NVIDIA/cutlass
1
0
Fork 0
mirror of https://github.com/NVIDIA/cutlass.git synced 2026-04-19 22:38:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUTLASS 3.1 Python interface documentation (#917)

Browse Source 
* Add 12.1 Dockerfile

* Add 3.1 docs
This commit is contained in:
Jack Kosaian
2023-04-18 15:11:35 -04:00
committed by GitHub
parent 54bebe417d
commit 9a83bd3381
81 changed files with 18997 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
python/README.md
View File

@@ -12,7 +12,7 @@ A, B, C, D = [np.ones((4096, 4096), dtype=np.float16) for i in range(4)]
plan.run(A, B, C, D)
```
**NOTE** The CUTLASS Python interface is currently an experimental release. The API may change in the future.
**NOTE:** The CUTLASS Python interface is currently an experimental release. The API may change in the future.
We welcome feedback from the community.
## Overview
@@ -93,11 +93,11 @@ The CUTLASS Python interface currently supports the following operations:
We recommend using the CUTLASS Python interface via one of the Docker images located in the [docker](/python/docker) directory.
```bash
docker build -t cutlass-cuda12.0:latest -f docker/Dockerfile-cuda12.0-pytorch .
docker run --gpus all -it --rm cutlass-cuda12.0:latest
docker build -t cutlass-cuda12.1:latest -f docker/Dockerfile-cuda12.1-pytorch .
docker run --gpus all -it --rm cutlass-cuda12.1:latest
```
The CUTLASS Python interface has been tested with CUDA 11.8 and CUDA 12.0 on Python 3.8.10 and 3.9.7.
The CUTLASS Python interface has been tested with CUDA 11.8, 12.0, and 12.1 on Python 3.8.10 and 3.9.7.
### Optional environment variables
Prior to installing the CUTLASS Python interface, one may optionally set the following environment variables: