Any-to-Any
Transformers
Safetensors
English
Korean
raon
feature-extraction
speech
audio
multimodal
mixture-of-experts
text-to-speech
automatic-speech-recognition
custom_code
Instructions to use KRAFTON/A.X-K2-Raon-Speech-21B-A3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use KRAFTON/A.X-K2-Raon-Speech-21B-A3B with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("KRAFTON/A.X-K2-Raon-Speech-21B-A3B", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
| { | |
| "checkpoint": { | |
| "training_stage": "stage_6_tts_dpo", | |
| "tensor_count": 19632, | |
| "total_parameters": 21197920833, | |
| "active_parameters": 3456014913, | |
| "tensor_bytes": 42429527298, | |
| "shards": 9 | |
| }, | |
| "components": { | |
| "text_backbone": "AXK1", | |
| "speech_encoder_integrated": true, | |
| "custom_audio_tokenizer_integrated": true, | |
| "audio_tokenizer_tensors": 350, | |
| "audio_tokenizer_patched_shard": "model-00009-of-00009.safetensors" | |
| } | |
| } | |