What is the difference between sd21-unclip-h.ckpt and sd21-unclip-l.ckpt

#7
by youxun - opened

image.png

I'm just throwing that in here to at least have a simple answer
ChatGPT:
sd21-unclip-h.ckpt was trained with a high level of regularization, which makes the model more robust to overfitting, but may also result in slightly lower performance on certain tasks.
On the other hand, sd21-unclip-l.ckpt was trained with a lower level of regularization, which may result in higher performance on certain tasks, but could also make the model more prone to overfitting.

Pros:
sd21-unclip-h.ckpt: High accuracy, Can handle high-resolution images, Better for fine-grained image generation
sd21-unclip-l.ckpt: Faster inference speed, Requires less memory during inference, Suitable for real-time applications

these corpo shill devs put all this useless bullshit info about muh biases and muh CO2 emissions and muh licenses but never include what actually matters: the explanations

Sign up or log in to comment