@@ -0,0 +1,41 @@
+{
+    "diffusers": {
+        "global_rename_dict": {
+            "patch_embedding": "patch_embedding",
+            "condition_embedder.text_embedder.linear_1": "text_embedding.0",
+            "condition_embedder.text_embedder.linear_2": "text_embedding.2",
+            "condition_embedder.time_embedder.linear_1": "time_embedding.0",
+            "condition_embedder.time_embedder.linear_2": "time_embedding.2",
+            "condition_embedder.time_proj": "time_projection.1",
+            "condition_embedder.image_embedder.norm1": "img_emb.proj.0",
+            "condition_embedder.image_embedder.ff.net.0.proj": "img_emb.proj.1",
+            "condition_embedder.image_embedder.ff.net.2": "img_emb.proj.3",
+            "condition_embedder.image_embedder.norm2": "img_emb.proj.4",
+            "condition_embedder.image_embedder.pos_embed": "img_emb.emb_pos",
+            "proj_out": "head.head",
+            "scale_shift_table": "head.modulation"
+        },
+        "rename_dict": {
+            "attn1.to_q": "self_attn.q",
+            "attn1.to_k": "self_attn.k",
+            "attn1.to_v": "self_attn.v",
+            "attn1.to_out.0": "self_attn.o",
+            "attn1.norm_q": "self_attn.norm_q",
+            "attn1.norm_k": "self_attn.norm_k",
+            "to_gate_compress": "self_attn.gate_compress",
+            "attn2.to_q": "cross_attn.q",
+            "attn2.to_k": "cross_attn.k",
+            "attn2.to_v": "cross_attn.v",
+            "attn2.to_out.0": "cross_attn.o",
+            "attn2.norm_q": "cross_attn.norm_q",
+            "attn2.norm_k": "cross_attn.norm_k",
+            "attn2.add_k_proj": "cross_attn.k_img",
+            "attn2.add_v_proj": "cross_attn.v_img",
+            "attn2.norm_added_k": "cross_attn.norm_k_img",
+            "norm2": "norm3",
+            "ffn.net.0.proj": "ffn.0",
+            "ffn.net.2": "ffn.2",
+            "scale_shift_table": "modulation"
+        }
+    }
+}