infiniflow
[Question]: Parsing batches of files stuck at 50%
Describe your problem
When a batch of documents are parsed all of them are stalled at 50% forever. Manual restart to reparse one by one again (and sometime 3 times) is necessary. In addition pls kindly add support for AMD GPU acceleration. Ragflow appears to only support Nvidia by using docker-compose-gpu.yml. I cannot find corresponding yml file for AMD. Am I missing something? Much appreciated for your assistance. I installed ragflow docker locally with Ollama deploying "nomic-embed-text "in Ubuntu 24.04.
- driver: nvidia Maybe change the driver in docker compose yaml? You need to start multiple task executor to enlarge the throughput of parsing, please enter the container and check this out.
The solution to this problem is to start multiple instances of the task_executor.py program, correct?
- driver: nvidia Maybe change the driver in docker compose yaml? You need to start multiple task executor to enlarge the throughput of parsing, please enter the container and check this out.
Sorry. I changed the driver to amdgpu. Docker compose did not fire up. Out of ideas I even attempted to add these and that did not work either: devices: - /dev/kfd:/dev/kfd - /dev/dri:/dev/dri environment: - 'HSA_OVERRIDE_GFX_VERSION=${HSA_OVERRIDE_GFX_VERSION-11.0.0}' Sorry I have no idea how to run multiple task_executor.py (thanks to Darry's help) in a container.
- driver: nvidia Maybe change the driver in docker compose yaml? You need to start multiple task executor to enlarge the throughput of parsing, please enter the container and check this out.
Sorry. I changed the driver to amdgpu. Docker compose did not fire up. Out of ideas I even attempted to add these and that did not work either: devices: - /dev/kfd:/dev/kfd - /dev/dri:/dev/dri environment: - 'HSA_OVERRIDE_GFX_VERSION=${HSA_OVERRIDE_GFX_VERSION-11.0.0}' Sorry I have no idea how to run multiple task_executor.py (thanks to Darry's help) in a container.
We're shrinking the RAGFlow docker image size by removing some python modules and some of the built-in modeling. This will be finished before next release. This might make RAGFlow run in your environment.
