Commit ·
2a4064d
1
Parent(s): 91c1c23
Upload lhq-sg3-410.yaml
Browse files- lhq-sg3-410.yaml +42 -0
lhq-sg3-410.yaml
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model:
|
| 2 |
+
network:
|
| 3 |
+
dim: 512
|
| 4 |
+
num_timesteps: 1000
|
| 5 |
+
depth: 12
|
| 6 |
+
dim_head: 64
|
| 7 |
+
heads: 12
|
| 8 |
+
diffusion:
|
| 9 |
+
image_embed_dim: ${model.network.dim}
|
| 10 |
+
timesteps: ${model.network.num_timesteps}
|
| 11 |
+
cond_drop_prob: 0.2
|
| 12 |
+
image_embed_scale: 1.0
|
| 13 |
+
text_embed_scale: 1.0
|
| 14 |
+
beta_schedule: cosine
|
| 15 |
+
predict_x_start: true
|
| 16 |
+
data:
|
| 17 |
+
bs: 512
|
| 18 |
+
format: webdataset
|
| 19 |
+
path: data/webdataset/sg3-lhq-256-clip/{00000..99}.tar
|
| 20 |
+
embed_noise_scale: 1.0
|
| 21 |
+
sg_pkl: https://huggingface.co/justinpinkney/stylegan3-t-lhq-256/resolve/main/lhq-256-stylegan3-t-25Mimg.pkl
|
| 22 |
+
clip_variant: ViT-B/32
|
| 23 |
+
n_latents: 1
|
| 24 |
+
latent_dim: 512
|
| 25 |
+
latent_repeats:
|
| 26 |
+
- 16
|
| 27 |
+
val_im_samples: 64
|
| 28 |
+
val_text_samples: data/text/landscape-val.txt
|
| 29 |
+
val_samples_per_text: 4
|
| 30 |
+
wandb_project: clip2latent
|
| 31 |
+
wandb_entity: null
|
| 32 |
+
name: lhq_noise_1
|
| 33 |
+
device: cuda:0
|
| 34 |
+
train:
|
| 35 |
+
znorm_embed: false
|
| 36 |
+
znorm_latent: true
|
| 37 |
+
max_it: 1000000
|
| 38 |
+
val_it: 10000
|
| 39 |
+
lr: 0.0001
|
| 40 |
+
weight_decay: 0.01
|
| 41 |
+
ema_update_every: 1
|
| 42 |
+
ema_beta: 0.99999
|