.10.1, enable custom_logits_processors
/usr/local/lib/python3.12/dist-packages/transformers/utils/hub.py:111: FutureWarning: Using `TRANSFORMERS_CACHE` is deprecated and will be removed in v5 of Transformers. Use `HF_HOME` instead.
  warnings.warn(
INFO 03-26 18:25:51 [__init__.py:241] Automatically detected platform cuda.
INFO 03-26 18:25:51 [utils.py:326] non-default args: {'model': '/root/.cache/modelscope/models/OpenDataLab/MinerU2___5-2509-1___2B', 'gpu_memory_utilization': 0.5, 'disable_log_stats': True, 'logits_processors': [<class 'mineru_vl_utils.logits_processor.vllm_v1_no_repeat_ngram.VllmV1NoRepeatNGramLogitsProcessor'>]}
INFO 03-26 18:25:57 [__init__.py:711] Resolved architecture: Qwen2VLForConditionalGeneration
INFO 03-26 18:25:57 [__init__.py:1750] Using max model len 16384
INFO 03-26 18:25:57 [scheduler.py:222] Chunked prefill is enabled with max_num_batched_tokens=16384.
WARNING 03-26 18:25:59 [__init__.py:2921] 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
2026-03-26 18:26:00.497008484 [W:onnxruntime:Default, device_discovery.cc:325 DiscoverDevicesForPlatform] GPU device discovery failed: device_discovery.cc:92 ReadFileContents Failed to open file: "/sys/class/drm/card0/device/vendor"
/usr/local/lib/python3.12/dist-packages/transformers/utils/hub.py:111: FutureWarning: Using `TRANSFORMERS_CACHE` is deprecated and will be removed in v5 of Transformers. Use `HF_HOME` instead.
  warnings.warn(
INFO 03-26 18:26:03 [__init__.py:241] Automatically detected platform cuda.
(EngineCore_0 pid=1468) INFO 03-26 18:26:04 [core.py:636] Waiting for init message from front-end.
(EngineCore_0 pid=1468) INFO 03-26 18:26:04 [core.py:74] Initializing a V1 LLM engine (v0.10.1.1) with config: model='/root/.cache/modelscope/models/OpenDataLab/MinerU2___5-2509-1___2B', speculative_config=None, tokenizer='/root/.cache/modelscope/models/OpenDataLab/MinerU2___5-2509-1___2B', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config={}, 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, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, kv_cache_dtype=auto, device_config=cuda, decoding_config=DecodingConfig(backend='auto', disable_fallback=False, disable_any_whitespace=False, disable_additional_properties=False, reasoning_backend=''), observability_config=ObservabilityConfig(show_hidden_metrics_for_version=None, otlp_traces_endpoint=None, collect_detailed_traces=None), seed=0, served_model_name=/root/.cache/modelscope/models/OpenDataLab/MinerU2___5-2509-1___2B, enable_prefix_caching=True, chunked_prefill_enabled=True, use_async_output_proc=True, pooler_config=None, compilation_config={"level":3,"debug_dump_path":"","cache_dir":"","backend":"","custom_ops":[],"splitting_ops":["vllm.unified_attention","vllm.unified_attention_with_output","vllm.mamba_mixer2"],"use_inductor":true,"compile_sizes":[],"inductor_compile_config":{"enable_auto_functionalized_v2":false},"inductor_passes":{},"cudagraph_mode":1,"use_cudagraph":true,"cudagraph_num_of_warmups":1,"cudagraph_capture_sizes":[512,504,496,488,480,472,464,456,448,440,432,424,416,408,400,392,384,376,368,360,352,344,336,328,320,312,304,296,288,280,272,264,256,248,240,232,224,216,208,200,192,184,176,168,160,152,144,136,128,120,112,104,96,88,80,72,64,56,48,40,32,24,16,8,4,2,1],"cudagraph_copy_inputs":false,"full_cuda_graph":false,"pass_config":{},"max_capture_size":512,"local_cache_dir":null}
(EngineCore_0 pid=1468) INFO 03-26 18:26:06 [parallel_state.py:1134] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, TP rank 0, EP rank 0
(EngineCore_0 pid=1468) INFO 03-26 18:26:06 [topk_topp_sampler.py:50] Using FlashInfer for top-p & top-k sampling.
