Any way to save the output markdown into folder ?

[drmetro09]

any way to save the output into folder ?

[xiaoyao9184]

You should directly reference the marker project

or use this project to run as a gradio service and use gradio_client.

There will be an API interface at the bottom of each gradio project.

[drmetro09]

You should directly reference the marker project

or use this project to run as a gradio service and use gradio_client.

There will be an API interface at the bottom of each gradio project.

Is it possible to mount the output folder ?

[drmetro09]

I just want it to save the markdown file to a folder , how can i do it ?

[xiaoyao9184]

you need to call convert_single.py or convert.py

[drmetro09]

you need to call convert_single.py or convert.py

How to do that from gradio ui ? Can you please explain?

[xiaoyao9184]

You can run any commands provided by Marker. However, unless it is a one-time task, it is not recommended to use command scripts for execution, as the model will be reloaded each time. Below is an example of using an offline model to process batch PDF file inputs and output them as MD files.

services:
  marker_pdf_convert:
    image: xiaoyao9184/marker:1.0.0
    container_name: marker_pdf_convert
    command: marker /pdf_in --output_dir /md_out
    environment:
      - TORCH_DEVICE=cuda
      - HF_HUB_OFFLINE=true
      - DETECTOR_MODEL_CHECKPOINT=/root/.cache/huggingface/hub/models--vikp--surya_det3/snapshots/467ee9ec33e6e6c5f73e57dbc1415b14032f5b95
      - LAYOUT_MODEL_CHECKPOINT=/root/.cache/huggingface/hub/models--datalab-to--surya_layout0/snapshots/421ac206a400227ea714d47a405e53ce74374957
      - RECOGNITION_MODEL_CHECKPOINT=/root/.cache/huggingface/hub/models--vikp--surya_rec2/snapshots/6611509b2c3a32c141703ce19adc899d9d0abf41
      - TABLE_REC_MODEL_CHECKPOINT=/root/.cache/huggingface/hub/models--vikp--surya_tablerec/snapshots/8bca165f81e9cee5fb382413eb23175079917d14
      - TEXIFY_MODEL_NAME=/root/.cache/huggingface/hub/models--vikp--texify/snapshots/ce49c1fe10842e78b8be61f9e762b85ac952807d
    volumes:
      - ./../../cache:/root/.cache
      - ./../../pdf_in:/pdf_in
      - ./../../md_out:/md_out
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              capabilities: [ gpu ]

[drmetro09]

The md output generated is not accurate it has all spelling errors

[xiaoyao9184]

You can try to update the latest version. If the marker project itself reports an error, I can't help you.

[drmetro09]

You can try to update the latest version. If the marker project itself reports an error, I can't help you.

Im using your latest docker image , How to update the environment variables to latest ?