You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+31-14Lines changed: 31 additions & 14 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -49,7 +49,14 @@ In addition, for layers to speed up on specific hardware (such as CUDA devices,
49
49
-**[MLX](https://github.com/ml-explore/mlx)** for mlx-based layers on MacOS
50
50
-**[CUTLASS](https://github.com/NVIDIA/cutlass)** for cutlass-based layers
51
51
52
-
Currently, the engine **needs to be built from source**.
52
+
### Binary Releases (coming soon)
53
+
54
+
We are currently preparing experimental binary releases.
55
+
Their installation will be documented in this section.
56
+
For now, please follow the guide below to build from source.
57
+
58
+
### Build From Source
59
+
53
60
We provide instructions for the following options:
54
61
55
62
- Conda + Linux (with CUDA and cutlass)
@@ -60,7 +67,7 @@ We recommend managing your BITorch Engine installation in a conda environment (o
60
67
You may want to keep everything (environment, code, etc.) in one directory or use the default directory for conda environments.
61
68
You may wish to adapt the CUDA version to 12.1 where applicable.
62
69
63
-
### Conda on Linux (with CUDA)
70
+
####Conda on Linux (with CUDA)
64
71
65
72
To use these instructions, you need to have [conda](https://conda.io/projects/conda/en/latest/user-guide/getting-started.html) and a suitable C++ compiler installed.
3.[Download customized Torch 2.1.0](https://drive.google.com/drive/folders/1T22b8JhN-E3xbn3h332rI1VjqXONZeB7?usp=sharing) (it allows gradients on INT tensors, built for Python 3.9 and CUDA 11.8) and install it with pip:
83
+
3. Download our customized torch for CUDA 11.8 and Python 3.9, it allows gradients on INT tensors and install it with pip (you can find other versions [here](https://packages.greenbit.ai/whl/)):
select the package fit for the cuda version you installed in the previous step
117
-
(it allows gradients on INT tensors, built for Python 3.9 and CUDA 11.8) and install it with pip:
123
+
3. Download our customized torch for CUDA 11.8 and Python 3.9, it allows gradients on INT tensors and install it with pip (you can find other versions [here](https://packages.greenbit.ai/whl/)):
2. Download [customized Torch for MacOS/arm](https://drive.google.com/drive/folders/1T22b8JhN-E3xbn3h332rI1VjqXONZeB7?usp=sharing) (it allows gradients on INT tensors, built for Python 3.9 and CUDA 11.8) and install it with pip:
0 commit comments