diffsynth-engine 0.4.0__tar.gz → 0.4.1__tar.gz

{diffsynth_engine-0.4.0 → diffsynth_engine-0.4.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffsynth_engine
-Version: 0.4.0
+Version: 0.4.1
 Author: MuseAI x ModelScope
 Classifier: Programming Language :: Python :: 3
 Classifier: Operating System :: OS Independent
@@ -28,6 +28,7 @@ Requires-Dist: opencv-python
 Provides-Extra: dev
 Requires-Dist: diffusers==0.31.0; extra == "dev"
 Requires-Dist: transformers==4.45.2; extra == "dev"
+Requires-Dist: accelerate; extra == "dev"
 Requires-Dist: build; extra == "dev"
 Requires-Dist: ruff; extra == "dev"
 Requires-Dist: scikit-image; extra == "dev"

{diffsynth_engine-0.4.0 → diffsynth_engine-0.4.1}/README.md

@@ -23,6 +23,7 @@ and offloading strategies, enabling loading of larger diffusion models (e.g., Fl
 
 ## News
 
+- **[v0.4.1](https://github.com/modelscope/DiffSynth-Engine/releases/tag/v0.4.1)** | **August 4, 2025**: 🔥Supports [Qwen-Image](https://www.modelscope.cn/models/Qwen/Qwen-Image), an image generation model excels at complex text rendering and creating images in a wide range of artistic styles.
 - **[v0.4.0](https://github.com/modelscope/DiffSynth-Engine/releases/tag/v0.4.0)** | **August 1, 2025**:
   - 🔥Supports [Wan2.2](https://modelscope.cn/collections/tongyiwanxiang-22--shipinshengcheng-2bb5b1adef2840) video generation model
   - ⚠️[**Breaking Change**] Improved `from_pretrained` method pipeline initialization
@@ -49,21 +50,24 @@ pip3 install -e .
 ### Usage
 Text to image
 ```python
-from diffsynth_engine import fetch_model, FluxImagePipeline
+from diffsynth_engine import fetch_model, FluxImagePipeline, FluxPipelineConfig
 
 model_path = fetch_model("muse/flux-with-vae", path="flux1-dev-with-vae.safetensors")
-pipe = FluxImagePipeline.from_pretrained(model_path, device='cuda:0')
+
+config = FluxPipelineConfig.basic_config(model_path=model_path, device="cuda:0")
+pipe = FluxImagePipeline.from_pretrained(config)
 image = pipe(prompt="a cat")
 image.save("image.png")
 ```
 Text to image with LoRA
 ```python
-from diffsynth_engine import fetch_model, FluxImagePipeline
+from diffsynth_engine import fetch_model, FluxImagePipeline, FluxPipelineConfig
 
 model_path = fetch_model("muse/flux-with-vae", path="flux1-dev-with-vae.safetensors")
 lora_path = fetch_model("DonRat/MAJICFLUS_SuperChinesestyleheongsam", path="麦橘超国风旗袍.safetensors")
 
-pipe = FluxImagePipeline.from_pretrained(model_path, device='cuda:0')
+config = FluxPipelineConfig.basic_config(model_path=model_path, device="cuda:0")
+pipe = FluxImagePipeline.from_pretrained(config)
 pipe.load_lora(path=lora_path, scale=1.0)
 image = pipe(prompt="a girl, qipao")
 image.save("image.png")

{diffsynth_engine-0.4.0 → diffsynth_engine-0.4.1}/diffsynth_engine/__init__.py

@@ -3,6 +3,7 @@ from .configs import (
     SDXLPipelineConfig,
     FluxPipelineConfig,
     WanPipelineConfig,
+    QwenImagePipelineConfig,
     ControlNetParams,
     ControlType,
 )
@@ -11,6 +12,7 @@ from .pipelines import (
     SDXLImagePipeline,
     SDImagePipeline,
     WanVideoPipeline,
+    QwenImagePipeline,
 )
 from .models.flux import FluxControlNet, FluxIPAdapter, FluxRedux
 from .models.sd import SDControlNet
@@ -31,6 +33,7 @@ __all__ = [
     "FluxPipelineConfig",
     "WanPipelineConfig",
     "FluxImagePipeline",
+    "QwenImagePipelineConfig",
     "FluxControlNet",
     "FluxIPAdapter",
     "FluxRedux",
@@ -39,6 +42,7 @@ __all__ = [
     "SDXLImagePipeline",
     "SDImagePipeline",
     "WanVideoPipeline",
+    "QwenImagePipeline",
     "FluxInpaintingTool",
     "FluxOutpaintingTool",
     "FluxIPAdapterRefTool",

diffsynth_engine-0.4.1/diffsynth_engine/conf/models/qwen_image/qwen2_5_vl_config.json

@@ -0,0 +1,25 @@
+{
+    "hidden_size": 3584,
+    "intermediate_size": 18944,
