LAION-AI / LAION-AI/CLIP_benchmark

Eval custom open_clip architecture – Add model config

Open
#144 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
814
Forks
103
PR merge metrics
No merged PRs in 30d

Description

What should be the appropriate way of adding a new model architecture without having to clone the repo? With open_clip we can write the model config json and add it using use add_model_config from open_clip.factory. However, doing this before runing the clip_benchmark cli has no effect, since this opens a new process that doesn't see the new architecture added.

Simlarly, what is the correct way of passing my architecture and my pretrained weights to the cli?

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 open_clip.factory.add_model_config and the clip_benchmark CLI entry point, focusing on how the separate process discovers model configurations. Determine how a custom architecture and pretrained weights should be supplied to the CLI without cloning the repository. Done means the issue documents or enables a supported workflow for evaluating the custom model.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
cli, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.