update readme

update v1.1.0

.gitignore

@@ -9,12 +9,12 @@
 *.bin
 *.idea
 *.csv
+cache
 build
 dist
 dev
 scepter.egg-info
 .readthedocs.yml
-1.9
-MANIFEST.in
 *resources
 *.ipynb_checkpoints*
+*.vscode

asset/workflow/sdxl_base.json

@@ -0,0 +1,163 @@
+{
+  "last_node_id": 4,
+  "last_link_id": 3,
+  "nodes": [
+    {
+      "id": 2,
+      "type": "PreviewImage",
+      "pos": {
+        "0": 937,
+        "1": 243
+      },
+      "size": {
+        "0": 284.03680419921875,
+        "1": 246
+      },
+      "flags": {
+        "collapsed": false
+      },
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 1
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "Node name for S&R": "PreviewImage"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 4,
+      "type": "ParameterNode",
+      "pos": {
+        "0": 0,
+        "1": 244
+      },
+      "size": {
+        "0": 311.9849853515625,
+        "1": 236.4765625
+      },
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            3
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ParameterNode"
+      },
+      "widgets_values": [
+        "ddim",
+        50,
+        5,
+        0.5,
+        "trailing",
+        1024,
+        1024,
+        2024
+      ]
+    },
+    {
+      "id": 1,
+      "type": "ModelNode",
+      "pos": {
+        "0": 430,
+        "1": 243
+      },
+      "size": {
+        "0": 400,
+        "1": 234
+      },
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "parameters",
+          "type": "CONDITIONING",
+          "link": 3,
+          "shape": 7
+        },
+        {
+          "name": "mantras",
+          "type": "CONDITIONING",
+          "link": null,
+          "shape": 7
+        },
+        {
+          "name": "tuners",
+          "type": "CONDITIONING",
+          "link": null,
+          "shape": 7
+        },
+        {
+          "name": "controls",
+          "type": "CONDITIONING",
+          "link": null,
+          "shape": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            1
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ModelNode"
+      },
+      "widgets_values": [
+        "SD_XL1.0",
+        "ModelScope",
+        "a cat",
+        ""
+      ]
+    }
+  ],
+  "links": [
+    [
+      1,
+      1,
+      0,
+      2,
+      0,
+      "IMAGE"
+    ],
+    [
+      3,
+      4,
+      0,
+      1,
+      0,
+      "CONDITIONING"
+    ]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "ds": {
+      "scale": 0.9229599817706423,
+      "offset": [
+        128.91466506592707,
+        43.07074577975898
+      ]
+    }
+  },
+  "version": 0.4
+}

asset/workflow/sdxl_base_mantra.json

@@ -0,0 +1,202 @@
+{
+  "last_node_id": 5,
+  "last_link_id": 4,
+  "nodes": [
+    {
+      "id": 2,
+      "type": "PreviewImage",
+      "pos": {
+        "0": 937,
+        "1": 243
+      },
+      "size": {
+        "0": 284.03680419921875,
+        "1": 246
+      },
+      "flags": {
+        "collapsed": false
+      },
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 1
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "Node name for S&R": "PreviewImage"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 4,
+      "type": "ParameterNode",
+      "pos": {
+        "0": 0,
+        "1": 244
+      },
+      "size": {
+        "0": 311.9849853515625,
+        "1": 236.4765625
+      },
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            3
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ParameterNode"
+      },
+      "widgets_values": [
+        "ddim",
+        50,
+        5,
+        0.5,
+        "trailing",
+        1024,
+        1024,
+        2024
+      ]
+    },
+    {
+      "id": 1,
+      "type": "ModelNode",
+      "pos": {
+        "0": 430,
+        "1": 243
+      },
+      "size": {
+        "0": 400,
+        "1": 234
+      },
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "parameters",
+          "type": "CONDITIONING",
+          "link": 3,
+          "shape": 7
+        },
+        {
+          "name": "mantras",
+          "type": "CONDITIONING",
+          "link": 4,
+          "shape": 7
+        },
+        {
+          "name": "tuners",
+          "type": "CONDITIONING",
+          "link": null,
+          "shape": 7
+        },
+        {
+          "name": "controls",
+          "type": "CONDITIONING",
+          "link": null,
+          "shape": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            1
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ModelNode"
+      },
+      "widgets_values": [
+        "SD_XL1.0",
+        "ModelScope",
+        "a cat",
+        ""
+      ]
+    },
+    {
+      "id": 5,
+      "type": "MantrasNode",
+      "pos": {
+        "0": 3,
+        "1": 547
+      },
+      "size": {
+        "0": 315,
+        "1": 58
+      },
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            4
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MantrasNode"
+      },
+      "widgets_values": [
+        "Flat 2D Art"
+      ]
+    }
+  ],
+  "links": [
+    [
+      1,
+      1,
+      0,
+      2,
+      0,
+      "IMAGE"
+    ],
+    [
+      3,
+      4,
+      0,
+      1,
+      0,
+      "CONDITIONING"
+    ],
+    [
+      4,
+      5,
+      0,
+      1,
+      1,
+      "CONDITIONING"
+    ]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "ds": {
+      "scale": 1.0152559799477068,
+      "offset": [
+        58.35814123566077,
+        -65.12749162217047
+      ]
+    }
+  },
+  "version": 0.4
+}

