Open Ishihara-Masabumi opened 1 year ago
What does --sample_unet=1 look like? Depending on which unet looks better, the other will need more training.
Here's some of my samples from my new model after 40 epochs with 107k images:
unet1, 256 dim. cond_scale=1.1
Even after 40 epochs (4.3 million images seen) I need a lot more training.
@deepglugs How long it took you to train this much and what GPU are you using?
The training time is about 2 or 3 days wih NVIDIA A10.
Woah! In paper they trained for 2.5M steps with a batch size of 2048 (are u guys using same batch size?).
What dateset are you using? The LAION-400M seems 10 TB. That's HUGE.
I tried to train and generate images as the following command lines.
The first generated image is below.![girl2_256](https://user-images.githubusercontent.com/23301778/191196395-2e8f1f67-bcb4-4d25-b8cb-780b528a5162.png)
The second generated image is below.![girl3_256](https://user-images.githubusercontent.com/23301778/191196498-df5b9e8a-c405-4cd2-8f34-b22eafefe96c.png)
Are they OK? or lack of training?