File size: 108 Bytes
5fa1a76
1
In Transformers, the quantized models are initialized first on the "meta" device before loading the weights.