File size: 763 Bytes
708ee6a 9b5451a 708ee6a 9b5451a 708ee6a 9b5451a 708ee6a 9b5451a 708ee6a 9b5451a 128f5a8 9b5451a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 |
---
library_name: transformers
license: cc-by-4.0
base_model:
- HKUSTAudio/Llasa-1B
---
## Model Description
This is a 4bit GPTQ quantization of Llasa-1B by the HKUSTAudio team.
I tested using a script written by GitHub user nivibilla, linked below.
The tests were successful, but the quality of the generated voice is often unusable.
In case you don't believe me, I'll leave this model up here so you can test it yourself.
## Model Sources
<!-- Provide the basic links for the model. -->
- **Repository:** HKUSTAudio/Llasa-1B
- **Paper:** LLaSA: Scaling Train-Time and Inference-Time Compute for LLaMA-based Speech Synthesis (Coming soon)
- **Test Script:** https://github.com/slives-lab/local-llasa-tts_voice/blob/main/llasa_vllm_longtext_inference.ipynb |