+    "num_hidden_layers": 28,
+    "num_attention_heads": 28,
+    "num_key_value_heads": 4,
+    "mrope_section": [
+        16,
+        24,
+        24
+    ],
+    "rms_norm_eps": 1e-6,
+    "use_cache": true,
+    "use_sliding_window": false,
+    "sliding_window": 32768,
+    "max_window_layers": 28,
+    "vocab_size": 152064,
+    "pad_token_id": 151643,
+    "im_start_token_id": 151644,
+    "im_end_token_id": 151645,
+    "vision_start_token_id": 151652,
+    "vision_end_token_id": 151653,
+    "image_token_id": 151655,
+    "video_token_id": 151656
+}

diffsynth_engine-0.4.1/diffsynth_engine/conf/models/qwen_image/qwen2_5_vl_vision_config.json

@@ -0,0 +1,19 @@
+{
+    "in_channels": 3,
+    "hidden_size": 1280,
+    "intermediate_size": 3420,
+    "out_hidden_size": 3584,
+    "num_heads": 16,
+    "depth": 32,
+    "patch_size": 14,
+    "temporal_patch_size": 2,
+    "spatial_merge_size": 2,
+    "tokens_per_second": 2,
+    "window_size": 112,
+    "fullatt_block_indexes": [
+        7,
+        15,
+        23,
+        31
+    ]
+}

diffsynth_engine-0.4.1/diffsynth_engine/conf/models/qwen_image/qwen_image_vae_keymap.json

@@ -0,0 +1,244 @@
+{
+    "diffusers": {
+        "rename_dict": {
+            "model.post_quant_conv.bias": "model.conv2.bias",
+            "model.post_quant_conv.weight": "model.conv2.weight",
+            "model.quant_conv.bias": "model.conv1.bias",
+            "model.quant_conv.weight": "model.conv1.weight",
+            "model.encoder.conv_in.bias": "model.encoder.conv1.bias",
+            "model.encoder.conv_in.weight": "model.encoder.conv1.weight",
+            "model.encoder.down_blocks.0.conv_shortcut.bias": "model.encoder.downsamples.0.shortcut.bias",
+            "model.encoder.down_blocks.0.conv_shortcut.weight": "model.encoder.downsamples.0.shortcut.weight",
+            "model.encoder.down_blocks.0.conv1.bias": "model.encoder.downsamples.0.residual.2.bias",
+            "model.encoder.down_blocks.0.conv1.weight": "model.encoder.downsamples.0.residual.2.weight",
+            "model.encoder.down_blocks.0.conv2.bias": "model.encoder.downsamples.0.residual.6.bias",
+            "model.encoder.down_blocks.0.conv2.weight": "model.encoder.downsamples.0.residual.6.weight",
+            "model.encoder.down_blocks.0.norm1.gamma": "model.encoder.downsamples.0.residual.0.gamma",
+            "model.encoder.down_blocks.0.norm2.gamma": "model.encoder.downsamples.0.residual.3.gamma",
+            "model.encoder.down_blocks.1.conv_shortcut.bias": "model.encoder.downsamples.1.shortcut.bias",
+            "model.encoder.down_blocks.1.conv_shortcut.weight": "model.encoder.downsamples.1.shortcut.weight",
+            "model.encoder.down_blocks.1.conv1.bias": "model.encoder.downsamples.1.residual.2.bias",
+            "model.encoder.down_blocks.1.conv1.weight": "model.encoder.downsamples.1.residual.2.weight",
+            "model.encoder.down_blocks.1.conv2.bias": "model.encoder.downsamples.1.residual.6.bias",
+            "model.encoder.down_blocks.1.conv2.weight": "model.encoder.downsamples.1.residual.6.weight",
+            "model.encoder.down_blocks.1.norm1.gamma": "model.encoder.downsamples.1.residual.0.gamma",
+            "model.encoder.down_blocks.1.norm2.gamma": "model.encoder.downsamples.1.residual.3.gamma",
+            "model.encoder.down_blocks.2.resample.1.bias": "model.encoder.downsamples.2.resample.1.bias",
+            "model.encoder.down_blocks.2.resample.1.weight": "model.encoder.downsamples.2.resample.1.weight",
+            "model.encoder.down_blocks.2.time_conv.bias": "model.encoder.downsamples.2.time_conv.bias",
+            "model.encoder.down_blocks.2.time_conv.weight": "model.encoder.downsamples.2.time_conv.weight",
+            "model.encoder.down_blocks.3.conv_shortcut.bias": "model.encoder.downsamples.3.shortcut.bias",
