Push trained VAE model

README.md

@@ -0,0 +1,3 @@
+# Transfusion VAE
+
+A UNet-style VAE trained on CC3M with adversarial and perceptual losses.

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bccf2a2abec6b609f2a3afe0024dd852e48d6d0dcf82c9677b15062f64bed9c3
+size 322243354

vae_config.json

@@ -0,0 +1,29 @@
+{
+  "image_size": 256,
+  "in_channels": 3,
+  "out_channels": 3,
+  "hidden_channels": 128,
+  "num_res_blocks": 3,
+  "channel_multipliers": [
+    1,
+    2,
+    2,
+    4,
+    4
+  ],
+  "attention_resolutions": [
+    32
+  ],
+  "dropout": 0.0,
+  "use_checkpoint": false,
+  "use_fp16": false,
+  "resblock_updown": true,
+  "z_channels": 8,
+  "double_z": true,
+  "λ_recon": 1.0,
+  "λ_kl": 1e-6,
+  "λ_lpips": 1.0,
+  "λ_id": 0.1,
+  "λ_gan": 0.5,
+  "gan_start_step": 50000
+}