olmocr icon indicating copy to clipboard operation
olmocr copied to clipboard

[vLLM] Response did not finish with reason code 'stop', cannot use this response

Open hcnhcn012 opened this issue 1 month ago • 1 comments

🐛 Describe the bug

version: 0.4.4 vllm verison: v0.10.1.1 installation method: uv add olmocr command: python -m olmocr.pipeline ./olm --server http://localhost:8114/v1 --markdown --pdfs ./test/Extract[955-968].pdf --model vllm other info: No error on vLLM endpoint logs, seems like endless generating

Versions

0.4.4

logs

2025-11-13 09:00:47,969 - __main__ - WARNING - ValueError on attempt 3 for ./test/Extract[6].pdf-1: <class 'ValueError'> - Response did not finish with reason code 'stop', cannot use this response
2025-11-13 09:00:54,982 - __main__ - INFO - Queue remaining: 0
2025-11-13 09:00:54,982 - __main__ - INFO - 
Metric Name                        Lifetime (tokens/sec)     Recently (tokens/sec)
----------------------------------------------------------------------------------
2025-11-13 09:00:54,982 - __main__ - INFO - 
Worker ID | started
----------+--------
0         | 1      
2025-11-13 09:01:04,990 - __main__ - INFO - Queue remaining: 0
2025-11-13 09:01:04,990 - __main__ - INFO - 
Metric Name                        Lifetime (tokens/sec)     Recently (tokens/sec)
----------------------------------------------------------------------------------

a ValueError problem ?

hcnhcn012 avatar Nov 13 '25 09:11 hcnhcn012

It's probably a PDF that might have text which is too long or otherwise not parseable by the model. Can you send the PDF you are using?

jakep-allenai avatar Nov 14 '25 17:11 jakep-allenai