+            "model.encoder.down_blocks.3.conv_shortcut.weight": "model.encoder.downsamples.3.shortcut.weight",
+            "model.encoder.down_blocks.3.conv1.bias": "model.encoder.downsamples.3.residual.2.bias",
+            "model.encoder.down_blocks.3.conv1.weight": "model.encoder.downsamples.3.residual.2.weight",
+            "model.encoder.down_blocks.3.conv2.bias": "model.encoder.downsamples.3.residual.6.bias",
+            "model.encoder.down_blocks.3.conv2.weight": "model.encoder.downsamples.3.residual.6.weight",
+            "model.encoder.down_blocks.3.norm1.gamma": "model.encoder.downsamples.3.residual.0.gamma",
+            "model.encoder.down_blocks.3.norm2.gamma": "model.encoder.downsamples.3.residual.3.gamma",
+            "model.encoder.down_blocks.4.conv_shortcut.bias": "model.encoder.downsamples.4.shortcut.bias",
+            "model.encoder.down_blocks.4.conv_shortcut.weight": "model.encoder.downsamples.4.shortcut.weight",
+            "model.encoder.down_blocks.4.conv1.bias": "model.encoder.downsamples.4.residual.2.bias",
+            "model.encoder.down_blocks.4.conv1.weight": "model.encoder.downsamples.4.residual.2.weight",
+            "model.encoder.down_blocks.4.conv2.bias": "model.encoder.downsamples.4.residual.6.bias",
+            "model.encoder.down_blocks.4.conv2.weight": "model.encoder.downsamples.4.residual.6.weight",
+            "model.encoder.down_blocks.4.norm1.gamma": "model.encoder.downsamples.4.residual.0.gamma",
+            "model.encoder.down_blocks.4.norm2.gamma": "model.encoder.downsamples.4.residual.3.gamma",
+            "model.encoder.down_blocks.5.resample.1.bias": "model.encoder.downsamples.5.resample.1.bias",
+            "model.encoder.down_blocks.5.resample.1.weight": "model.encoder.downsamples.5.resample.1.weight",
+            "model.encoder.down_blocks.5.time_conv.bias": "model.encoder.downsamples.5.time_conv.bias",
+            "model.encoder.down_blocks.5.time_conv.weight": "model.encoder.downsamples.5.time_conv.weight",
+            "model.encoder.down_blocks.6.conv_shortcut.bias": "model.encoder.downsamples.6.shortcut.bias",
+            "model.encoder.down_blocks.6.conv_shortcut.weight": "model.encoder.downsamples.6.shortcut.weight",
+            "model.encoder.down_blocks.6.conv1.bias": "model.encoder.downsamples.6.residual.2.bias",
+            "model.encoder.down_blocks.6.conv1.weight": "model.encoder.downsamples.6.residual.2.weight",
+            "model.encoder.down_blocks.6.conv2.bias": "model.encoder.downsamples.6.residual.6.bias",
+            "model.encoder.down_blocks.6.conv2.weight": "model.encoder.downsamples.6.residual.6.weight",
+            "model.encoder.down_blocks.6.norm1.gamma": "model.encoder.downsamples.6.residual.0.gamma",
+            "model.encoder.down_blocks.6.norm2.gamma": "model.encoder.downsamples.6.residual.3.gamma",
+            "model.encoder.down_blocks.7.conv_shortcut.bias": "model.encoder.downsamples.7.shortcut.bias",
+            "model.encoder.down_blocks.7.conv_shortcut.weight": "model.encoder.downsamples.7.shortcut.weight",
+            "model.encoder.down_blocks.7.conv1.bias": "model.encoder.downsamples.7.residual.2.bias",
+            "model.encoder.down_blocks.7.conv1.weight": "model.encoder.downsamples.7.residual.2.weight",
+            "model.encoder.down_blocks.7.conv2.bias": "model.encoder.downsamples.7.residual.6.bias",
+            "model.encoder.down_blocks.7.conv2.weight": "model.encoder.downsamples.7.residual.6.weight",
+            "model.encoder.down_blocks.7.norm1.gamma": "model.encoder.downsamples.7.residual.0.gamma",
+            "model.encoder.down_blocks.7.norm2.gamma": "model.encoder.downsamples.7.residual.3.gamma",
+            "model.encoder.down_blocks.8.resample.1.bias": "model.encoder.downsamples.8.resample.1.bias",
