Awni Hannun
Awni Hannun
We have a discord link here: https://github.com/ml-explore/mlx/discussions/733 You shouldn't need to implement weight norm in C++. That can all be done in Python using existing ops.
Looks nice! I requested a review from @angeloskath as I'm not that familiar with this example.
That is definitely not expected. The adapter file should have a `npz` extension (our `savez` automatically adds this even if you don't specify it). Could you double check that the...
> @awni perhaps we can leave this as T5 and then make an attempt at swapping to Llama in a new PR? Yea that sounds like a great plan to...
Sorry for the delay!! I will review and get this in early next week
Strange.. I just ran it no problem on an M1 Max. Is it possible the weights are ill formatted? Maybe the conversion didn't work?
If this is still an issue, please reopen.
@Diniden we've updated MLX and fixed a bug with loading from NumPy so try `pip install --upgrade mlx` and see if that helps. I would also double check the weights...
Hmm, I wonder why it's so slow. I assume you are using the Metal back-end? ``` python -c "import mlx.core as mx; print(mx.default_device())" ``` Should give `Device(gpu, 0)`. If it's...
I have a 32GB machine and the Mistral example runs pretty quickly. It seems like somewhere between 16 and 32 is the cutoff right now to get good perf, but...