Instructions to use KitKatL/Hojo-ASR-Multi-V1-MLX-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use KitKatL/Hojo-ASR-Multi-V1-MLX-4bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download KitKatL/Hojo-ASR-Multi-V1-MLX-4bit --local-dir Hojo-ASR-Multi-V1-MLX-4bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download config.json from KitKatL/Hojo-ASR-Multi-V1-MLX-4bit: direct link, hf CLI and curl.
- Browser
- Download file 1.53 kB
-
https://huggingface.co/KitKatL/Hojo-ASR-Multi-V1-MLX-4bit/resolve/main/config.json
- Command line
-
hf download hf://KitKatL/Hojo-ASR-Multi-V1-MLX-4bit/config.json
-
curl -L -o config.json https://huggingface.co/KitKatL/Hojo-ASR-Multi-V1-MLX-4bit/resolve/main/config.json
1.53 kB
| { | |
| "model_type": "hojo_asr_mlx", | |
| "source_repository": "HojoAI/Hojo-ASR-Multi-V1", | |
| "source_revision": "aa1d0dfc369deb3b3235572fa151acd7b498481f", | |
| "source_sha256": "c046a5a26c142f5aded7331d36892293a025f579e342a27861e80509afaca88d", | |
| "quantization": { | |
| "bits": 4, | |
| "group_size": 64, | |
| "mode": "affine", | |
| "scope": "decoder_and_frontend_linear" | |
| }, | |
| "frontend_dtype": "float16", | |
| "audio_encoder": { | |
| "d_model": 1280, | |
| "encoder_attention_heads": 20, | |
| "encoder_ffn_dim": 5120, | |
| "encoder_layers": 32, | |
| "num_mel_bins": 128, | |
| "max_source_positions": 1500, | |
| "output_dim": 2048, | |
| "downsample_hidden_size": 480, | |
| "conv_chunksize": 500, | |
| "n_window": 1500, | |
| "n_window_infer": 3000, | |
| "layer_norm_eps": 1e-05, | |
| "conv_out_input_dim": 7680 | |
| }, | |
| "adapter": { | |
| "input_size": 2048, | |
| "output_size": 2560, | |
| "linear_units": 640, | |
| "num_blocks": 2, | |
| "attention_heads": 4, | |
| "convolution_kernel_size": 15, | |
| "positional_encoding_max_length": 5000 | |
| }, | |
| "text_decoder": { | |
| "vocab_size": 151670, | |
| "hidden_size": 2560, | |
| "num_hidden_layers": 36, | |
| "num_attention_heads": 32, | |
| "num_key_value_heads": 8, | |
| "head_dim": 128, | |
| "intermediate_size": 9728, | |
| "max_position_embeddings": 262144, | |
| "rms_norm_eps": 1e-06, | |
| "rope_theta": 5000000, | |
| "tie_word_embeddings": true | |
| }, | |
| "generation": { | |
| "bos_token_id": 151644, | |
| "eos_token_id": 151645, | |
| "max_new_tokens": 200, | |
| "num_beams": 4, | |
| "repetition_penalty": 2.0, | |
| "length_penalty": 1.0 | |
| } | |
| } | |