Upload folder using huggingface_hub

config.json

@@ -0,0 +1,40 @@
+{
+  "model_class": "SanaModel",
+  "input_size": 32,
+  "patch_size": 1,
+  "in_channels": 32,
+  "hidden_size": 2240,
+  "depth": 20,
+  "num_heads": 20,
+  "mlp_ratio": 2.5,
+  "caption_channels": 2304,
+  "model_max_length": 300,
+  "attn_type": "linear",
+  "cross_attn_type": "flash",
+  "linear_head_dim": 32,
+  "qk_norm": true,
+  "cross_norm": true,
+  "ffn_type": "glumbconv",
+  "mlp_acts": [
+    "silu",
+    "silu",
+    null
+  ],
+  "pred_sigma": false,
+  "learn_sigma": false,
+  "use_pe": false,
+  "pe_interpolation": 1.0,
+  "y_norm": true,
+  "y_norm_scale_factor": 1.0,
+  "use_connector": true,
+  "y_embedding_init_method": "zeros",
+  "connector_channels": 4096,
+  "norm_eps": 1e-06,
+  "class_dropout_prob": 0.1,
+  "drop_path": 0.0,
+  "vae_type": "AutoencoderDC",
+  "vae_pretrained": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
+  "vae_downsample_rate": 32,
+  "text_encoder_name": "google/gemma-2-2b-it",
+  "__class__": "SanaConfig"
+}

model-00000-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:009d698941c7c8c58f52858cf23e5fa3e52a196df749e63011706bf288fd5107
+size 2162601768

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:571094eb70c604316bec75f7ae1050586b7e5edcd909942e36ab294b7ebab2b7
+size 2161975312

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d6b7557ed55a480d3262d4a39b4e3cddffe5b2892e515d5b29d1e185d7c3cc6
+size 2161921544

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bcfb777ed6a8d387f6f60cd0ea535663d9141f7b438c64ae52df5bfb7e5c289
+size 6940688