+            "model.encoder.down_blocks.8.resample.1.weight": "model.encoder.downsamples.8.resample.1.weight",
+            "model.encoder.down_blocks.8.time_conv.bias": "model.encoder.downsamples.8.time_conv.bias",
+            "model.encoder.down_blocks.8.time_conv.weight": "model.encoder.downsamples.8.time_conv.weight",
+            "model.encoder.down_blocks.9.conv_shortcut.bias": "model.encoder.downsamples.9.shortcut.bias",
+            "model.encoder.down_blocks.9.conv_shortcut.weight": "model.encoder.downsamples.9.shortcut.weight",
+            "model.encoder.down_blocks.9.conv1.bias": "model.encoder.downsamples.9.residual.2.bias",
+            "model.encoder.down_blocks.9.conv1.weight": "model.encoder.downsamples.9.residual.2.weight",
+            "model.encoder.down_blocks.9.conv2.bias": "model.encoder.downsamples.9.residual.6.bias",
+            "model.encoder.down_blocks.9.conv2.weight": "model.encoder.downsamples.9.residual.6.weight",
+            "model.encoder.down_blocks.9.norm1.gamma": "model.encoder.downsamples.9.residual.0.gamma",
+            "model.encoder.down_blocks.9.norm2.gamma": "model.encoder.downsamples.9.residual.3.gamma",
+            "model.encoder.down_blocks.10.conv_shortcut.bias": "model.encoder.downsamples.10.shortcut.bias",
+            "model.encoder.down_blocks.10.conv_shortcut.weight": "model.encoder.downsamples.10.shortcut.weight",
+            "model.encoder.down_blocks.10.conv1.bias": "model.encoder.downsamples.10.residual.2.bias",
+            "model.encoder.down_blocks.10.conv1.weight": "model.encoder.downsamples.10.residual.2.weight",
+            "model.encoder.down_blocks.10.conv2.bias": "model.encoder.downsamples.10.residual.6.bias",
+            "model.encoder.down_blocks.10.conv2.weight": "model.encoder.downsamples.10.residual.6.weight",
+            "model.encoder.down_blocks.10.norm1.gamma": "model.encoder.downsamples.10.residual.0.gamma",
+            "model.encoder.down_blocks.10.norm2.gamma": "model.encoder.downsamples.10.residual.3.gamma",
+            "model.encoder.mid_block.attentions.0.norm.gamma": "model.encoder.middle.1.norm.gamma",
+            "model.encoder.mid_block.attentions.0.proj.bias": "model.encoder.middle.1.proj.bias",
+            "model.encoder.mid_block.attentions.0.proj.weight": "model.encoder.middle.1.proj.weight",
+            "model.encoder.mid_block.attentions.0.to_qkv.bias": "model.encoder.middle.1.to_qkv.bias",
+            "model.encoder.mid_block.attentions.0.to_qkv.weight": "model.encoder.middle.1.to_qkv.weight",
+            "model.encoder.mid_block.resnets.0.conv_shortcut.bias": "model.encoder.middle.0.shortcut.bias",
+            "model.encoder.mid_block.resnets.0.conv_shortcut.weight": "model.encoder.middle.0.shortcut.weight",
+            "model.encoder.mid_block.resnets.0.conv1.bias": "model.encoder.middle.0.residual.2.bias",
+            "model.encoder.mid_block.resnets.0.conv1.weight": "model.encoder.middle.0.residual.2.weight",
+            "model.encoder.mid_block.resnets.0.conv2.bias": "model.encoder.middle.0.residual.6.bias",
+            "model.encoder.mid_block.resnets.0.conv2.weight": "model.encoder.middle.0.residual.6.weight",
+            "model.encoder.mid_block.resnets.0.norm1.gamma": "model.encoder.middle.0.residual.0.gamma",
+            "model.encoder.mid_block.resnets.0.norm2.gamma": "model.encoder.middle.0.residual.3.gamma",
+            "model.encoder.mid_block.resnets.1.conv_shortcut.bias": "model.encoder.middle.2.shortcut.bias",
+            "model.encoder.mid_block.resnets.1.conv_shortcut.weight": "model.encoder.middle.2.shortcut.weight",
+            "model.encoder.mid_block.resnets.1.conv1.bias": "model.encoder.middle.2.residual.2.bias",
+            "model.encoder.mid_block.resnets.1.conv1.weight": "model.encoder.middle.2.residual.2.weight",
+            "model.encoder.mid_block.resnets.1.conv2.bias": "model.encoder.middle.2.residual.6.bias",
+            "model.encoder.mid_block.resnets.1.conv2.weight": "model.encoder.middle.2.residual.6.weight",
+            "model.encoder.mid_block.resnets.1.norm1.gamma": "model.encoder.middle.2.residual.0.gamma",
+            "model.encoder.mid_block.resnets.1.norm2.gamma": "model.encoder.middle.2.residual.3.gamma",
+            "model.encoder.norm_out.gamma": "model.encoder.head.0.gamma",
+            "model.encoder.conv_out.bias": "model.encoder.head.2.bias",
+            "model.encoder.conv_out.weight": "model.encoder.head.2.weight",
+            "model.decoder.conv_in.bias": "model.decoder.conv1.bias",
+            "model.decoder.conv_in.weight": "model.decoder.conv1.weight",
+            "model.decoder.mid_block.attentions.0.norm.gamma": "model.decoder.middle.1.norm.gamma",
+            "model.decoder.mid_block.attentions.0.proj.bias": "model.decoder.middle.1.proj.bias",
+            "model.decoder.mid_block.attentions.0.proj.weight": "model.decoder.middle.1.proj.weight",
+            "model.decoder.mid_block.attentions.0.to_qkv.bias": "model.decoder.middle.1.to_qkv.bias",
+            "model.decoder.mid_block.attentions.0.to_qkv.weight": "model.decoder.middle.1.to_qkv.weight",
+            "model.decoder.mid_block.resnets.0.conv_shortcut.bias": "model.decoder.middle.0.shortcut.bias",
+            "model.decoder.mid_block.resnets.0.conv_shortcut.weight": "model.decoder.middle.0.shortcut.weight",
+            "model.decoder.mid_block.resnets.0.conv1.bias": "model.decoder.middle.0.residual.2.bias",
+            "model.decoder.mid_block.resnets.0.conv1.weight": "model.decoder.middle.0.residual.2.weight",
+            "model.decoder.mid_block.resnets.0.conv2.bias": "model.decoder.middle.0.residual.6.bias",
+            "model.decoder.mid_block.resnets.0.conv2.weight": "model.decoder.middle.0.residual.6.weight",
+            "model.decoder.mid_block.resnets.0.norm1.gamma": "model.decoder.middle.0.residual.0.gamma",
+            "model.decoder.mid_block.resnets.0.norm2.gamma": "model.decoder.middle.0.residual.3.gamma",
+            "model.decoder.mid_block.resnets.1.conv_shortcut.bias": "model.decoder.middle.2.shortcut.bias",
+            "model.decoder.mid_block.resnets.1.conv_shortcut.weight": "model.decoder.middle.2.shortcut.weight",
+            "model.decoder.mid_block.resnets.1.conv1.bias": "model.decoder.middle.2.residual.2.bias",
+            "model.decoder.mid_block.resnets.1.conv1.weight": "model.decoder.middle.2.residual.2.weight",
+            "model.decoder.mid_block.resnets.1.conv2.bias": "model.decoder.middle.2.residual.6.bias",
+            "model.decoder.mid_block.resnets.1.conv2.weight": "model.decoder.middle.2.residual.6.weight",
+            "model.decoder.mid_block.resnets.1.norm1.gamma": "model.decoder.middle.2.residual.0.gamma",
+            "model.decoder.mid_block.resnets.1.norm2.gamma": "model.decoder.middle.2.residual.3.gamma",
+            "model.decoder.up_blocks.0.resnets.0.conv_shortcut.bias": "model.decoder.upsamples.0.shortcut.bias",
+            "model.decoder.up_blocks.0.resnets.0.conv_shortcut.weight": "model.decoder.upsamples.0.shortcut.weight",
+            "model.decoder.up_blocks.0.resnets.0.conv1.bias": "model.decoder.upsamples.0.residual.2.bias",
+            "model.decoder.up_blocks.0.resnets.0.conv1.weight": "model.decoder.upsamples.0.residual.2.weight",
+            "model.decoder.up_blocks.0.resnets.0.conv2.bias": "model.decoder.upsamples.0.residual.6.bias",
+            "model.decoder.up_blocks.0.resnets.0.conv2.weight": "model.decoder.upsamples.0.residual.6.weight",
+            "model.decoder.up_blocks.0.resnets.0.norm1.gamma": "model.decoder.upsamples.0.residual.0.gamma",
