大鹏
> 来我们公司吧，配有AED，并且离北医三院近，猝死概率大大减少😂 我们公司也买了AED了！
可以按照这个算一下： https://huggingface.co/docs/transformers/perf_train_gpu_one#anatomy-of-models-memory > 模型进一步训练的化，需要多少卡？8张A100 42G显存够不够？
marker_single '****.pdf' ./ --batch_multiplier 2 Loaded detection model vikp/surya_det2 on device cuda with dtype torch.float16 Loaded detection model vikp/surya_layout2 on device cuda with dtype torch.float16 Loaded reading order model vikp/surya_order...
The problem has been solved, refer to issues #185 , pdf files cannot be placed in subdirectories
> marker_single '****.pdf' ./ --batch_multiplier 2 > refer to issues #209 ， pip install transformers==4.41.2
I had the same problem. Marker used all my memory, so I lost control of the server through ssh. 😭 
I've built a Docker image and run the tool within a container with limited memory, which should work well.
Look this func ([func (s *SSEServer) ServeHTTP(w http.ResponseWriter, r *http.Request) ](https://github.com/mark3labs/mcp-go/blob/71b910bee8fee098e3412177dac8548453eee5c0/server/sse.go#L457)), you could fetch `http.Request` and `http.ResponseWriter` from `echo.Context` and call it