model.safetensors.index.json

@@ -0,0 +1,490 @@
+{
+  "metadata": {
+    "total_size": 510507505408
+  },
+  "weight_map": {
+    "pos_embed": "model-00000-of-00004.safetensors",
+    "x_embedder.proj.weight": "model-00000-of-00004.safetensors",
+    "x_embedder.proj.bias": "model-00000-of-00004.safetensors",
+    "t_embedder.mlp.0.weight": "model-00000-of-00004.safetensors",
+    "t_embedder.mlp.0.bias": "model-00000-of-00004.safetensors",
+    "t_embedder.mlp.2.weight": "model-00000-of-00004.safetensors",
+    "t_embedder.mlp.2.bias": "model-00000-of-00004.safetensors",
+    "t_block.1.weight": "model-00000-of-00004.safetensors",
+    "t_block.1.bias": "model-00000-of-00004.safetensors",
+    "diffusion_connector.0.weight": "model-00000-of-00004.safetensors",
+    "diffusion_connector.0.bias": "model-00000-of-00004.safetensors",
+    "diffusion_connector.2.weight": "model-00000-of-00004.safetensors",
+    "diffusion_connector.2.bias": "model-00000-of-00004.safetensors",
+    "diffusion_connector.3.weight": "model-00000-of-00004.safetensors",
+    "y_embedder.y_proj.fc1.weight": "model-00000-of-00004.safetensors",
+    "y_embedder.y_proj.fc1.bias": "model-00000-of-00004.safetensors",
+    "y_embedder.y_proj.fc2.weight": "model-00000-of-00004.safetensors",
+    "y_embedder.y_proj.fc2.bias": "model-00000-of-00004.safetensors",
+    "attention_y_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.scale_shift_table": "model-00000-of-00004.safetensors",
+    "blocks.0.attn.to_q.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.0.attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.q_linear.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.q_linear.bias": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.to_k.bias": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.to_v.bias": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.cross_attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.mlp.inverted_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.mlp.inverted_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.0.mlp.depth_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.0.mlp.depth_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.0.mlp.point_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.scale_shift_table": "model-00000-of-00004.safetensors",
+    "blocks.1.attn.to_q.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.1.attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.q_linear.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.q_linear.bias": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.to_k.bias": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.to_v.bias": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.cross_attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.mlp.inverted_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.mlp.inverted_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.1.mlp.depth_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.1.mlp.depth_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.1.mlp.point_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.scale_shift_table": "model-00000-of-00004.safetensors",
+    "blocks.2.attn.to_q.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.2.attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.q_linear.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.q_linear.bias": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.to_k.bias": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.to_v.bias": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.cross_attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.mlp.inverted_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.mlp.inverted_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.2.mlp.depth_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.2.mlp.depth_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.2.mlp.point_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.scale_shift_table": "model-00000-of-00004.safetensors",
+    "blocks.3.attn.to_q.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.3.attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.q_linear.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.q_linear.bias": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.to_k.bias": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.to_v.bias": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.cross_attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.mlp.inverted_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.mlp.inverted_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.3.mlp.depth_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.3.mlp.depth_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.3.mlp.point_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.scale_shift_table": "model-00000-of-00004.safetensors",
+    "blocks.4.attn.to_q.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.4.attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.q_linear.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.q_linear.bias": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.to_k.bias": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.to_v.bias": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.cross_attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.mlp.inverted_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.mlp.inverted_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.4.mlp.depth_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.4.mlp.depth_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.4.mlp.point_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.scale_shift_table": "model-00000-of-00004.safetensors",
+    "blocks.5.attn.to_q.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.5.attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.q_linear.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.q_linear.bias": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.to_k.bias": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.to_v.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.to_v.bias": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.proj.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.proj.bias": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.q_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.cross_attn.k_norm.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.mlp.inverted_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.mlp.inverted_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.5.mlp.depth_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.5.mlp.depth_conv.conv.bias": "model-00000-of-00004.safetensors",
+    "blocks.5.mlp.point_conv.conv.weight": "model-00000-of-00004.safetensors",
+    "blocks.6.scale_shift_table": "model-00000-of-00004.safetensors",
+    "blocks.6.attn.to_q.weight": "model-00000-of-00004.safetensors",
+    "blocks.6.attn.to_k.weight": "model-00000-of-00004.safetensors",
+    "blocks.6.attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.6.attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.q_linear.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.q_linear.bias": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.to_k.bias": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.to_v.bias": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.cross_attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.mlp.inverted_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.mlp.inverted_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.6.mlp.depth_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.6.mlp.depth_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.6.mlp.point_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.scale_shift_table": "model-00001-of-00004.safetensors",
+    "blocks.7.attn.to_q.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.7.attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.q_linear.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.q_linear.bias": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.to_k.bias": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.to_v.bias": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.cross_attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.mlp.inverted_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.mlp.inverted_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.7.mlp.depth_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.7.mlp.depth_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.7.mlp.point_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.scale_shift_table": "model-00001-of-00004.safetensors",
+    "blocks.8.attn.to_q.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.8.attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.q_linear.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.q_linear.bias": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.to_k.bias": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.to_v.bias": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.cross_attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.mlp.inverted_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.mlp.inverted_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.8.mlp.depth_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.8.mlp.depth_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.8.mlp.point_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.scale_shift_table": "model-00001-of-00004.safetensors",