asset/workflow/sdxl_base_mantra_tuner.json

@@ -0,0 +1,242 @@
+{
+  "last_node_id": 6,
+  "last_link_id": 5,
+  "nodes": [
+    {
+      "id": 2,
+      "type": "PreviewImage",
+      "pos": {
+        "0": 937,
+        "1": 243
+      },
+      "size": {
+        "0": 284.03680419921875,
+        "1": 246
+      },
+      "flags": {
+        "collapsed": false
+      },
+      "order": 4,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 1
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "Node name for S&R": "PreviewImage"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 4,
+      "type": "ParameterNode",
+      "pos": {
+        "0": 0,
+        "1": 244
+      },
+      "size": {
+        "0": 311.9849853515625,
+        "1": 236.4765625
+      },
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            3
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ParameterNode"
+      },
+      "widgets_values": [
+        "ddim",
+        50,
+        5,
+        0.5,
+        "trailing",
+        1024,
+        1024,
+        2024
+      ]
+    },
+    {
+      "id": 5,
+      "type": "MantrasNode",
+      "pos": {
+        "0": 3,
+        "1": 547
+      },
+      "size": {
+        "0": 315,
+        "1": 58
+      },
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            4
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MantrasNode"
+      },
+      "widgets_values": [
+        "Flat 2D Art"
+      ]
+    },
+    {
+      "id": 1,
+      "type": "ModelNode",
+      "pos": {
+        "0": 430,
+        "1": 243
+      },
+      "size": {
+        "0": 400,
+        "1": 234
+      },
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "parameters",
+          "type": "CONDITIONING",
+          "link": 3,
+          "shape": 7
+        },
+        {
+          "name": "mantras",
+          "type": "CONDITIONING",
+          "link": 4,
+          "shape": 7
+        },
+        {
+          "name": "tuners",
+          "type": "CONDITIONING",
+          "link": 5,
+          "shape": 7
+        },
+        {
+          "name": "controls",
+          "type": "CONDITIONING",
+          "link": null,
+          "shape": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            1
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ModelNode"
+      },
+      "widgets_values": [
+        "SD_XL1.0",
+        "ModelScope",
+        "a cat",
+        ""
+      ]
+    },
+    {
+      "id": 6,
+      "type": "TunerNode",
+      "pos": {
+        "0": 8,
+        "1": 680
+      },
+      "size": {
+        "0": 315,
+        "1": 82
+      },
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            5
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "TunerNode"
+      },
+      "widgets_values": [
+        "SD_XL1.0_Pencil Sketch Drawing",
+        1
+      ]
+    }
+  ],
+  "links": [
+    [
+      1,
+      1,
+      0,
+      2,
+      0,
+      "IMAGE"
+    ],
+    [
+      3,
+      4,
+      0,
+      1,
+      0,
+      "CONDITIONING"
+    ],
+    [
+      4,
+      5,
+      0,
+      1,
+      1,
+      "CONDITIONING"
+    ],
+    [
+      5,
+      6,
+      0,
+      1,
+      2,
+      "CONDITIONING"
+    ]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "ds": {
+      "scale": 0.6934334949441366,
+      "offset": [
+        450.0923948318967,
+        30.783653239467935
+      ]
+    }
+  },
+  "version": 0.4
+}

asset/workflow/sdxl_base_mantra_tuner_control.json

@@ -0,0 +1,366 @@
+{
+  "last_node_id": 11,
+  "last_link_id": 9,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "ModelNode",
+      "pos": {
+        "0": 439,
+        "1": 123
+      },
+      "size": {
+        "0": 400,
+        "1": 234
+      },
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "parameters",
+          "type": "CONDITIONING",
+          "link": 3,
+          "shape": 7
+        },
+        {
+          "name": "mantras",
+          "type": "CONDITIONING",
+          "link": 4,
+          "shape": 7
+        },
+        {
+          "name": "tuners",
+          "type": "CONDITIONING",
+          "link": 5,
+          "shape": 7
+        },
+        {
+          "name": "controls",
+          "type": "CONDITIONING",
+          "link": 6,
+          "shape": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            1
+          ],
+          "slot_index": 0
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ModelNode"
+      },
+      "widgets_values": [
+        "SD_XL1.0",
+        "ModelScope",
+        "a cat",
+        ""
+      ]
+    },
+    {
+      "id": 5,
+      "type": "MantrasNode",
+      "pos": {
+        "0": 13,
+        "1": 336
+      },
+      "size": {
+        "0": 315,
+        "1": 58
+      },
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            4
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MantrasNode"
+      },
+      "widgets_values": [
+        "Flat 2D Art"
+      ]
+    },
+    {
+      "id": 6,
+      "type": "TunerNode",
+      "pos": {
+        "0": 12,
+        "1": 455
+      },
+      "size": {
+        "0": 315,
+        "1": 82
+      },
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            5
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "TunerNode"
+      },
+      "widgets_values": [
+        "SD_XL1.0_Pencil Sketch Drawing",
+        1
+      ]
+    },
+    {
+      "id": 11,
+      "type": "NoteNode",
+      "pos": {
+        "0": 845,
+        "1": 506
+      },
+      "size": {
+        "0": 398.3059387207031,
+        "1": 210.83267211914062
+      },
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "properties": {
+        "Node name for S&R": "NoteNode"
+      },
+      "widgets_values": [
+        "This is a sample for quickly setting up ComfyUI using the Scepter open-source library:\n\n1) The first run involves downloading the models. By default, we will automatically pull models from ModelScope. The initial download may take some time, and you can also adjust the model source to change the model address.\n\n2) Currently, it supports various base models, basic settings for some inference hyperparameters, mantra settings, tuning model settings, and conditional generation node.\n\n3) In the future, we will gradually integrate interesting features to enrich the use cases.\n\n"
+      ]
+    },
+    {
+      "id": 8,
+      "type": "LoadImage",
+      "pos": {
+        "0": 471,
+        "1": 451
+      },
+      "size": {
+        "0": 315,
+        "1": 314
+      },
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            7
+          ]
+        },
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoadImage"
+      },
+      "widgets_values": [
+        "cat.jpg",
+        "image"
+      ]
+    },
+    {
+      "id": 7,
+      "type": "ControlNode",
+      "pos": {
+        "0": 15,
+        "1": 600
+      },
+      "size": {
+        "0": 330,
+        "1": 198
+      },
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "source_image",
+          "type": "IMAGE",
+          "link": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            6
+          ]
+        },
+        {
+          "name": "Control Image",
+          "type": "IMAGE",
+          "links": [],
+          "slot_index": 1
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ControlNode"
+      },
+      "widgets_values": [
+        "SD_XL1.0_color",
+        "Color",
+        "CenterCrop",
+        1,
+        1024,
+        1024
+      ]
+    },
+    {
+      "id": 2,
+      "type": "PreviewImage",
+      "pos": {
+        "0": 959,
+        "1": 121
+      },
+      "size": {
+        "0": 284.03680419921875,
+        "1": 246
+      },
+      "flags": {
+        "collapsed": false
+      },
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 1
+        }
+      ],
+      "outputs": [],
+      "properties": {
+        "Node name for S&R": "PreviewImage"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 4,
+      "type": "ParameterNode",
+      "pos": {
+        "0": 13,
+        "1": 43
+      },
+      "size": {
+        "0": 311.9849853515625,
+        "1": 236.4765625
+      },
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "Result",
+          "type": "CONDITIONING",
+          "links": [
+            3
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ParameterNode"
+      },
+      "widgets_values": [
+        "ddim",
+        50,
+        5,
+        0.5,
+        "trailing",
+        1024,
+        1024,
+        2024
+      ]
+    }
+  ],
+  "links": [
+    [
+      1,
+      1,
+      0,
+      2,
+      0,
+      "IMAGE"
+    ],
+    [
+      3,
+      4,
+      0,
+      1,
+      0,
+      "CONDITIONING"
+    ],
+    [
+      4,
+      5,
+      0,
+      1,
+      1,
+      "CONDITIONING"
+    ],
+    [
+      5,
+      6,
+      0,
+      1,
+      2,
+      "CONDITIONING"
+    ],
+    [
+      6,
+      7,
+      0,
+      1,
+      3,
+      "CONDITIONING"
+    ],
+    [
+      7,
+      8,
+      0,
+      7,
+      0,
+      "IMAGE"
+    ]
+  ],
+  "groups": [],
+  "config": {},
+  "extra": {
+    "ds": {
+      "scale": 0.7627768444385667,
+      "offset": [
+        254.62860058494985,
+        87.05734194144193
+      ]
+    }
+  },
+  "version": 0.4
+}

