README tells that contributions are welcome, but in practice I see some obstacles for it. For example:
- My PR with Qwen3-VL implementation has been submitted 1.5 months ago, but so far even CI/CD workflow has not been approved.
- Similar PR for multimodal support for Gemma 3 took 6 months to merge.
- For the past 3 weeks, all merged PRs come either from copybara-service, or from Google employees.
- Even though Tunix runs perfectly well on GPU, the development setup for it doesn't exist (40+ failing tests, strict dependency on libtpu, etc.)
I totally understand that reviewing external contributions may not be aligned with the company goals at the moment, but can you help me understand what is the best way to extend Tunix ecosystem:
- Committing code to this repo as a single source of truth, or
- Spinning up a new independent repository with all the extensions?
README tells that contributions are welcome, but in practice I see some obstacles for it. For example:
I totally understand that reviewing external contributions may not be aligned with the company goals at the moment, but can you help me understand what is the best way to extend Tunix ecosystem: