awaaz-se-alfaaz/whisper-turbo-v3-urdu
Adapter vs openai/whisper-large-v3-turbo. awaaz-se-alfaaz/whisper-turbo-v3-urdu is a LoRA adapter, not full weights: rank 32, on k_proj, out_proj, q_proj, v_proj in 32 layers (160 adapted matrices). It modifies openai/whisper-large-v3-turbo at load time.