docs/en/conf.py

@@ -1,4 +1,5 @@
 # -*- coding: utf-8 -*-
+# Copyright (c) Alibaba, Inc. and its affiliates.
 # Configuration file for the Sphinx documentation builder.
 #
 # This file only contains a selection of the most common options. For a full

docs/en/tasks/stylebooth.md

@@ -29,8 +29,21 @@ Given an original image, image editing aims to generate an image that align with
     <td><img src="../../../asset/images/stylebooth/retrogame.jpeg" width="240"></td>
     <td><img src="../../../asset/images/stylebooth/vangogh.jpeg" width="240"></td>
   </tr>
+  <tr>
+    <td><strong>Origin Image</strong></td>
+    <td><strong>Lowpoly</strong></td>
+    <td><strong>Colored Pencil Art</strong></td>
+    <td><strong>Watercolor</strong></td>
+    <td><strong>misc-disco</strong></td>
+  </tr>
+  <tr>
+    <td><img src="../../../asset/images/stylebooth/mountain.jpg" width="240"></td>
+    <td><img src="../../../asset/images/stylebooth/lowpoly.jpg" width="240"></td>
+    <td><img src="../../../asset/images/stylebooth/colorpencil.jpeg" width="240"></td>
+    <td><img src="../../../asset/images/stylebooth/watercolor.jpeg" width="240"></td>
+    <td><img src="../../../asset/images/stylebooth/disco.jpeg" width="240"></td>
+  </tr>
 </table>
-
 ## Features
 
 | **Text-Based** | **Exemplar-Based** |
@@ -43,7 +56,7 @@ Given an original image, image editing aims to generate an image that align with
 - More models will be released in the future.
 
 ## Run StyleBooth
-- Code implementation: See model configuration and code based on [🪄SCEPTER](https://github.com/modelscope/scepter).
+- Code implementation: See model configuration and code based on [🪄SCEPTER](https://github.com/modelscope/scepter/blob/main/docs/en/tasks/stylebooth.md).
 
 - Demo: Try [🖥️SCEPTER Studio](https://github.com/modelscope/scepter/tree/main?tab=readme-ov-file#%EF%B8%8F-scepter-studio).
 
@@ -92,7 +105,7 @@ class DiffusionInferenceTest(unittest.TestCase):
         output = diff_infer({'prompt': 'Let this image be in the style of sai-lowpoly'},
                             style_edit_image=Image.open('asset/images/inpainting_text_ref/ex4_scene_im.jpg'),
                             style_guide_scale_text=7.5,
-                            style_guide_scale_image=0.5)
+                            style_guide_scale_image=1.5)
         save_path = os.path.join(self.tmp_dir,
                                  'stylebooth_test_lowpoly_cute_dog.png')
         save_image(output['images'], save_path)
@@ -101,3 +114,17 @@ class DiffusionInferenceTest(unittest.TestCase):
 if __name__ == '__main__':
     unittest.main()
 ```
+
+## StyleTuner and De-StyleTuner.
+
+### Base I2I Model.
+
+For style and de-style tuning, we use a private high-resolution I2I model trained with [InstructPix2Pix dataset](https://instruct-pix2pix.eecs.berkeley.edu/) as base model. However, one can try the same tunning process using this [yaml](https://github.com/modelscope/scepter/blob/main/scepter/methods/edit/edit_512_lora.yaml) based on any other I2I model, such as StyleBooth (shown in this yaml), [InstructPix2Pix](https://github.com/timothybrooks/instruct-pix2pix) or [MagicBrush](https://github.com/OSU-NLP-Group/MagicBrush).
+
+### Training Data.
+
+Please check the zips for correct format: [De-Text](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets%2Fdetext.zip), [Image2Hed](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets%2Fhed_pair.zip), [Image2Depth](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets%2Fimage2depth.zip), [Depth2Image](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets%2Fdepth2image.zip).
+
+### Launch.
+
+See [code](https://github.com/modelscope/scepter/blob/cbbef4a2da5b66fc33b9f8ece7f2fb4aac9d6e3c/tests/tools/test_train.py#L220) for more information.

docs/en/tutorials/dataset.md

@@ -5,7 +5,7 @@ Below are examples for each format, illustrating their details and basic usage.
 
 ## Modelscope Format
 
-We use a [custom-stylized dataset](https://modelscope.cn/datasets/damo/style_custom_dataset/summary), which included classes 3D, anime, flat illustration, oil painting, sketch, and watercolor, each with 30 image-text pairs.
+We use a [custom-stylized dataset](https://modelscope.cn/datasets/iic/style_custom_dataset/summary), which included classes 3D, anime, flat illustration, oil painting, sketch, and watercolor, each with 30 image-text pairs.
 
 ```python
 # pip install modelscope
@@ -16,7 +16,11 @@ print(next(iter(ms_train_dataset)))
 
 ## CSV Format
 
-For the data format used by SCEPTER Studio, please refer to [3D_example_csv.zip](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets/3D_example_csv.zip).
+For the data format used by SCEPTER Studio, please refer to [3D_example_csv.zip](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets/3D_example_csv.zip) and [hed_pair.zip](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets%2Fhed_pair.zip).
+```shell
+mkdir -p cache/datasets/ && wget 'https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets/3D_example_csv.zip' -O cache/datasets/3D_example_csv.zip && unzip cache/datasets/3D_example_csv.zip -d cache/datasets/ && rm cache/datasets/3D_example_csv.zip
+mkdir -p cache/datasets/ && wget 'https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets/hed_pair.zip' -O cache/datasets/hed_pair.zip && unzip cache/datasets/hed_pair.zip -d cache/datasets/ && rm cache/datasets/hed_pair.zip
+```
 
