-
I've noticed that MatX needs to include some CUDA-related abstractions and facilities (like error handling) which are actually not specific to matrix/tensor/numeric computations in any way.
I woul…
-
Hello Jake,
Long time no speak. Actually, do I even have your email address? I think I don't. would you mind emailing me something?
Anyway... as I was looking for your address in your repos, I c…
-
This is the error log
```
0%| …
-
### 🐛 Describe the bug
We encountered an illegal memory access issue with `torch.compile` and customized torch library operator.
Here's one minimal example to reproduce:
```python
import torch…
-
As part of the warnings seen in #468, with a bit of trial and error I have narrowed down the cause of the "unsupported operand `|`" errors being produced by sphinx, despite running with a python 3.12 …
-
With version v4.1.2, during `./configure`, Extrae build system finds correctly location of `xml2` installation, with headers under `${prefix}/include/libxml2` and it sets correctly
```
XML2_CFLAGS='…
-
I'm trying to compile one of the examples of my [cuda-api-wrappers library](https://github.com/eyalroz/cuda-api-wrappers), which uses string-view-lite, using C++14 instead of C++11 like I was compilin…
-
### 🐛 Describe the bug
I understand from https://github.com/pytorch/pytorch/issues/134756 that it's necessary to compile `create_block_mask` to avoid materialising the full NxM mask. However when doi…
-
### 🐛 Describe the bug
Flex attention on FSDP works without compile, but not with compile. The key error seems to be `ValueError: Pointer argument (at 2) cannot be accessed from Triton (cpu tensor?)`…
-
### Description
I'm the author of cuda-api-wrappers. The library was added to conan-center this year (great!) - but without telling me about it (not great).
I have now noticed this happened, but m…