Failed for 5 of 120 audio samples

[lokesh12345678910]

I ran CrisperWhisper on a folder of 120 audio files. I wrote a for-loop, transcribing each file one at a time (i.e. batch size of 1). There are 5 audio files where I saw the following error message: The expanded size of the tensor (4) must match the existing size (5) at non-singleton dimension 0. Target sizes: [4]. Tensor sizes: [5]

I can confirm the audio files (which I can't share for PHI reasons) aren't silent. 4 of 5 had significant background noise. Please let me know if there is a fix for this.

[a-rogalska]

This solution worked for me #12 (comment)
You have to modify their transformers fork. Hope it helps