search

modelscope / DiffSynth-Studio

Enjoy the magic of Diffusion models!
Apache License 2.0
6.4k stars 575 forks source link

无法通过kolors官方代码加载Diffsynth训练的lora权重 #143

Closed JoshonSmith closed 2 months ago

JoshonSmith commented 2 months ago

请问是否可以将epoch=0-step=500.ckpt 转为 pytorch_lora_weights.bin

Artiprocher commented 2 months ago

我们训练的 LoRA 权重只能通过我们自己的代码加载，请参考 examples/train 中给出的代码样例