(EngineCore_0 pid=1468) INFO 03-26 18:26:08 [gpu_model_runner.py:1953] Starting to load model /root/.cache/modelscope/models/OpenDataLab/MinerU2___5-2509-1___2B...
(EngineCore_0 pid=1468) INFO 03-26 18:26:08 [gpu_model_runner.py:1985] Loading model from scratch...
(EngineCore_0 pid=1468) WARNING 03-26 18:26:08 [cuda.py:211] Current `vllm-flash-attn` has a bug inside vision module, so we use xformers backend instead. You can run `pip install flash-attn` to use flash-attention backend.
(EngineCore_0 pid=1468) INFO 03-26 18:26:09 [cuda.py:328] Using Flash Attention backend on V1 engine.
Loading safetensors checkpoint shards:   0% Completed | 0/1 [00:00<?, ?it/s]
Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00,  2.93it/s]
Loading safetensors checkpoint shards: 100% Completed | 1/1 [00:00<00:00,  2.93it/s]
(EngineCore_0 pid=1468) 
(EngineCore_0 pid=1468) INFO 03-26 18:26:09 [default_loader.py:262] Loading weights took 0.34 seconds
(EngineCore_0 pid=1468) INFO 03-26 18:26:09 [gpu_model_runner.py:2007] Model loading took 2.1647 GiB and 1.063956 seconds
(EngineCore_0 pid=1468) INFO 03-26 18:26:10 [gpu_model_runner.py:2591] Encoder cache will be initialized with a budget of 16384 tokens, and profiled with 1 video items of the maximum feature size.
(EngineCore_0 pid=1468) INFO 03-26 18:26:18 [backends.py:548] Using cache directory: /root/.cache/vllm/torch_compile_cache/b19ac70222/rank_0_0/backbone for vLLM's torch.compile
(EngineCore_0 pid=1468) INFO 03-26 18:26:18 [backends.py:559] Dynamo bytecode transform time: 2.84 s
(EngineCore_0 pid=1468) INFO 03-26 18:26:20 [backends.py:161] Directly load the compiled graph(s) for dynamic shape from the cache, took 2.385 s
(EngineCore_0 pid=1468) INFO 03-26 18:26:21 [monitor.py:34] torch.compile takes 2.84 s in total
(EngineCore_0 pid=1468) /usr/local/lib/python3.12/dist-packages/torch/utils/cpp_extension.py:2356: UserWarning: TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation. 
(EngineCore_0 pid=1468) If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST'].
(EngineCore_0 pid=1468)   warnings.warn(
(EngineCore_0 pid=1468) INFO 03-26 18:26:21 [gpu_worker.py:276] Available KV cache memory: 65.05 GiB
(EngineCore_0 pid=1468) INFO 03-26 18:26:22 [kv_cache_utils.py:849] GPU KV cache size: 5,684,144 tokens
(EngineCore_0 pid=1468) INFO 03-26 18:26:22 [kv_cache_utils.py:853] Maximum concurrency for 16,384 tokens per request: 346.93x
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 67/67 [00:03<00:00, 19.56it/s](EngineCore_0 pid=1468) INFO 03-26 18:26:25 [gpu_model_runner.py:2708] Graph capturing finished in 4 secs, took 0.39 GiB
(EngineCore_0 pid=1468) INFO 03-26 18:26:25 [core.py:214] init engine (profile, create kv cache, warmup model) took 15.81 seconds
INFO 03-26 18:26:26 [llm.py:298] Supported_tasks: ['generate']
2026-03-26 18:26:26.844 | INFO     | mineru.backend.vlm.vlm_analyze:get_model:218 - get vllm-engine predictor cost: 37.8s
2026-03-26 18:26:35.780 | INFO     | mineru.backend.hybrid.hybrid_analyze:get_batch_ratio:365 - hybrid batch ratio (auto, vram=140GB): 16
Adding requests: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 188/188 [00:05<00:00, 33.11it/s]Processed prompts: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 188/188 [00:28<00:00,  6.58it/s, est. speed input: 9178.14 toks/s, output: 3284.83 toks/s]Adding requests: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 155/155 [00:01<00:00, 94.59it/s]Processed prompts: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 155/155 [00:12<00:00, 12.09it/s, est. speed input: 6065.56 toks/s, output: 2325.30 toks/s]Downloading Model from https://www.modelscope.cn to directory: /root/.cache/modelscope/models/OpenDataLab/PDF-Extract-Kit-1.0
2026-03-26 18:27:35,559 - modelscope - INFO - Target directory already exists, skipping creation.