 ## TXT Format
 
@@ -24,3 +28,4 @@ To facilitate starting training in command-line mode, you can use a dataset in t
 
 ```shell
 mkdir -p cache/datasets/ && wget 'https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=datasets/3D_example_txt.zip' -O cache/datasets/3D_example_txt.zip && unzip cache/datasets/3D_example_txt.zip -d cache/datasets/ && rm cache/datasets/3D_example_txt.zip
+```

docs/en/tutorials/inference.md

@@ -40,6 +40,6 @@ python scepter/tools/run_inference.py --cfg scepter/methods/scedit/t2i/sd15_512_
 
 - SCEdit
 ```shell
-python scepter/tools/run_inference.py --cfg scepter/methods/scedit/ctr/sd21_768_sce_ctr_canny.yaml --num_samples 1 --prompt 'a single flower is shown in front of a tree' --save_folder 'test_flower_canny' --image_size 768 --task control --image 'asset/images/flower.jpg' --control_mode canny --pretrained_model ms://damo/scepter_scedit@controllable_model/SD2.1/canny_control/0_SwiftSCETuning/pytorch_model.bin   # canny
-python scepter/tools/run_inference.py --cfg scepter/methods/scedit/ctr/sd21_768_sce_ctr_pose.yaml --num_samples 1 --prompt 'super mario' --save_folder 'test_mario_pose' --image_size 768 --task control --image 'asset/images/pose_source.png' --control_mode source --pretrained_model ms://damo/scepter_scedit@controllable_model/SD2.1/pose_control/0_SwiftSCETuning/pytorch_model.bin   # pose
+python scepter/tools/run_inference.py --cfg scepter/methods/scedit/ctr/sd21_768_sce_ctr_canny.yaml --num_samples 1 --prompt 'a single flower is shown in front of a tree' --save_folder 'test_flower_canny' --image_size 768 --task control --image 'asset/images/flower.jpg' --control_mode canny --pretrained_model ms://iic/scepter_scedit@controllable_model/SD2.1/canny_control/0_SwiftSCETuning/pytorch_model.bin   # canny
+python scepter/tools/run_inference.py --cfg scepter/methods/scedit/ctr/sd21_768_sce_ctr_pose.yaml --num_samples 1 --prompt 'super mario' --save_folder 'test_mario_pose' --image_size 768 --task control --image 'asset/images/pose_source.png' --control_mode source --pretrained_model ms://iic/scepter_scedit@controllable_model/SD2.1/pose_control/0_SwiftSCETuning/pytorch_model.bin   # pose
 ```

docs/zh_cn/conf.py

@@ -1,4 +1,5 @@
 # -*- coding: utf-8 -*-
+# Copyright (c) Alibaba, Inc. and its affiliates.
 # Configuration file for the Sphinx documentation builder.
 #
 # This file only contains a selection of the most common options. For a full

environment.yaml

@@ -6,4 +6,5 @@ dependencies:
   - pip>=20.3
   - numpy>=1.23.1
   - pip:
+      - -r requirements/recommended.txt
       - -r requirements.txt

example/classifier/classifier_dataset.py

@@ -2,7 +2,6 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 import numpy as np
 import torchvision
-
 from scepter.modules.data.dataset.base_dataset import BaseDataset
 from scepter.modules.data.dataset.registry import DATASETS
 from scepter.modules.utils.config import dict_to_yaml

install.py

@@ -0,0 +1,22 @@
+# -*- coding: utf-8 -*-
+# Copyright (c) Alibaba, Inc. and its affiliates.
+import os
+import shutil
+import subprocess
+import sys
+
+if sys.argv[0] == 'install.py':
+    sys.path.append('.')   # for portable version
+
+source_folder = os.path.join(os.path.dirname(__file__), "scepter/workflow")
+current_dir = os.path.dirname(__file__)
+destination_folder = os.path.join(os.path.dirname(current_dir), "ComfyUI-Scepter")
+
+if not os.path.exists(destination_folder):
+    shutil.copytree(source_folder, destination_folder)
+    print(f"{os.path.abspath(source_folder)} copy to {os.path.abspath(destination_folder)} success!")
+else:
+    print(f"{os.path.abspath(destination_folder)} exist.")
+
+# pip install scepter
+subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'scepter'])

readme.md

@@ -14,10 +14,15 @@ SCEPTER integrates popular community-driven implementations as well as proprieta
 SCEPTER offers 3 core components:
 - [Generative training and inference framework](#tutorials)
 - [Easy implementation of popular approaches](#currently-supported-approaches)
-- [Interactive user interface: SCEPTER Studio](#launch)
+- [Interactive user interface: SCEPTER Studio & Comfy UI](#launch)
 
 
 ## 🎉 News
+- [🔥🔥🔥2024.10]: We are pleased to announce the release of the code for [ACE](https://arxiv.org/abs/2410.00086), supporting Customized Training / Comfy UI Workflow / gradio-based ChatBot Interface. The detailed documents can be found at [ACE repo](https://github.com/ali-vilab/ACE.git).
+- [2024.10]: Support for inference and tuning with [FLUX](https://huggingface.co/black-forest-labs/FLUX.1-dev), as well as for building [ComfyUI](https://github.com/comfyanonymous/ComfyUI) workflows using this framework.
+- [2024.09]: We introduce **ACE**, an **A**ll-round **C**reator and **E**ditor adept at executing a diverse array of image editing tasks tailored to your specifications. Built upon the cutting-edge Diffusion Transformer architecture, ACE has been extensively trained on a comprehensive dataset to seamlessly interpret and execute any natural language instruction. For further information, please consult the [project page](https://ali-vilab.github.io/ace-page/).
+- [2024.07]: Support the inference and training of open-source generative models based on the [DiT](https://arxiv.org/abs/2212.09748) architecture, such as [SD3](https://arxiv.org/pdf/2403.03206) and [PixArt](https://arxiv.org/abs/2310.00426).
+- [2024.05]: Introducing SCEPTER v1, supporting customized image edit tasks! Simply provide 10 image pairs, SCEPTER will tune an edit tuner for your own Image-to-Image tasks, like `Clay Style`, `De-Text`, `Segmentation`, etc.
 - [2024.04]: New [StyleBooth](https://ali-vilab.github.io/stylebooth-page/) demo on SCEPTER Studio for`Text-Based Style Editing`.
 - [2024.03]: We optimize the training UI and checkpoint management. New [LAR-Gen](https://arxiv.org/abs/2403.19534) model has been added on SCEPTER Studio, supporting `zoom-out`, `virtual try on`, `inpainting`.
 - [2024.02]: We release new SCEdit controllable image synthesis models for SD v2.1 and SD XL. Multiple strategies applied to accelerate inference time for SCEPTER Studio.
@@ -29,59 +34,164 @@ SCEPTER offers 3 core components:
 
 ## 🖼 Gallery for Recent Works
 
-### StyleBooth
-<table>
+### ACE
+
+ACE is a unified foundational model framework that supports a wide range of visual generation tasks. By defining CU for unifying multi-modal inputs across different tasks and incorporating long-context CU, we introduce historical contextual information into visual generation tasks, paving the way for ChatGPT-like dialog systems in visual generation.
+
+[![Watch the demo](https://ali-vilab.github.io/ace-page/static/images/tasks.png)](https://ali-vilab.github.io/ace-page/)
+
+#### ACE Training
+
