Upload folder using huggingface_hub

dit.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:992de40836906ad7c4565d2052f149c229cc88d8d6411abc085f51b6b4ffc9c7
+size 406159595

dit_config.json

@@ -0,0 +1,12 @@
+{
+  "model_type": "mmdit",
+  "latent_dim": 16,
+  "latent_size": 16,
+  "hidden_dim": 512,
+  "n_layers": 16,
+  "n_heads": 8,
+  "text_embed_dim": 768,
+  "patch_size": 128,
+  "sine_encoding_frequency": 10000,
+  "num_classes": 1081
+}

encoder_config.json

@@ -0,0 +1,3 @@
+{
+  "encoder_path": "google/embeddinggemma-300m"
+}

vae.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94231d899b54ecad421904f3de3115b3d1c63a8496c775216eb4b17a89572055
+size 53156779

vae_config.json

@@ -0,0 +1,21 @@
+{
+  "in_channels": 3,
+  "image_size": 128,
+  "latent_dim": 16,
+  "depths": [
+    3,
+    5,
+    7,
+    3
+  ],
+  "dims": [
+    48,
+    96,
+    192,
+    384
+  ],
+  "ln_eps": 1e-06,
+  "iaf": false,
+  "iaf_n_blocks": null,
+  "iaf_timesteps": null
+}