This repository contains the implementation of Whisper-Large ASR model on the Gramvaani and Kathbath dataset for evaluation
- The notebooks used for running the evaluation for the datasets are present here :
Kathbath here
Gramvaani here
Dataset | Time Taken for evaluation | Overall WER |
---|---|---|
Kathbath | ~4.5 hours | 0.2971 |
Gramvaani | ~2 hours | 0.5017 |
RAM : 30GB
GPUs : T4x2