+            "model.decoder.up_blocks.0.resnets.0.norm2.gamma": "model.decoder.upsamples.0.residual.3.gamma",
+            "model.decoder.up_blocks.0.resnets.1.conv_shortcut.bias": "model.decoder.upsamples.1.shortcut.bias",
+            "model.decoder.up_blocks.0.resnets.1.conv_shortcut.weight": "model.decoder.upsamples.1.shortcut.weight",
+            "model.decoder.up_blocks.0.resnets.1.conv1.bias": "model.decoder.upsamples.1.residual.2.bias",
+            "model.decoder.up_blocks.0.resnets.1.conv1.weight": "model.decoder.upsamples.1.residual.2.weight",
+            "model.decoder.up_blocks.0.resnets.1.conv2.bias": "model.decoder.upsamples.1.residual.6.bias",
+            "model.decoder.up_blocks.0.resnets.1.conv2.weight": "model.decoder.upsamples.1.residual.6.weight",
+            "model.decoder.up_blocks.0.resnets.1.norm1.gamma": "model.decoder.upsamples.1.residual.0.gamma",
+            "model.decoder.up_blocks.0.resnets.1.norm2.gamma": "model.decoder.upsamples.1.residual.3.gamma",
+            "model.decoder.up_blocks.0.resnets.2.conv_shortcut.bias": "model.decoder.upsamples.2.shortcut.bias",
+            "model.decoder.up_blocks.0.resnets.2.conv_shortcut.weight": "model.decoder.upsamples.2.shortcut.weight",
+            "model.decoder.up_blocks.0.resnets.2.conv1.bias": "model.decoder.upsamples.2.residual.2.bias",
+            "model.decoder.up_blocks.0.resnets.2.conv1.weight": "model.decoder.upsamples.2.residual.2.weight",
+            "model.decoder.up_blocks.0.resnets.2.conv2.bias": "model.decoder.upsamples.2.residual.6.bias",
+            "model.decoder.up_blocks.0.resnets.2.conv2.weight": "model.decoder.upsamples.2.residual.6.weight",
+            "model.decoder.up_blocks.0.resnets.2.norm1.gamma": "model.decoder.upsamples.2.residual.0.gamma",
+            "model.decoder.up_blocks.0.resnets.2.norm2.gamma": "model.decoder.upsamples.2.residual.3.gamma",
+            "model.decoder.up_blocks.0.upsamplers.0.resample.1.bias": "model.decoder.upsamples.3.resample.1.bias",
+            "model.decoder.up_blocks.0.upsamplers.0.resample.1.weight": "model.decoder.upsamples.3.resample.1.weight",
+            "model.decoder.up_blocks.0.upsamplers.0.time_conv.bias": "model.decoder.upsamples.3.time_conv.bias",
+            "model.decoder.up_blocks.0.upsamplers.0.time_conv.weight": "model.decoder.upsamples.3.time_conv.weight",
+            "model.decoder.up_blocks.1.resnets.0.conv_shortcut.bias": "model.decoder.upsamples.4.shortcut.bias",
+            "model.decoder.up_blocks.1.resnets.0.conv_shortcut.weight": "model.decoder.upsamples.4.shortcut.weight",
+            "model.decoder.up_blocks.1.resnets.0.conv1.bias": "model.decoder.upsamples.4.residual.2.bias",
+            "model.decoder.up_blocks.1.resnets.0.conv1.weight": "model.decoder.upsamples.4.residual.2.weight",
+            "model.decoder.up_blocks.1.resnets.0.conv2.bias": "model.decoder.upsamples.4.residual.6.bias",
+            "model.decoder.up_blocks.1.resnets.0.conv2.weight": "model.decoder.upsamples.4.residual.6.weight",
+            "model.decoder.up_blocks.1.resnets.0.norm1.gamma": "model.decoder.upsamples.4.residual.0.gamma",
+            "model.decoder.up_blocks.1.resnets.0.norm2.gamma": "model.decoder.upsamples.4.residual.3.gamma",
+            "model.decoder.up_blocks.1.resnets.1.conv_shortcut.bias": "model.decoder.upsamples.5.shortcut.bias",
+            "model.decoder.up_blocks.1.resnets.1.conv_shortcut.weight": "model.decoder.upsamples.5.shortcut.weight",
+            "model.decoder.up_blocks.1.resnets.1.conv1.bias": "model.decoder.upsamples.5.residual.2.bias",
+            "model.decoder.up_blocks.1.resnets.1.conv1.weight": "model.decoder.upsamples.5.residual.2.weight",
+            "model.decoder.up_blocks.1.resnets.1.conv2.bias": "model.decoder.upsamples.5.residual.6.bias",
