Add voicebench evaluation suite https://github.com/MatthewCYM/VoiceBench

 (#312)

This merge brings Voicebench eval pipeline into the main repository. To run this eval, use 
poetry run python -m ultravox.evaluation.eval --config_path ultravox/evaluation/configs/eval_config_voicebench.yaml

update ultravox model and config for v0.5 (#276)

Add whisper masking (#146)

- Added masking in whisper encoder to ensure consistency in training and inference.
- Simplified release_config.yaml to serve as an example configuration.

Make v0.4 release (#99)

Two changes for training the ultravox v0.4 model:
- Update the speech transcription instructions
- Update the training config file to include more training data

Co-authored-by: Zhongqiang Huang <[email protected]>

Update default config to ultravox_v0.3 (#84)

Add a release configuration to train `ultravox-v0_3` (https://huggingface.co/fixie-ai/ultravox-v0_3). Simply run 
`mcli run -f mcloud.yaml --follow` after following the instructions in README.md