I used the GGML version of Whisper and I had to revert back to the PyTorch version released by OpenAI. The GGML version simply didn't work well even for the sam

I used the GGML version of Whisper and I had to revert back to the PyTorch version released by OpenAI. The GGML version simply didn't work well even for the same model. I am assuming it has to do with the quantization.