Merge pull request #142 from bofenghuang/patch-1

Merge pull request #161 from peregilk/patch-4

Merge pull request #160 from peregilk/patch-3

Merge pull request #159 from peregilk/patch-2

Merge pull request #163 from huggingface/whisper-turbo

add turbo

Another Flash attention typo

Flash attention typo

Update run_pseudo_labelling.py

Merge pull request #156 from Rolv-Arild/patch-1

Merge pull request #155 from hannan72/fix_sorted_best_checkpoints_for_wer_below_ten_percent

Fix snapshot_download

Fix sorted_best_checkpoints method in run_distillation to resolve issue for wer below 10%

Merge pull request #141 from eustlb/save-n-best

base dir model is the best one, end of training weights in dedicated dir