Instructions to use zerodegress/NanoJev-mlx-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use zerodegress/NanoJev-mlx-4bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download zerodegress/NanoJev-mlx-4bit --local-dir NanoJev-mlx-4bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download MANIFEST.json from zerodegress/NanoJev-mlx-4bit: direct link, hf CLI and curl.
- Browser
- Download file 1.03 kB
-
https://huggingface.co/zerodegress/NanoJev-mlx-4bit/resolve/main/MANIFEST.json
- Command line
-
hf download hf://zerodegress/NanoJev-mlx-4bit/MANIFEST.json
-
curl -L -o MANIFEST.json https://huggingface.co/zerodegress/NanoJev-mlx-4bit/resolve/main/MANIFEST.json
1.03 kB
| { | |
| "schema_version": "nanojev-mlx-quant-v1", | |
| "source_checkpoint": "NanoJev", | |
| "source_weights_sha256": "fff62d1412685c1714eaa386acb603f9690371fb3cc8ad03dc41319302597c28", | |
| "bits": 4, | |
| "group_size": 64, | |
| "include_embed": true, | |
| "scales_dtype": "float16", | |
| "note": "backbone linears + embed_tokens in MLX affine 4-bit (group 64); head and norms kept in fp16", | |
| "quantized_params_by_bucket": { | |
| "mlp": 264241152, | |
| "self_attn": 176160768, | |
| "embed_tokens": 155582464 | |
| }, | |
| "relative_max_error_by_bucket": { | |
| "embed_tokens": { | |
| "max": 0.06698610633611679, | |
| "mean": 0.06698610633611679 | |
| }, | |
| "mlp": { | |
| "max": 0.12449125945568085, | |
| "mean": 0.06786140923698743 | |
| }, | |
| "self_attn": { | |
| "max": 0.11749649047851562, | |
| "mean": 0.07736335305630096 | |
| } | |
| }, | |
| "packed_weights_bytes": 335857708, | |
| "packed_weights_sha256": "332d47eb23c81b0370f580e87e0e55ef737ba1baa206a8d300eadf9617307b62", | |
| "dequant_fp32_bytes": 2384507132, | |
| "dequant_fp32_sha256": "2e5bf94be4fa2264e95e4ee3ba763d6240a41faaff6f6cd643f9e1eea29fe291", | |
| "seconds": 87.46 | |
| } | |