CAYTU/whosper-large
Adapter vs openai/whisper-large-v2. CAYTU/whosper-large is a LoRA adapter, not full weights: rank 32, on q_proj, v_proj in 32 layers (192 adapted matrices). It modifies openai/whisper-large-v2 at load time.
Adapter vs openai/whisper-large-v2. CAYTU/whosper-large is a LoRA adapter, not full weights: rank 32, on q_proj, v_proj in 32 layers (192 adapted matrices). It modifies openai/whisper-large-v2 at load time.