+            "model.decoder.up_blocks.1.resnets.1.conv2.weight": "model.decoder.upsamples.5.residual.6.weight",
+            "model.decoder.up_blocks.1.resnets.1.norm1.gamma": "model.decoder.upsamples.5.residual.0.gamma",
+            "model.decoder.up_blocks.1.resnets.1.norm2.gamma": "model.decoder.upsamples.5.residual.3.gamma",
+            "model.decoder.up_blocks.1.resnets.2.conv_shortcut.bias": "model.decoder.upsamples.6.shortcut.bias",
+            "model.decoder.up_blocks.1.resnets.2.conv_shortcut.weight": "model.decoder.upsamples.6.shortcut.weight",
+            "model.decoder.up_blocks.1.resnets.2.conv1.bias": "model.decoder.upsamples.6.residual.2.bias",
+            "model.decoder.up_blocks.1.resnets.2.conv1.weight": "model.decoder.upsamples.6.residual.2.weight",
+            "model.decoder.up_blocks.1.resnets.2.conv2.bias": "model.decoder.upsamples.6.residual.6.bias",
+            "model.decoder.up_blocks.1.resnets.2.conv2.weight": "model.decoder.upsamples.6.residual.6.weight",
+            "model.decoder.up_blocks.1.resnets.2.norm1.gamma": "model.decoder.upsamples.6.residual.0.gamma",
+            "model.decoder.up_blocks.1.resnets.2.norm2.gamma": "model.decoder.upsamples.6.residual.3.gamma",
+            "model.decoder.up_blocks.1.upsamplers.0.resample.1.bias": "model.decoder.upsamples.7.resample.1.bias",
+            "model.decoder.up_blocks.1.upsamplers.0.resample.1.weight": "model.decoder.upsamples.7.resample.1.weight",
+            "model.decoder.up_blocks.1.upsamplers.0.time_conv.bias": "model.decoder.upsamples.7.time_conv.bias",
+            "model.decoder.up_blocks.1.upsamplers.0.time_conv.weight": "model.decoder.upsamples.7.time_conv.weight",
+            "model.decoder.up_blocks.2.resnets.0.conv_shortcut.bias": "model.decoder.upsamples.8.shortcut.bias",
+            "model.decoder.up_blocks.2.resnets.0.conv_shortcut.weight": "model.decoder.upsamples.8.shortcut.weight",
+            "model.decoder.up_blocks.2.resnets.0.conv1.bias": "model.decoder.upsamples.8.residual.2.bias",
+            "model.decoder.up_blocks.2.resnets.0.conv1.weight": "model.decoder.upsamples.8.residual.2.weight",
+            "model.decoder.up_blocks.2.resnets.0.conv2.bias": "model.decoder.upsamples.8.residual.6.bias",
+            "model.decoder.up_blocks.2.resnets.0.conv2.weight": "model.decoder.upsamples.8.residual.6.weight",
+            "model.decoder.up_blocks.2.resnets.0.norm1.gamma": "model.decoder.upsamples.8.residual.0.gamma",
+            "model.decoder.up_blocks.2.resnets.0.norm2.gamma": "model.decoder.upsamples.8.residual.3.gamma",
+            "model.decoder.up_blocks.2.resnets.1.conv_shortcut.bias": "model.decoder.upsamples.9.shortcut.bias",
+            "model.decoder.up_blocks.2.resnets.1.conv_shortcut.weight": "model.decoder.upsamples.9.shortcut.weight",
+            "model.decoder.up_blocks.2.resnets.1.conv1.bias": "model.decoder.upsamples.9.residual.2.bias",
+            "model.decoder.up_blocks.2.resnets.1.conv1.weight": "model.decoder.upsamples.9.residual.2.weight",
+            "model.decoder.up_blocks.2.resnets.1.conv2.bias": "model.decoder.upsamples.9.residual.6.bias",
+            "model.decoder.up_blocks.2.resnets.1.conv2.weight": "model.decoder.upsamples.9.residual.6.weight",
+            "model.decoder.up_blocks.2.resnets.1.norm1.gamma": "model.decoder.upsamples.9.residual.0.gamma",
+            "model.decoder.up_blocks.2.resnets.1.norm2.gamma": "model.decoder.upsamples.9.residual.3.gamma",
+            "model.decoder.up_blocks.2.resnets.2.conv_shortcut.bias": "model.decoder.upsamples.10.shortcut.bias",
+            "model.decoder.up_blocks.2.resnets.2.conv_shortcut.weight": "model.decoder.upsamples.10.shortcut.weight",
