replicate / replicate/cog

CUDA version changes when upgrading Cog

Open
#330 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement go type/bug
Dominant language
Go
Stars
9.5k
Forks
696
Avg merge
7d 19h
Merged PRs (30d)
2

Description

If you upgrade Cog and the latest known CUDA version changes in Cog, then Cog will start using a more recent version of CUDA for your version.

As a user, I would expect my Cog environment to remain consistent. This is unexpected behavior, and might cause breakage.

A couple of potential solutions:

  1. We could encourage users to specify a cuda version.
  2. We could lock the CUDA version somehow, akin to package-lock.json.

See https://github.com/replicate/cog/pull/329#discussion_r746731800

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reading the discussion linked from pull request #329 and tracing how Cog selects its known CUDA version during an upgrade. Done should mean the CUDA version remains consistent across Cog upgrades, either through an explicit user setting or a lock mechanism, with the chosen behavior documented and tested.

Written by the indexing model from the issue text.

Assessment

Tech stack
docker, go
Domain
devops, machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
28/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.