Andreas Jansson
Andreas Jansson
Hi @xinntao! Really great work on ESRGAN, I still use it regularly for image upscaling. This pull request makes it possible to run your model inside a Docker environment, which...
Cog currently only works on Mac and Linux. We should support Windows as well. ## User data - https://canary.discord.com/channels/775512803439280149/882336879343706122/963166127213674556 ## See also - #363 - https://www.docker.com/blog/wsl-2-gpu-support-for-docker-desktop-on-nvidia-gpus/ (This is edited by...
The loop here takes a full CPU while async polling for messages from Redis: https://github.com/replicate/cog/blob/main/python/cog/server/redis_queue.py#L223-L224 Can we do this with a blocking pop instead?
This is a follow-up from https://github.com/replicate/cog/issues/404, where I've tried to distill the discussion into separate issues and prioritize issues into required and future work. ## Required now - [x] https://github.com/replicate/cog/issues/473...
When nvidia-docker isn't installed and you're trying to run a GPU model, Cog blows up with a big unparseable stack trace. We should check for nvidia-docker and show instructions for...
Since code and weights are stored in the same Docker layer, making small changes can result in really long push times when model weights are in the hundreds of megabytes...
It's currently impossible to mount directories outside of the working directory when doing `cog run`. It's common to keep datasets in separate folders than the working directory when training models,...
If the user hasn't [set up Docker to work without sudo](https://docs.docker.com/engine/install/linux-postinstall/#manage-docker-as-a-non-root-user), `cog push` will fail unless you use `sudo cog push`. However, `cog login` still works so the user will...
Cog has reference documentation for the Python API but no reference for the command line interface. The tutorials explain parts of the CLI, but it would be great to have...
The original [AMQP proposal](https://github.com/replicate/cog/issues/404) sends metrics (model start up time, request processing time, and system metrics like CPU, RAM, and GPU memory used, etc.) to the response queue in the...