mirror of
https://github.com/huggingface/candle.git
synced 2025-06-16 10:38:54 +00:00
feat: parse Cuda compute cap from env (#1066)
* feat: add support for multiple compute caps * Revert to one compute cap * fmt * fix
This commit is contained in:
@ -12,6 +12,9 @@ compute_cap
|
||||
8.9
|
||||
```
|
||||
|
||||
You can also compile the Cuda kernels for a specific compute cap using the
|
||||
`CUDA_COMPUTE_CAP=<compute cap>` environment variable.
|
||||
|
||||
If any of the above commands errors out, please make sure to update your Cuda version.
|
||||
|
||||
2. Create a new app and add [`candle-core`](https://github.com/huggingface/candle/tree/main/candle-core) with Cuda support.
|
||||
|
Reference in New Issue
Block a user