Upload . with huggingface_hub

.gitattributes

@@ -30,3 +30,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+ffhq-sg2-510.ckpt filter=lfs diff=lfs merge=lfs -text
+lhq-sg3-410-best.ckpt filter=lfs diff=lfs merge=lfs -text

ffhq-sg2-510.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:126afd0477ab874921064cf8717308d3f715c80d0e25e2b966fd7e9a1746d71d
+size 390606755

ffhq-sg2-510.yaml

@@ -0,0 +1,42 @@
+model:
+  network:
+    dim: 512
+    num_timesteps: 1000
+    depth: 12
+    dim_head: 64
+    heads: 12
+  diffusion:
+    image_embed_dim: ${model.network.dim}
+    timesteps: ${model.network.num_timesteps}
+    cond_drop_prob: 0.2
+    image_embed_scale: 1.0
+    text_embed_scale: 1.0
+    beta_schedule: cosine
+    predict_x_start: true
+data:
+  bs: 512
+  format: webdataset
+  path: data/webdataset/sg2-ffhq-1024-clip/{00000..99}.tar
+  embed_noise_scale: 1.0
+  sg_pkl: https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan2/versions/1/files/stylegan2-ffhq-1024x1024.pkl
+  clip_variant: ViT-B/32
+  n_latents: 1
+  latent_dim: 512
+  latent_repeats:
+  - 18
+  val_im_samples: 64
+  val_text_samples: data/text/face-val.txt
+  val_samples_per_text: 4
+wandb_project: clip2latent
+wandb_entity: null
+name: baseline_noise_1
+device: cuda:0
+train:
+  znorm_embed: false
+  znorm_latent: true
+  max_it: 1000000
+  val_it: 10000
+  lr: 0.0001
+  weight_decay: 0.01
+  ema_update_every: 1
+  ema_beta: 0.99999

lhq-sg3-410-best.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:740220d88d810c362f1aa575c2d488d48d14fe6c9b5e40712049841efe92d910
+size 390606755

lhq-sg3-410-best.yaml

@@ -0,0 +1,42 @@
+model:
+  network:
+    dim: 512
+    num_timesteps: 1000
+    depth: 12
+    dim_head: 64
+    heads: 12
+  diffusion:
+    image_embed_dim: ${model.network.dim}
+    timesteps: ${model.network.num_timesteps}
+    cond_drop_prob: 0.2
+    image_embed_scale: 1.0
+    text_embed_scale: 1.0
+    beta_schedule: cosine
+    predict_x_start: true
+data:
+  bs: 512
+  format: webdataset
+  path: data/webdataset/sg3-lhq-256-clip/{00000..99}.tar
+  embed_noise_scale: 1.0
+  sg_pkl: /home/jpinkney/code/clip2latent/data/models/lhq-256-stylegan3-t-25Mimg.pkl
+  clip_variant: ViT-B/32
+  n_latents: 1
+  latent_dim: 512
+  latent_repeats:
+  - 16
+  val_im_samples: 64
+  val_text_samples: data/text/landscape-val.txt
+  val_samples_per_text: 4
+wandb_project: clip2latent
+wandb_entity: null
+name: lhq_noise_1
+device: cuda:0
+train:
+  znorm_embed: false
+  znorm_latent: true
+  max_it: 1000000
+  val_it: 10000
+  lr: 0.0001
+  weight_decay: 0.01
+  ema_update_every: 1
+  ema_beta: 0.99999