Bohdan Kovalevskyi
Bohdan Kovalevskyi
Yes, that would be great. Can anyone send a PR?
It happens on the latest OpenAI version: 4.59.0 - 4.63.0. I added a minimal reproducible code example.
@tatethurston Thank you for adding support for the app router. How to use it with the app router? Dynamic href in this format is not supported now: ``` href={{ pathname:...
Hi! It’s hard to tell from the logs. I recommend printing each step to identify which one is failing. First, print all trajectories and confirm that rollouts are working correctly....
Hi, @johnson7788! How can we reproduce the issue? Can you share your setup?
Hi @jgbradley1, How are you planning to use it? For most use cases, like LLM-as-a-judge, you can use any clients directly, without relying on the `Model` abstraction. I imagine the...
Hey @sr5434, I see that you unpinned `openpipe-art[backend]` in your notebook. This may cause issues when running on a T4 GPU, as the T4 doesn’t support the latest vLLM version...
Added a function to upload to W&B inference.
Hi, @maks-ivanov ! Yes, it’s on the roadmap, but there’s no ETA at this time.
Hey @casper-hansen, We don’t support MoE models yet. However, since vLLM has added MoE LoRA support, it should now be possible to enable this on ART.