+We offer a demonstration training YAML that enables the end-to-end training of ACE using a toy dataset. For a comprehensive overview of the hyperparameter configurations, please consult `scepter/methods/edit/dit_ace_0.6b_512.yaml`.
+
+##### Prepare datasets
+
+Please find the dataset class located in `scepter/modules/data/dataset/ms_dataset.py`,
+designed to facilitate end-to-end training using an open-source toy dataset.
+Download a dataset zip file from [modelscope](https://www.modelscope.cn/models/iic/scepter/resolve/master/datasets/hed_pair.zip), and then extract its contents into the `cache/datasets/` directory.
+
+Should you wish to prepare your own datasets, we recommend consulting `scepter/modules/data/dataset/ms_dataset.py` for detailed guidance on the required data format.
+
+##### Prepare initial weight
+The ACE checkpoint has been uploaded to both ModelScope and HuggingFace platforms:
+* [ModelScope](https://www.modelscope.cn/models/iic/ACE-0.6B-512px)
+* [HuggingFace](https://huggingface.co/scepter-studio/ACE-0.6B-512px)
+
+In the provided training YAML configuration, we have designated the Modelscope URL as the default checkpoint URL. Should you wish to transition to Hugging Face, you can effortlessly achieve this by modifying the PRETRAINED_MODEL value within the YAML file (replace the prefix "ms://iic" to "hf://scepter-studio").
+
+
+##### Start training
+
+You can easily start training procedure by executing the following command:
+```bash
+PYTHONPATH=. python scepter/tools/run_train.py --cfg scepter/methods/edit/dit_ace_0.6b_512.yaml
+```
+
+#### ACE Chat Bot
+
+We have developed a chatbot interface utilizing Gradio, designed to convert user input in natural language into visually captivating images that align semantically with the specified instructions. You can easily access this functionality by launching Scepter Studio with the following command:
+```bash
+PYTHONPATH=. python scepter/tools/webui.py --cfg scepter/methods/studio/scepter_ui.yaml --language zh
+```
+Upon starting, you will find a "ChatBot" tab within the Gradio application, which serves as a chat-based interface to handle any requests related to image editing or generation.
+
+#### ACE ComfyUI Workflow
+
+![Workflow](https://github.com/ali-vilab/ace-page/raw/main/assets/comfyui/ace_example.jpg)
+
+<table><tbody>
   <tr>
-    <td><strong>Origin Image</strong><br>Gold Dragon Tuner</td>
-    <td><strong>Graffiti Art</strong></td>
-    <td><strong>Adorable Kawaii</strong></td>
-    <td><strong>game-retro game</strong></td>
-    <td><strong>Vincent van Gogh</strong></td>
+    <th align="center" colspan="4">ACE Workflow Examples</th>
   </tr>
   <tr>
-    <td><img src="asset/images/scedit/tuner_gold_dragon.jpeg" width="240"></td>
-    <td><img src="asset/images/stylebooth/graffiti.jpeg" width="240"></td>
-    <td><img src="asset/images/stylebooth/kawaii.jpeg" width="240"></td>
-    <td><img src="asset/images/stylebooth/retrogame.jpeg" width="240"></td>
-    <td><img src="asset/images/stylebooth/vangogh.jpeg" width="240"></td>
+    <th align="center" colspan="1">Control</th>
+    <th align="center" colspan="1">Semantic</th>
+    <th align="center" colspan="1">Element</th>
   </tr>
+  <tr>
+    <td>
+      <a href="https://github.com/ali-vilab/ace-page/raw/main/assets/comfyui/ace_control.png" target="_blank">
+        <img src="https://github.com/ali-vilab/ace-page/raw/main/assets/comfyui/ace_control.png" width="200">
+      </a>
+    </td>
+    <td>
+      <a href="https://github.com/ali-vilab/ace-page/raw/main/assets/comfyui/ace_semantic.png" target="_blank">
+        <img src="https://github.com/ali-vilab/ace-page/raw/main/assets/comfyui/ace_semantic.png" width="200">
+      </a>
+    </td>
+    <td>
+      <a href="https://github.com/ali-vilab/ace-page/raw/main/assets/comfyui/ace_element.png" target="_blank">
+        <img src="https://github.com/ali-vilab/ace-page/raw/main/assets/comfyui/ace_element.png" width="200">
+      </a>
+    </td>
+  </tr>
+</tbody>
 </table>
 
-<table>
+### FLUX Tuners
+
+<table><tbody>
   <tr>
-    <td><strong>Origin Image</strong></td>
-    <td><strong>Lowpoly</strong></td>
-    <td><strong>Colored Pencil Art</strong></td>
-    <td><strong>Watercolor</strong></td>
-    <td><strong>misc-disco</strong></td>
+    <th align="center" colspan="3">Yarn Style</th>
+    <th align="center" colspan="3">Soft Watercolor Style</th>
   </tr>
   <tr>
-    <td><img src="asset/images/stylebooth/mountain.jpg" width="240"></td>
-    <td><img src="asset/images/stylebooth/lowpoly.jpg" width="240"></td>
-    <td><img src="asset/images/stylebooth/colorpencil.jpeg" width="240"></td>
-    <td><img src="asset/images/stylebooth/watercolor.jpeg" width="240"></td>
-    <td><img src="asset/images/stylebooth/disco.jpeg" width="240"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_2_1.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_2_2.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_2_3.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_1_1.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_1_2.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_1_3.webp" width="200"></td>
   </tr>
+  <tr>
+    <th align="center" colspan="3">Travel Style</th>
+    <th align="center" colspan="3">WuKong Style</th>
+  </tr>
+  <tr>
+    <td><img src="asset/images/flux_tuner/flux_tuner_3_1.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_3_2.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_3_3.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_4_1.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_4_2.webp" width="200"></td>
+    <td><img src="asset/images/flux_tuner/flux_tuner_4_3.webp" width="200"></td>
+  </tr>
+</tbody>
+</table>
+
+### ComfyUI Workflow
+
+![Workflow](asset/workflow/workflow.jpg)
+
+<table><tbody>
+  <tr>
+    <th align="center" colspan="4">Example Workflow Case</th>
+  </tr>
+  <tr>
+    <th align="center" colspan="1">Base</th>
+    <th align="center" colspan="1">+Mantra</th>
+    <th align="center" colspan="1">+Tuner</th>
+    <th align="center" colspan="1">+Control</th>
+  </tr>
+  <tr>
+    <td>
+      <a href="asset/workflow/sdxl_base.json" target="_blank">
+        <img src="asset/workflow/sdxl_base.jpg" width="200">
+      </a>
+    </td>
+    <td>
+      <a href="asset/workflow/sdxl_base_mantra.json" target="_blank">
+        <img src="asset/workflow/sdxl_base_mantra.jpg" width="200">
+      </a>
+    </td>
+    <td>
+      <a href="asset/workflow/sdxl_base_mantra_tuner.json" target="_blank">
+        <img src="asset/workflow/sdxl_base_mantra_tuner.jpg" width="200">
+      </a>
+    </td>
+    <td>
+      <a href="asset/workflow/sdxl_base_mantra_tuner_control.json" target="_blank">
+        <img src="asset/workflow/sdxl_base_mantra_tuner_control.jpg" width="200">
+      </a>
+    </td>
+  </tr>
+</tbody>
 </table>
 
 
 ## 🛠️ Installation
 
-- Create new environment
+- Create new environment with `conda` command:
 
