-
### 🐛 Describe the bug
1. Policy (related): We need to agree on how to handle refs/decomps like `binary_cross_entropy` where the ATen op and the Python frontend have different number/type of argument…
-
1. Investigate how hard it would be to integrate https://github.com/pytorch/pytorch/blob/1c1c68b1e31351030843c773e73fab736cbfaf1b/torch/_meta_registrations.py formulas to compute `is_dynamic` in LTC
…
-
### 🚀 The feature, motivation and pitch
This RFC proposes an enhancement to torch.compile to improve its backend agnosticism. The goal is to enable a more seamless experience for users working with d…
-
CC0 gives everyone the right to do whatever, including for example, selling commercial products based on the decompilation (that would violate Nintendo's copyright anyway) without source code.
It's…
-
I was starting to look at failing inductor models with `functorch.config.use_functionalize=True` turned on, and one failure that I noticed is:
```
// run this
python benchmarks/timm_models.py --flo…
-
### Expected behavior
We are simulating QAOA on an NVIDIA DGX system.
Since the new pennylane version (v0.22) supports cuQuantum using the "lightning.gpu" device, we want to use it for potential sp…
-
Support for `export(..., pre_dispatch=True)` was added to make sure autograd (and autocast) functionality (e.g. `torch.no_grad`) continue to work on exported graphs. This is needed for training on an …
-
After enabling aten2aten decomps:
===== dlrm_backward_0 ======
Generating testing data...
aten2aten decomp: aten.detach.default
aten2aten decomp: aten.detach.default
aten2aten decomp: aten.deta…
-
…before Nintendō catches wind and issues a takedown notice. Please. 😟
-
With only scene collision visible, in the collision viewer settings, hitting pause shows everything behind the pause menu.
![image](https://github.com/user-attachments/assets/65bf56c2-e381-42f2-a724-…