Downloading Model from https://www.modelscope.cn to directory: /root/.cache/modelscope/models/OpenDataLab/PDF-Extract-Kit-1.0
2026-03-26 18:27:36,943 - modelscope - INFO - Target directory already exists, skipping creation.
Downloading Model from https://www.modelscope.cn to directory: /root/.cache/modelscope/models/OpenDataLab/PDF-Extract-Kit-1.0
2026-03-26 18:27:39,619 - modelscope - INFO - Target directory already exists, skipping creation.
Downloading Model from https://www.modelscope.cn to directory: /root/.cache/modelscope/models/OpenDataLab/PDF-Extract-Kit-1.0
2026-03-26 18:27:41,085 - modelscope - INFO - Target directory already exists, skipping creation.
MFD Predict: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 188/188 [00:07<00:00, 24.88it/s]MFR Predict: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 634/634 [00:04<00:00, 136.88it/s]OCR-det: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 55/55 [00:12<00:00,  4.47it/s]OCR-rec Predict: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2219/2219 [00:04<00:00, 507.17it/s]2026-03-26 18:28:25.281 | INFO     | mineru.cli.common:_process_output:168 - local output dir is /app/output/fa56f7a8-bd36-4140-8cb1-f9f973ce8745/hybrid_auto
ERROR 03-26 18:28:27 [core_client.py:562] Engine core proc EngineCore_0 died unexpectedly, shutting down client.

real    2m45.350s
user    0m0.045s
sys     0m0.032s
(vllm) root@ubuntu:/data/app_workspace/minerU# cd output/
(vllm) root@ubuntu:/data/app_workspace/minerU/output# ls
fa56f7a8-bd36-4140-8cb1-f9f973ce8745  tmp
(vllm) root@ubuntu:/data/app_workspace/minerU/output# cd fa56f7a8-bd36-4140-8cb1-f9f973ce8745/
(vllm) root@ubuntu:/data/app_workspace/minerU/output/fa56f7a8-bd36-4140-8cb1-f9f973ce8745# ls
hybrid_auto
(vllm) root@ubuntu:/data/app_workspace/minerU/output/fa56f7a8-bd36-4140-8cb1-f9f973ce8745# cd hybrid_auto/
(vllm) root@ubuntu:/data/app_workspace/minerU/output/fa56f7a8-bd36-4140-8cb1-f9f973ce8745/hybrid_auto# ls
fa56f7a8-bd36-4140-8cb1-f9f973ce8745_content_list.json     fa56f7a8-bd36-4140-8cb1-f9f973ce8745_layout.pdf  fa56f7a8-bd36-4140-8cb1-f9f973ce8745_middle.json  fa56f7a8-bd36-4140-8cb1-f9f973ce8745_origin.pdf
fa56f7a8-bd36-4140-8cb1-f9f973ce8745_content_list_v2.json  fa56f7a8-bd36-4140-8cb1-f9f973ce8745.md          fa56f7a8-bd36-4140-8cb1-f9f973ce8745_model.json   images
(vllm) root@ubuntu:/data/app_workspace/minerU/output/fa56f7a8-bd36-4140-8cb1-f9f973ce8745/hybrid_auto# vim fa56f7a8-bd36-4140-8cb1-f9f973ce8745.md 
(vllm) root@ubuntu:/data/app_workspace/minerU/output/fa56f7a8-bd36-4140-8cb1-f9f973ce8745/hybrid_auto#