Commit
•
a2b3d6b
1
Parent(s):
cef726f
Updated Readme
Browse files
README.md
CHANGED
|
@@ -4,7 +4,23 @@
|
|
| 4 |
|
| 5 |
A 512x512 unconditional ImageNet diffusion model, fine-tuned for 900.000 samples from the 512x512 unconditional ImageNet diffusion model. It was fine-tuned using 60.000 images of architecture of the AIDA dataset from Harvard x ArchDaily.
|
| 6 |
|
|
|
|
| 7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
---
|
| 10 |
license: cc
|
|
|
|
| 4 |
|
| 5 |
A 512x512 unconditional ImageNet diffusion model, fine-tuned for 900.000 samples from the 512x512 unconditional ImageNet diffusion model. It was fine-tuned using 60.000 images of architecture of the AIDA dataset from Harvard x ArchDaily.
|
| 6 |
|
| 7 |
+
## Config (as used in Disco Diffusion)
|
| 8 |
|
| 9 |
+
{
|
| 10 |
+
"attention_resolutions": '32, 16, 8',
|
| 11 |
+
"class_cond": False,
|
| 12 |
+
"diffusion_steps": 1000,
|
| 13 |
+
"image_size": 512,
|
| 14 |
+
"learn_sigma": True,
|
| 15 |
+
"noise_schedule": "linear",
|
| 16 |
+
"num_channels": 256,
|
| 17 |
+
"num_head_channels": 64,
|
| 18 |
+
"num_res_blocks": 2,
|
| 19 |
+
"resblock_updown": True,
|
| 20 |
+
"rescale_timesteps": True,
|
| 21 |
+
"timestep_respacing": "250",
|
| 22 |
+
"use_scale_shift_norm": True
|
| 23 |
+
}
|
| 24 |
|
| 25 |
---
|
| 26 |
license: cc
|