Johnson Zhou

icon location Melbourne AU icon location Health ⋂ Tech. PhD candidate (deep learning in neuroscience) at the University of Melbourne.

Results 6 comments of Johnson Zhou

Adjusted baseline upwards

Hello! Yes sure thing. If there's a better way to adjust the baseline for web browsers, please let me know as well.

Segmentation fault during training

Hi @angeloskath, @awni pointed me to this issue as I was having a similar seg fault issue. My seg fault is also intermittent and difficult to reproduce but it does...

Segmentation fault during training

Here is some code that can reproduce the above error semi-regularly on my machine: ```python import mlx.core as mx import mlx.nn as nn import mlx.optimizers as optim import random from...

Segmentation fault during training

Were you able to reproduce it on your end?

Segmentation fault during training

If it will show, it will show within the first few epochs, most frequently around 30-40% of the batch in the first epoch. If not, it will run fine forever....

Segmentation fault during training

Try these weights. It seg faults fairly consistently now within the first half of the first epoch with these weights for some reason. [test_weights.safetensors.zip](https://github.com/user-attachments/files/16076322/test_weights.safetensors.zip)