 ```shell
 conda env create -f environment.yaml
 conda activate scepter
 ```
-- We recommend installing the specific version of PyTorch and accelerate toolbox [xFormers](https://pypi.org/project/xformers/). You can install these recommended version by pip:
 
+- Install with `pip` command:
+
+We recommend installing the specific version of PyTorch and accelerate toolbox [xFormers](https://pypi.org/project/xformers/). You can install these recommended version by pip:
 ```shell
 pip install -r requirements/recommended.txt
-```
-
-- Install SCEPTER by the `pip` command:
-
-```shell
 pip install scepter
 ```
 
@@ -100,16 +210,18 @@ pip install scepter
 
 ### Currently supported approaches
 
-|            Tasks             |                   Methods                    | Links                                                                                                                                                                                                                                                       |
-|:----------------------------:|:--------------------------------------------:|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-|   Text-to-image generation   |                   SD v1.5                    | [![Hugging Face Repo](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Repo-blue)](https://huggingface.co/runwayml/stable-diffusion-v1-5)                                                                                                         |
-|   Text-to-image generation   |                   SD v2.1                    | [![Hugging Face Repo](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Repo-blue)](https://huggingface.co/runwayml/stable-diffusion-v1-5)                                                                                                         |
-|   Text-to-image generation   |                    SD-XL                     | [![Hugging Face Repo](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Repo-blue)](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)                                                                                               |
-|       Efficient Tuning       |                     LoRA                     | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=LoRA&color=red&logo=arxiv)](https://arxiv.org/abs/2106.09685)                                                                                                                         |
-|       Efficient Tuning       |            Res-Tuning(NeurIPS23)             | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=Res-Tuing&color=red&logo=arxiv)](https://arxiv.org/abs/2310.19859) [![Page link](https://img.shields.io/badge/Page-ResTuning-Gree)](https://res-tuning.github.io/)                    |
-| Controllable image synthesis | [🌟SCEdit(CVPR24)](docs/en/tasks/scedit.md)  | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=SCEdit&color=red&logo=arxiv)](https://arxiv.org/abs/2312.11392)   [![Page link](https://img.shields.io/badge/Page-SCEdit-Gree)](https://scedit.github.io/)                            |
-|        Image editing         |     [🌟LAR-Gen](docs/en/tasks/largen.md)     | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=LARGen&color=red&logo=arxiv)](https://arxiv.org/abs/2403.19534)   [![Page link](https://img.shields.io/badge/Page-LARGen-Gree)](https://ali-vilab.github.io/largen-page/)             |
-|        Image editing         |  [🌟StyleBooth](docs/en/tasks/stylebooth.md) | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=StyleBooth&color=red&logo=arxiv)](https://arxiv.org/abs/2404.12154)   [![Page link](https://img.shields.io/badge/Page-StyleBooth-Gree)](https://ali-vilab.github.io/stylebooth-page/) |
+|            Tasks             |                    Methods                     | Links                                                                                                                                                                                                                                                       |
+|:----------------------------:|:----------------------------------------------:|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+|   Text-to-image Generation   |                    SD v1.5                     | [![Hugging Face Repo](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Repo-blue)](https://huggingface.co/runwayml/stable-diffusion-v1-5)                                                                                                         |
+|   Text-to-image Generation   |                    SD v2.1                     | [![Hugging Face Repo](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Repo-blue)](https://huggingface.co/runwayml/stable-diffusion-v1-5)                                                                                                         |
+|   Text-to-image Generation   |                     SD-XL                      | [![Hugging Face Repo](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Repo-blue)](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)                                                                                               |
+|   Text-to-image Generation   |                      FLUX                      | [![Hugging Face Repo](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Repo-blue)](https://huggingface.co/black-forest-labs/FLUX.1-dev)                                                                                               |
+|       Efficient Tuning       |                      LoRA                      | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=LoRA&color=red&logo=arxiv)](https://arxiv.org/abs/2106.09685)                                                                                                                         |
+|       Efficient Tuning       |             Res-Tuning(NeurIPS23)              | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=Res-Tuing&color=red&logo=arxiv)](https://arxiv.org/abs/2310.19859) [![Page link](https://img.shields.io/badge/Page-ResTuning-Gree)](https://res-tuning.github.io/)                    |
+| Controllable Image Synthesis |  [🌟SCEdit(CVPR24)](docs/en/tasks/scedit.md)   | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=SCEdit&color=red&logo=arxiv)](https://arxiv.org/abs/2312.11392)   [![Page link](https://img.shields.io/badge/Page-SCEdit-Gree)](https://scedit.github.io/)                            |
+|        Image Editing         |      [🌟LAR-Gen](docs/en/tasks/largen.md)      | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=LARGen&color=red&logo=arxiv)](https://arxiv.org/abs/2403.19534)   [![Page link](https://img.shields.io/badge/Page-LARGen-Gree)](https://ali-vilab.github.io/largen-page/)             |
+|        Image Editing         |  [🌟StyleBooth](docs/en/tasks/stylebooth.md)   | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=StyleBooth&color=red&logo=arxiv)](https://arxiv.org/abs/2404.12154)   [![Page link](https://img.shields.io/badge/Page-StyleBooth-Gree)](https://ali-vilab.github.io/stylebooth-page/) |
+| Image Generation and Editing | [🌟ACE](https://ali-vilab.github.io/ace-page/) | [![Arxiv   link](https://img.shields.io/static/v1?label=arXiv&message=ACE&color=red&logo=arxiv)](https://arxiv.org/abs/2410.00086)   [![Page link](https://img.shields.io/badge/Page-ACE-Gree)](https://ali-vilab.github.io/ace-page/) [![Demo link](https://img.shields.io/badge/Demo-ACE-purple)](https://huggingface.co/spaces/scepter-studio/ACE-Chat) <br> [![ModelScope link](https://img.shields.io/badge/ModelScope-Model-blue)](https://www.modelscope.cn/models/iic/ACE-0.6B-512px)  [![HuggingFace link](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Model-yellow)](https://huggingface.co/scepter-studio/ACE-0.6B-512px) |
 
 
 ## 🖥️ SCEPTER Studio
@@ -132,8 +244,6 @@ The startup of **SCEPTER Studio** eliminates the need for manual downloading and
 Depending on the network and hardware situation, the initial startup usually requires 15-60 minutes, primarily involving the download and processing of SDv1.5, SDv2.1, and SDXL models.
 Therefore, subsequent startups will become much faster (about one minute) as downloading is no longer required.
 
-To support the sharing and downloading of models,
-please make sure that you have installed **zip** and Git Large File Storage (**git lfs**).
 ### Usage Demo
 
 |              [Image Editing](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets%2Fscepter_studio%2Fimage_editing_20240419.webm)              |                [Training](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets%2Fscepter_studio%2Ftraining_20240419.webm)                 |              [Model Sharing](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets%2Fscepter_studio%2Fmodel_sharing_20240419.webm)               |             [Model Inference](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets%2Fscepter_studio%2Fmodel_inference_20240419.webm)              |             [Data Management](https://www.modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets%2Fscepter_studio%2Fdata_management_20240419.webm)              |
@@ -142,7 +252,23 @@ please make sure that you have installed **zip** and Git Large File Storage (**g
 
 ### Modelscope Studio & Huggingface Space
 
-We deploy a work studio on Modelscope that includes only the inference tab, please refer to [ms_scepter_studio](https://www.modelscope.cn/studios/damo/scepter_studio/summary) and [hf_scepter_studio](https://huggingface.co/spaces/modelscope/scepter_studio)
+We deploy a work studio on Modelscope that includes only the inference tab, please refer to [ms_scepter_studio](https://www.modelscope.cn/studios/iic/scepter_studio/summary) and [hf_scepter_studio](https://huggingface.co/spaces/modelscope/scepter_studio)
+
+
+
+## ⚙️️ ComfyUI Workflow
+
+### Launch
+
+Manually install by moving custom_nodes to ComfyUI.
+```shell
+cd path/to/scepter
+pip install -e .
+cp -r path/to/scepter/workflow/ path/to/ComfyUI/custom_nodes/ComfyUI-Scepter
+cd path/to/ComfyUI
+python main.py
+```
+In addition, we also support installation and usage through the ComfyUI Manager.
 
 
 ## 🔍 Learn More
@@ -177,4 +303,4 @@ This project is licensed under the [Apache License (Version 2.0)](https://github
 
 
 ## Acknowledgement
-Thanks to [Stability-AI](https://github.com/Stability-AI), [SWIFT library](https://github.com/modelscope/swift/) and [Fooocus](https://github.com/lllyasviel/Fooocus) for their awesome work.
+Thanks to [Stability-AI](https://github.com/Stability-AI), [SWIFT library](https://github.com/modelscope/swift/), [Fooocus](https://github.com/lllyasviel/Fooocus) and [ComfyUI](https://github.com/comfyanonymous/ComfyUI) for their awesome work.

requirements/framework.txt

@@ -1,8 +1,9 @@
 albumentations
+beautifulsoup4
 bezier
 einops
 modelscope
-ms-swift>=2.0.1
+ms-swift
 numpy
 open_clip_torch
 opencv-python
@@ -13,3 +14,4 @@ pyyaml>=5.3.1
 scikit-image
 torchsde
 transformers
+scikit-learn

requirements/scepter_studio.txt

@@ -1,5 +1,6 @@
 bitsandbytes
-gradio>=3.47.1,<4.0.0
+gradio==4.44.1
+gradio_imageslider
 imagehash
 psutil
 tiktoken

scepter/methods/edit/dit_ace_0.6b_512.yaml

@@ -0,0 +1,161 @@
+ENV:
+  BACKEND: nccl
+  SEED: 2024
+#
+SOLVER:
+  NAME: ACESolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 500
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: 50
+  RESCALE_LR: False
+  #
+  WORK_DIR: ./cache/save_data/ace_0.6b_512
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    - NAME: "HuggingfaceFs"
+      TEMP_DIR: ./cache/cache_data
+    - NAME: "LocalFs"
+      TEMP_DIR: ./cache/cache_data
+    - NAME: "ModelscopeFs"
+      TEMP_DIR: ./cache/cache_data
+
+  #
+  MODEL:
+    NAME: LatentDiffusionACE
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    SCALE_FACTOR: 0.18215
+    SIZE_FACTOR: 8
+    DECODER_BIAS: 0.5
+    DEFAULT_N_PROMPT:
+    USE_EMA: True
+    EVAL_EMA: False
+    TEXT_IDENTIFIER: [ '{image}', '{image1}', '{image2}', '{image3}', '{image4}', '{image5}', '{image6}', '{image7}', '{image8}', '{image9}' ]
+    USE_TEXT_POS_EMBEDDINGS: True
+    #
+    DIFFUSION:
+      NAME: BaseDiffusion
+      PREDICTION_TYPE: eps
+      MIN_SNR_GAMMA:
+      NOISE_SCHEDULER:
+        NAME: LinearScheduler
+        NUM_TIMESTEPS: 1000
+        BETA_MIN: 0.0001
+        BETA_MAX: 0.02
+    #
+    DIFFUSION_MODEL:
+      NAME: ACE
+      PRETRAINED_MODEL: ms://iic/ACE-0.6B-512px@models/dit/ace_0.6b_512px.pth
+      IGNORE_KEYS: [ ]
+      PATCH_SIZE: 2
+      IN_CHANNELS: 4
+      HIDDEN_SIZE: 1152
+      DEPTH: 28
+      NUM_HEADS: 16
+      MLP_RATIO: 4.0
+      PRED_SIGMA: True
+      DROP_PATH: 0.0
+      WINDOW_DIZE: 0
+      Y_CHANNELS: 4096
+      MAX_SEQ_LEN: 1024
+      QK_NORM: True
+      USE_GRAD_CHECKPOINT: True
+      ATTENTION_BACKEND: flash_attn
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      EMBED_DIM: 4
+      PRETRAINED_MODEL: ms://iic/ACE-0.6B-512px@models/vae/vae.bin
+      IGNORE_KEYS: []
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      NAME: T5EmbedderHF
+      PRETRAINED_MODEL: ms://iic/ACE-0.6B-512px@models/text_encoder/t5-v1_1-xxl/
