search

ziplab / PTQD

The official implementation of PTQD: Accurate Post-Training Quantization for Diffusion Models
85 stars 4 forks source link

what is the seed for experiments? #14

Open doobluhc opened 4 months ago

doobluhc commented 4 months ago

Hello, for the experiment you did for ImageNet256x256 LDM-4(steps=250,eta=1.0,scale=1.5), i am getting FID=3.62 but you reported 5.05. Could you please tell me what is the seed you fixed?

TaylorJocelyn commented 3 months ago

Seeing that you have achieved better results，I would like to ask you a few question.

  1. In collect_diffusion_input_imagenet.py file, how did you set classes and n_samples_per_class?

    image

  2. In collect_quant_error_brecq.py file, how did you set classes and n_samples_per_class?

    image

Looking forward to your answer. Thank you very much.