Instructions to use labhamlet/speech-jepa-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use labhamlet/speech-jepa-base with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="labhamlet/speech-jepa-base", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("labhamlet/speech-jepa-base", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Links to fine-tuned CTC checkpoints lead to missing repos
Hi, thanks for releasing Speech-JEPA! The model card links to fine-tuned CTC checkpoints like, speech-jepa-base-asr-100h, but it don’t appear to be public yet. Are there plans to release them? I’d like to benchmark them for edge ASR.
Hey Shubham,
We are working on releasing the CTC fine-tuned checkpoints and finalizing the paper details this week. They will be available the latest next Monday!
Meanwhile, you could try to fine-tune the base model using the ASR fine-tuning scripts we've provided in the repo.
Kind Regards,
Goksenin Yuksel
They are live now, I'd recommend using the 100hour one. We are also working on the full fine-tune of the model on the 960hours of librispeech. When it becomes available I ll let u know
Hi Goksenin,
Thank you so much for the information. I'll use the 100hrs one and also I'll wait for the 960hrs fine tuned model :)