Instructions to use mlx-community/Mistral-Nemo-Instruct-2407-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use mlx-community/Mistral-Nemo-Instruct-2407-4bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir Mistral-Nemo-Instruct-2407-4bit mlx-community/Mistral-Nemo-Instruct-2407-4bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
ValueError
#2
by ceoofcapybaras - opened
Expected shape (5120, 640) but received shape (4096, 640) for parameter model.layers.0.self_attn.q_proj.weight
after manually editing hidden state in config.json:
Expected shape (131072, 512) but received shape (131072, 640) for parameter model.embed_tokens.weight
same here : ValueError: Expected shape (5120, 640) but received shape (4096, 640) for parameter model.layers.0.self_attn.q_proj.weight
Please install mlx-lm from source, the PR is not merged yet: