Note that mlx will not load this model for fine tuning
#5
by
LeeHarrold
- opened
README.md
CHANGED
@@ -35,3 +35,7 @@ python mlx-examples/llms/llama/llama.py --prompt "My name is " --model-path Llam
|
|
35 |
|
36 |
Please, refer to the [original model card](https://huggingface.co/meta-llama/Llama-2-7b-chat) for details on Llama 2.
|
37 |
|
|
|
|
|
|
|
|
|
|
35 |
|
36 |
Please, refer to the [original model card](https://huggingface.co/meta-llama/Llama-2-7b-chat) for details on Llama 2.
|
37 |
|
38 |
+
# Beware
|
39 |
+
|
40 |
+
This model's weights were converted to .npz, the old mlx format. You can use it to generate text but you cannot load it for fine tuning. [npz was adapted to the safetensor format](https://github.com/ml-explore/mlx-examples/issues/480#issuecomment-1960694170).
|
41 |
+
If you are looking for a quantized model, you can either use `mlx_lm.convert` on a hugging face model or find one that someone has already quantized and uploaded - but make sure to look in the Files and Versions tab for safetensors!
|