openai/whisper-large-v3-turbo
Full fine-tune (fewer layers) vs openai/whisper-large-v3. openai/whisper-large-v3-turbo matches part of openai/whisper-large-v3 (587 of its 1259 tensors; the rest of openai/whisper-large-v3 has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 2/3, Attention 280/280, Norms 156/156, MLP 144/144, other 4/4
95 derivatives checked
- handy-computer/whisper-large-v3-turbo-gguf: Not comparable yet
- naazimsnh02/whisper-large-v3-turbo-ar-quran: LoRA-style
- sbintuitions/kana-whisper: Full fine-tune
- ivrit-ai/whisper-large-v3-turbo: Full fine-tune
- shyngys879/kazakh-whisper-large-v3-turbo: LoRA-style
- typhoon-ai/typhoon-whisper-turbo: Full fine-tune
- mlx-community/whisper-large-v3-turbo-4bit: Quantization
- BELLE-2/Belle-whisper-large-v3-turbo-zh: Full fine-tune
- mesolitica/Malaysian-whisper-large-v3-turbo-v3: Full fine-tune
- TheStageAI/thewhisper-large-v3-turbo: Full fine-tune
- RedHatAI/whisper-large-v3-turbo-FP8-dynamic: Quantization
- Oriserve/Whisper-Hindi2Hinglish-Apex: Partial fine-tune
- oxide-lab/whisper-large-v3-turbo-GGUF: Quantization
- efficient-speech/lite-whisper-large-v3-turbo-acc: Partial fine-tune
- marianbasti/whisper-large-v3-turbo-latam: Full fine-tune
- LogosAccessibleExpression/logos-turbo-lora-d43df745: Adapter
- bond005/whisper-podlodka-turbo: Full fine-tune
- vhdm/whisper-large-fa-v1: Full fine-tune
- anaszil/whisper-large-v3-turbo-darija: Adapter
- abilmansplus/whisper-turbo-ksc2: Full fine-tune
- Flurin17/whisper-large-v3-turbo-swiss-german: Full fine-tune
- sachmatkris/agirdi: Full fine-tune
- RedHatAI/whisper-large-v3-turbo-quantized.w4a16: Quantization
- mlx-community/whisper-large-v3-turbo-8bit: Quantization
- ele-sage/whisper-large-v3-turbo-fr-quebecois: Full fine-tune
- efficient-speech/lite-whisper-large-v3-turbo: Partial fine-tune
- ghost613/whisper-large-v3-turbo-korean: Full fine-tune
- LocalDoc/azerbaijani-whisper-turbo: LoRA-style
- efficient-speech/lite-whisper-large-v3-turbo-fast: Partial fine-tune
- oddadmix/whisper-large-v3-turbo-arabic-dialectal-v2: Full fine-tune
51 other repos hold byte-for-byte identical weight files.