PraxySante/Qwen3-0.6B-SFT-ASR-Correction-FR-v2
PraxySante/Qwen3-0.6B-SFT-ASR-Correction-FR-v2 is a 0.6 billion parameter Qwen3-based language model developed by PraxySante. This model is a fine-tuned version of PraxySante/Qwen3-0.6B-ASR-PostTrain-Medical-FR, specifically optimized for ASR correction in French. It leverages Supervised Fine-Tuning (SFT) to enhance its performance in correcting speech-to-text errors, making it suitable for applications requiring accurate French transcription refinement.
Loading preview...
Model Overview
PraxySante/Qwen3-0.6B-SFT-ASR-Correction-FR-v2 is a specialized language model built upon the Qwen3 architecture, developed by PraxySante. It is a fine-tuned iteration of the PraxySante/Qwen3-0.6B-ASR-PostTrain-Medical-FR model, with a primary focus on improving Automatic Speech Recognition (ASR) correction for French language inputs.
Key Capabilities
- French ASR Correction: Specifically trained to refine and correct errors in French speech-to-text outputs.
- Supervised Fine-Tuning (SFT): Utilizes SFT for targeted performance enhancement in its domain.
- Qwen3 Base: Benefits from the underlying capabilities of the Qwen3 model family.
Training Details
The model was trained using the TRL library, a framework for Transformer Reinforcement Learning. This SFT process aims to adapt the model for its specific ASR correction task. Developers can quickly integrate and test the model using the provided transformers pipeline example for text generation.
Use Cases
This model is particularly well-suited for applications requiring high accuracy in correcting French ASR transcripts, such as:
- Post-processing of speech-to-text outputs in French.
- Improving the reliability of voice assistants or dictation software for French speakers.
- Enhancing transcription services where precision in French is critical.