wehub-resource-sync
7 lines
414 B
Markdown
7 lines
414 B
Markdown
Loading model weights with fastsafetensors
|
|
===================================================================
|
|
|
|
Using fastsafetensors library enables loading model weights to GPU memory by leveraging GPU direct storage. See [their GitHub repository](https://github.com/foundation-model-stack/fastsafetensors) for more details.
|
|
|
|
To enable this feature, use the `--load-format fastsafetensors` command-line argument
|