| --- |
| license: other |
| license_name: openmdw-1.1 |
| license_link: https://huggingface.co/DeepBeepMeep/Nemotron/blob/main/nemotron_transcribe/LICENSE_NEMOTRON_TRANSCRIBE.txt |
| base_model: nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B |
| tags: |
| - wangp |
| - automatic-speech-recognition |
| - speaker-diarization |
| --- |
| |
| # Nemotron 3.5 Transcribe 30B-A4B for WanGP |
|
|
| Repackaged weights of [nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B](https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B) for [WanGP](https://github.com/deepbeepmeep/Wan2GP): transcription with speaker attribution, word timestamps, subtitles and translation. WanGP downloads these files itself. |
|
|
| | File | Content | |
| |---|---| |
| | `NemotronTranscribe_30B_A4B_lm_bf16.safetensors` | Language model without its routed experts, BF16 | |
| | `NemotronTranscribe_30B_A4B_lm_int8_convrot.safetensors` | Same, INT8 ConvRot | |
| | `NemotronTranscribe_30B_A4B_experts_int8_convrot.safetensors` | Routed experts, INT8 ConvRot (default variant) | |
| | `NemotronTranscribe_30B_A4B_experts_int4.safetensors` | Routed experts, 4-bit (Q4 Experts variant) | |
| | `nemotron_transcribe/` | Audio encoder, DFlash2 draft model, tokenizer, configuration, licenses and notices | |
|
|
| The model materials are licensed by NVIDIA under the OpenMDW License Agreement, version 1.1: see `nemotron_transcribe/LICENSE_NEMOTRON_TRANSCRIBE.txt` and `nemotron_transcribe/THIRD_PARTY_NOTICES.md`. |
|
|
