Generation from noise give poor result

[lchunleo]

Hi
I had trained on my custom dataset and when the model trained , it produced the checkpoint and sample image file which contains a grid of images. My questions are

1. is the grid of images is all generated or denoised images from my training data ?
2. when viewed the grid images , they are visually show reasonable correct images but when I used the check point , the produced images are poor than those in the grid of images (generated during training ) .

Thanks