Repo for the Whisper vdi transcription service
beam_size = trial.suggest_int("beam_size", 1, 10)
temperature = trial.suggest_float("temp", 0.1, 1.0)
batch_size = trial.suggest_int("batch_size", 1, 32)
best_of = trial.suggest_int("best_of", 1, 10)
patience = trial.suggest_float("patience", 0.1, 1.0)
chunk_length = trial.suggest_int("chunk_length", 1, 10)
compression_ratio_threshold = trial.suggest_float("compression_ratio_threshold", 0.1, 1.0)
repetition_penalty = trial.suggest_float("repetition_penalty", 0.1, 1.0)
length_penalty = trial.suggest_float("length_penalty", 0.1, 1.0)
no_speech_threshold = trial.suggest_float("no_speech_threshold", 0.1, 1.0)
and prompting
A folder in the CeR drive contains benchmark corpus and reference transcripts used for testing. Download and place them in /benchmark
.