Awni Hannun
Awni Hannun
Of course, sounds good to me! Thank you!
@Blaizzy just to understand your concern a bit. This PR has changed to give another option to the user to basically force use the default template if the model doesn't...
Really nice! When you say RoBERTA is not working is it giving bad output or just crashing?
Diffusers MLX back-end? 👀 That would be very cool
It's not a bug.. at the risk of being redundant, the last dimension of the matrix has to be divisible by the quantization group size. For the size 4304 there...
You can use `class_predicate` for that. Just put the condition you want in the predicate. For example if you are trying to skip weights of a certain shape: `class_predicate =...
@madrob can we close this or are you still investigating?
Did you manually make a command encoder from the command buffer? MLX manages an active command encoder so you should not make it directly. Rather call the `device.get_command_encoder()` to get...
@nicolov are you planning to come back to this?
That's pretty awesome that it's faster. A rare example of mixing CPU / GPU speeding things up! I'm not sure what to do with it. On the one-hand, it's a...