Apple Son
Reputation: 1
OpenAI whisper medium WER does not decay for low resource language
Does anyone know why the WER does not decay? I'm trying to finetune the OpenAI Whisper medium model for low resource language.
I'm using the following parameters:
per_device_train_batch_size="32"per_device_eval_batch_size="16"learning_rate="1e-5"
Upvotes: 0
Views: 137
Answers (0)
Related Questions
- How to successfully transcribe audio files using Whisper for OpenAI in Python?
- Azure openai speech to text -Whisper "code":"404","message": "Resource not found"
- OpenAI Whisper API error: "AttributeError: module 'openai' has no attribute 'Audio'"
- Request for help Getting Whisper Api (NextJs)- Could not parse multipart form error