+    "blocks.9.attn.to_q.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.9.attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.q_linear.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.q_linear.bias": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.to_k.bias": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.to_v.bias": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.cross_attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.mlp.inverted_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.mlp.inverted_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.9.mlp.depth_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.9.mlp.depth_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.9.mlp.point_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.scale_shift_table": "model-00001-of-00004.safetensors",
+    "blocks.10.attn.to_q.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.10.attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.q_linear.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.q_linear.bias": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.to_k.bias": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.to_v.bias": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.cross_attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.mlp.inverted_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.mlp.inverted_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.10.mlp.depth_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.10.mlp.depth_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.10.mlp.point_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.scale_shift_table": "model-00001-of-00004.safetensors",
+    "blocks.11.attn.to_q.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.11.attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.q_linear.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.q_linear.bias": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.to_k.bias": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.to_v.bias": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.cross_attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.mlp.inverted_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.mlp.inverted_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.11.mlp.depth_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.11.mlp.depth_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.11.mlp.point_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.scale_shift_table": "model-00001-of-00004.safetensors",
+    "blocks.12.attn.to_q.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.12.attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.q_linear.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.q_linear.bias": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.to_k.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.to_k.bias": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.to_v.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.to_v.bias": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.proj.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.proj.bias": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.q_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.cross_attn.k_norm.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.mlp.inverted_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.mlp.inverted_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.12.mlp.depth_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.12.mlp.depth_conv.conv.bias": "model-00001-of-00004.safetensors",
+    "blocks.12.mlp.point_conv.conv.weight": "model-00001-of-00004.safetensors",
+    "blocks.13.scale_shift_table": "model-00001-of-00004.safetensors",
+    "blocks.13.attn.to_q.weight": "model-00001-of-00004.safetensors",
+    "blocks.13.attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.13.attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.q_linear.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.q_linear.bias": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.to_k.bias": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.to_v.bias": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.cross_attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.mlp.inverted_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.mlp.inverted_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.13.mlp.depth_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.13.mlp.depth_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.13.mlp.point_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.scale_shift_table": "model-00002-of-00004.safetensors",
+    "blocks.14.attn.to_q.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.14.attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.q_linear.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.q_linear.bias": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.to_k.bias": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.to_v.bias": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.cross_attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.mlp.inverted_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.mlp.inverted_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.14.mlp.depth_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.14.mlp.depth_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.14.mlp.point_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.scale_shift_table": "model-00002-of-00004.safetensors",
+    "blocks.15.attn.to_q.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.15.attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.q_linear.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.q_linear.bias": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.to_k.bias": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.to_v.bias": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.cross_attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.mlp.inverted_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.mlp.inverted_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.15.mlp.depth_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.15.mlp.depth_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.15.mlp.point_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.scale_shift_table": "model-00002-of-00004.safetensors",
+    "blocks.16.attn.to_q.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.16.attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.q_linear.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.q_linear.bias": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.to_k.bias": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.to_v.bias": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.cross_attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.mlp.inverted_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.mlp.inverted_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.16.mlp.depth_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.16.mlp.depth_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.16.mlp.point_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.scale_shift_table": "model-00002-of-00004.safetensors",
+    "blocks.17.attn.to_q.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.17.attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.q_linear.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.q_linear.bias": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.to_k.bias": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.to_v.bias": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.cross_attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.mlp.inverted_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.mlp.inverted_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.17.mlp.depth_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.17.mlp.depth_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.17.mlp.point_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.scale_shift_table": "model-00002-of-00004.safetensors",
+    "blocks.18.attn.to_q.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.18.attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.q_linear.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.q_linear.bias": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.to_k.bias": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.to_v.bias": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.cross_attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.mlp.inverted_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.mlp.inverted_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.18.mlp.depth_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.18.mlp.depth_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.18.mlp.point_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.scale_shift_table": "model-00002-of-00004.safetensors",
+    "blocks.19.attn.to_q.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.19.attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.q_linear.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.q_linear.bias": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.to_k.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.to_k.bias": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.to_v.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.to_v.bias": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.proj.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.proj.bias": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.q_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.cross_attn.k_norm.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.mlp.inverted_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.mlp.inverted_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.19.mlp.depth_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "blocks.19.mlp.depth_conv.conv.bias": "model-00002-of-00004.safetensors",
+    "blocks.19.mlp.point_conv.conv.weight": "model-00002-of-00004.safetensors",
+    "final_layer.scale_shift_table": "model-00003-of-00004.safetensors",
+    "final_layer.linear.weight": "model-00003-of-00004.safetensors",
+    "final_layer.linear.bias": "model-00003-of-00004.safetensors",
+    "y_embedder.y_embedding": "model-00003-of-00004.safetensors"
+  }
+}