SPEAK-ASR/whisper-si-exp-1

Updated 18 days ago • 94

Note Used dataset: 1/3 OpenSLR Dataset (~60h) Lora Projection Layers to: Query (Q) and Value (V) layers

SPEAK-ASR/whisper-si-exp-2

Updated 18 days ago • 35

Note Use `pranay-j/whisper-small-hindi` as base model

SPEAK-ASR/whisper-si-exp-3

Updated 18 days ago • 27

Note Used dataset: Full OpenSLR Dataset (~180h)

SPEAK-ASR/whisper-si-exp-4

Updated 10 days ago • 30

Note Used dataset: OpenSLR + YouTube Dataset (~185h)

SPEAK-ASR/whisper-si-exp-5

Updated 18 days ago • 46

Note Used WER to choose the best model instead of loss Used dataset: OpenSLR + YouTube Dataset (~185h) Changed Lora Projection Layers to: Q, V, Key (K), and Output (O) layers

SPEAK-ASR/whisper-si-exp-6

Updated 10 days ago • 76

Note Revert Lora Projection Layers to: Q, V layers Change num_epoch from 5 to 15