mirror of
https://github.com/priyanshujain/torchmlx.git
synced 2026-10-02 11:07:13 +00:00
16 lines
500 B
Markdown
16 lines
500 B
Markdown
# torchmlx
|
|
|
|
torchmlx is a pytorch-shaped compatibility layer that uses mlx on apple silicon and pytorch elsewhere.
|
|
|
|
NOTE: it is experimental and built first for educational use. If you find a bug, please create an issue on the repo.
|
|
|
|
```python
|
|
import torchmlx as torch
|
|
from torchmlx import nn, optim
|
|
|
|
model = nn.Linear(4, 2)
|
|
optimizer = optim.AdamW(model.parameters(), lr=3e-4)
|
|
```
|
|
|
|
see the [tinystories example](examples/tinystories-llm/train.py) and [compatibility details](docs/compatibility.md).
|