+            "model.decoder.up_blocks.2.resnets.2.conv1.bias": "model.decoder.upsamples.10.residual.2.bias",
+            "model.decoder.up_blocks.2.resnets.2.conv1.weight": "model.decoder.upsamples.10.residual.2.weight",
+            "model.decoder.up_blocks.2.resnets.2.conv2.bias": "model.decoder.upsamples.10.residual.6.bias",
+            "model.decoder.up_blocks.2.resnets.2.conv2.weight": "model.decoder.upsamples.10.residual.6.weight",
+            "model.decoder.up_blocks.2.resnets.2.norm1.gamma": "model.decoder.upsamples.10.residual.0.gamma",
+            "model.decoder.up_blocks.2.resnets.2.norm2.gamma": "model.decoder.upsamples.10.residual.3.gamma",
+            "model.decoder.up_blocks.2.upsamplers.0.resample.1.bias": "model.decoder.upsamples.11.resample.1.bias",
+            "model.decoder.up_blocks.2.upsamplers.0.resample.1.weight": "model.decoder.upsamples.11.resample.1.weight",
+            "model.decoder.up_blocks.3.resnets.0.conv_shortcut.bias": "model.decoder.upsamples.12.shortcut.bias",
+            "model.decoder.up_blocks.3.resnets.0.conv_shortcut.weight": "model.decoder.upsamples.12.shortcut.weight",
+            "model.decoder.up_blocks.3.resnets.0.conv1.bias": "model.decoder.upsamples.12.residual.2.bias",
+            "model.decoder.up_blocks.3.resnets.0.conv1.weight": "model.decoder.upsamples.12.residual.2.weight",
+            "model.decoder.up_blocks.3.resnets.0.conv2.bias": "model.decoder.upsamples.12.residual.6.bias",
+            "model.decoder.up_blocks.3.resnets.0.conv2.weight": "model.decoder.upsamples.12.residual.6.weight",
+            "model.decoder.up_blocks.3.resnets.0.norm1.gamma": "model.decoder.upsamples.12.residual.0.gamma",
+            "model.decoder.up_blocks.3.resnets.0.norm2.gamma": "model.decoder.upsamples.12.residual.3.gamma",
+            "model.decoder.up_blocks.3.resnets.1.conv_shortcut.bias": "model.decoder.upsamples.13.shortcut.bias",
+            "model.decoder.up_blocks.3.resnets.1.conv_shortcut.weight": "model.decoder.upsamples.13.shortcut.weight",
+            "model.decoder.up_blocks.3.resnets.1.conv1.bias": "model.decoder.upsamples.13.residual.2.bias",
+            "model.decoder.up_blocks.3.resnets.1.conv1.weight": "model.decoder.upsamples.13.residual.2.weight",
+            "model.decoder.up_blocks.3.resnets.1.conv2.bias": "model.decoder.upsamples.13.residual.6.bias",
+            "model.decoder.up_blocks.3.resnets.1.conv2.weight": "model.decoder.upsamples.13.residual.6.weight",
+            "model.decoder.up_blocks.3.resnets.1.norm1.gamma": "model.decoder.upsamples.13.residual.0.gamma",
+            "model.decoder.up_blocks.3.resnets.1.norm2.gamma": "model.decoder.upsamples.13.residual.3.gamma",
+            "model.decoder.up_blocks.3.resnets.2.conv_shortcut.bias": "model.decoder.upsamples.14.shortcut.bias",
+            "model.decoder.up_blocks.3.resnets.2.conv_shortcut.weight": "model.decoder.upsamples.14.shortcut.weight",
+            "model.decoder.up_blocks.3.resnets.2.conv1.bias": "model.decoder.upsamples.14.residual.2.bias",
+            "model.decoder.up_blocks.3.resnets.2.conv1.weight": "model.decoder.upsamples.14.residual.2.weight",
+            "model.decoder.up_blocks.3.resnets.2.conv2.bias": "model.decoder.upsamples.14.residual.6.bias",
+            "model.decoder.up_blocks.3.resnets.2.conv2.weight": "model.decoder.upsamples.14.residual.6.weight",
+            "model.decoder.up_blocks.3.resnets.2.norm1.gamma": "model.decoder.upsamples.14.residual.0.gamma",
+            "model.decoder.up_blocks.3.resnets.2.norm2.gamma": "model.decoder.upsamples.14.residual.3.gamma",
+            "model.decoder.norm_out.gamma": "model.decoder.head.0.gamma",
+            "model.decoder.conv_out.bias": "model.decoder.head.2.bias",
+            "model.decoder.conv_out.weight": "model.decoder.head.2.weight"
+        }
+    }
+}