Markus / Mark

Fix incorrect order of robot for languages `alias` and `en`

1

Refering to https://github.com/carpedm20/emoji/pull/102/files#diff-48bad5f533ed8ca39c7260b339db68d37fa13b547ed4f14ccfaafd5cb99b97d8 I noticed that in the current version, the alias name is ":robot_face:" and not ":robot:" as was addressed in #102 This PR sets 🤖 to ":robot_face:" for...

Fix issue 3355 assertion error

2

## Summary This PR fixes issue #3355, which caused an `AssertionError` when running evaluations with OpenAI-compatible endpoints when API responses have `"content": null`. This happens with vllm if using a...

Eval breaking on openai-compatible model run - AssertionError.

2

I'm trying to run evaluation using local openai-compatible endpoint with tokenization on the server side. Both these examples run vllm, with the second being behind llama-swap. Not sure what the...

[MODEL EVALUATION REQUEST] GLM-4.6

1

### Model ID zai-org/GLM-4.6 ### Evaluation languages - [x] Baltic languages (Latvian, Lithuanian) - [x] Finnic languages (Estonian, Finnish) - [x] Hellenic languages (Greek) - [x] Romance languages (French, Italian,...

### Model ID moonshotai/Kimi-K2-Thinking ### Evaluation languages - [x] Baltic languages (Latvian, Lithuanian) - [x] Finnic languages (Estonian, Finnish) - [x] Hellenic languages (Greek) - [x] Romance languages (French, Italian,...

Fix `ImportError` and `NameError` in `examples/run_simple_mcore_train_loop.py`

8

# What does this PR do ? This PR fixes two breaking bugs (`ImportError` and `NameError`) in `examples/run_simple_train_loop.py` that were introduced when type hints were added. Resolves #1979 ### Pre-checks...

https://github.com/vibevoice-community/VibeVoice

Async-GRPO: Decouple rollout generation and reward computation/model update for faster training

1

### Method description Im not sure if it is better as a feature request or trainer addition, but i think it would be great if trl supported async-grpo. It is...