Awni Hannun comments

Results 1014 comments of


                                            Awni Hannun

Example of a Convolutional Variational Autoencoder (CVAE) on MNIST

I left a couple of high level questions, could you check? I will do a more detailed review after hearing from you. FYI usually with the examples reivewing for nits...

Example of a Convolutional Variational Autoencoder (CVAE) on MNIST

@menzHSE is this ready for review?

Example of a Convolutional Variational Autoencoder (CVAE) on MNIST

@menzHSE this is really nice! I took a deeper look today and simplifies / reorganized the example to be more self-contained and in the style of our other small examples....

Example of a Convolutional Variational Autoencoder (CVAE) on MNIST

@menzHSE you might be interested to check the updated README log. I notice the throughput I get on my M1 Max is much much better than the numbers you had...

whisper: No such file or directory: 'mlx_models/config.json'

Maybe you are using an outdated whisper? The `transcribe` function takes an `path_or_hf_repo` as the second argument which you should set to `mlx_models/large` if that's where you stored the model.

whisper: No such file or directory: 'mlx_models/config.json'

I think the README for this example needs a little work, I will update it.

whisper: No such file or directory: 'mlx_models/config.json'

For future reference `help(whisper.transcribe)` will tell you about the arguments it can accept 😄

Lora: generating from local fused model fails

Could you share the training command you used?

Lora: generating from local fused model fails

I think the problem here is that our `lora.py` script assumes you are sending it an "unfused" model so it is loading the adapters into a fused model and hence...

lora README.md `--num-tokens` should be ` --max-tokens`

Fix in #282, thanks!