+      TOKENIZER_PATH: ms://iic/ACE-0.6B-512px@models/tokenizer/t5-v1_1-xxl
+      LENGTH: 120
+      T5_DTYPE: bfloat16
+      ADDED_IDENTIFIER: [ '{image}', '{caption}', '{mask}', '{ref_image}', '{image1}', '{image2}', '{image3}', '{image4}', '{image5}', '{image6}', '{image7}', '{image8}', '{image9}' ]
+      CLEAN: whitespace
+      USE_GRAD: False
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: ddim
+    SAMPLE_STEPS: 20
+    GUIDE_SCALE: 4.5
+    GUIDE_RESCALE: 0.5
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 1e-7
+    EPS: 1e-10
+    WEIGHT_DECAY: 5e-4
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDatasetForACE
+    MODE: train
+    MS_DATASET_NAME: cache/datasets/hed_pair
+    MS_DATASET_NAMESPACE: ""
+    MS_DATASET_SPLIT: "train"
+    MS_DATASET_SUBNAME: ""
+    PROMPT_PREFIX: ""
+    REPLACE_STYLE: False
+    MAX_SEQ_LEN: 1024
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 1
+    SAMPLER:
+      NAME: LoopSampler
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 0
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 50
+    -
+      NAME: CheckpointHook
+      INTERVAL: 100
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 100

scepter/methods/edit/edit_512_lora.yaml

@@ -0,0 +1,250 @@
+ENV:
+  BACKEND: nccl
+SOLVER:
+  NAME: LatentDiffusionSolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 2000
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: 100
+  #
+  WORK_DIR: ./cache/save_data/edit_512_lora
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  TUNER:
+    -
+      NAME: SwiftLoRA
+      R: 64
+      LORA_ALPHA: 64
+      LORA_DROPOUT: 0.0
+      BIAS: "none"
+      TARGET_MODULES: model.*(to_q|to_k|to_v|to_out.0|net.0.proj|net.2)$
+  #
+  MODEL:
+    NAME: LatentDiffusionEdit
+    PARAMETERIZATION: eps
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL: ms://iic/stylebooth@models/stylebooth-tb-5000-0.bin
+    IGNORE_KEYS: [ ]
+    CONCAT_NO_SCALE_FACTOR: True
+    SCALE_FACTOR: 0.18215
+    SIZE_FACTOR: 8
+    # DEFAULT_N_PROMPT: 'lowres, error, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, username, watermark, signature'
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "scaled_linear"
+      "BETA_MIN": 0.00085
+      "BETA_MAX": 0.012
+    USE_EMA: False
+    #
+    DIFFUSION_MODEL:
+      NAME: DiffusionUNet
+      IN_CHANNELS: 8
+      OUT_CHANNELS: 4
+      MODEL_CHANNELS: 320
+      NUM_HEADS: 8
+      NUM_RES_BLOCKS: 2
+      ATTENTION_RESOLUTIONS: [ 4, 2, 1 ]
+      CHANNEL_MULT: [ 1, 2, 4, 4 ]
+      CONV_RESAMPLE: True
+      DIMS: 2
+      USE_CHECKPOINT: False
+      USE_SCALE_SHIFT_NORM: False
+      RESBLOCK_UPDOWN: False
+      USE_SPATIAL_TRANSFORMER: True
+      TRANSFORMER_DEPTH: 1
+      CONTEXT_DIM: 768
+      DISABLE_MIDDLE_SELF_ATTN: False
+      USE_LINEAR_IN_TRANSFORMER: False
+      PRETRAINED_MODEL:
+      IGNORE_KEYS: []
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      EMBED_DIM: 4
+      PRETRAINED_MODEL:
+      IGNORE_KEYS: []
+      BATCH_SIZE: 4
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    TOKENIZER:
+      NAME: ClipTokenizer
+      PRETRAINED_PATH: ms://AI-ModelScope/clip-vit-large-patch14
+      LENGTH: 77
+      CLEAN: True
+    #
+    COND_STAGE_MODEL:
+      NAME: FrozenCLIPEmbedder
+      FREEZE: True
+      LAYER: last
+      PRETRAINED_MODEL: ms://AI-ModelScope/clip-vit-large-patch14
+    #
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: ddim
+    SAMPLE_STEPS: 50
+    SEED: 2023
+    GUIDE_SCALE: #7.5
+      image: 1.5
+      text: 7.5
+    GUIDE_RESCALE: 0.5
+    DISCRETIZATION: trailing
+    IMAGE_SIZE: [512, 512]
+    RUN_TRAIN_N: False
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 0.064
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: cache/datasets/hed_pair
+    MS_DATASET_NAMESPACE: ""
+    MS_DATASET_SPLIT: "train"
+    MS_DATASET_SUBNAME: ""
+    PROMPT_PREFIX: ""
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFileList
+        FILE_KEYS: ['img_path', 'src_path']
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'img', 'src' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'img', 'src' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'img', 'src' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'image', 'condition_cat' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'condition_cat', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 0
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 50
+    -
+      NAME: CheckpointHook
+      INTERVAL: 1000
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "Convert to an edge map#;#cache/datasets/hed_pair/images/src_001.jpeg" ]
+    IMAGE_SIZE: [ 512, 512 ]
+    FIELDS: [ "prompt", "src_path" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: LoadImageFromFileList
+        FILE_KEYS: [ 'src_path' ]
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'src' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'src' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'src' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'condition_cat' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'condition_cat', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  EVAL_HOOKS:
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/examples/classification/example.yaml

@@ -11,7 +11,7 @@ SOLVER:
   # NUM_FOLDS DESCRIPTION: Num folds for training. TYPE: int default: 0
   NUM_FOLDS: 1
   # WORK_DIR DESCRIPTION: Save dir of the training log or model. TYPE: str default: ''
-  WORK_DIR: ./exp12/
+  WORK_DIR: ./cache/save_data/example/
   LOG_FILE: std_log.txt
   # EVAL_INTERVAL DESCRIPTION: Eval the model interval. TYPE: int default: 1
   EVAL_INTERVAL: 1
@@ -102,7 +102,7 @@ SOLVER:
     # DATASET DESCRIPTION: the public dataset name TYPE: str default: 'cifar10'
     DATASET: cifar10
     # DATA_ROOT DESCRIPTION: the download data save path TYPE: str default: ''
-    DATA_ROOT: ./local_data/cifar10
+    DATA_ROOT: ./cache/cache_data/cifar10
     # MODE DESCRIPTION: test TYPE: str default: test
     MODE: test
     # PIN_MEMORY DESCRIPTION: pin_memory for data loader TYPE: bool default: False

scepter/methods/examples/generation/dit_flux1.0_dev_1024_lora.yaml

@@ -0,0 +1,303 @@
+ENV:
+  BACKEND: nccl
+  SEED: 166666
+SOLVER:
+  # NAME DESCRIPTION:  TYPE:  default: 'LatentUfitSolver'
+  NAME: LatentDiffusionSolver
+  # MAX_STEPS DESCRIPTION: The total steps for training. TYPE: int default: 100000
+  MAX_STEPS: 100000
+  # USE_AMP DESCRIPTION: Use amp to surpport mix precision or not, default is False. TYPE: bool default: False
+  USE_AMP: True
+  # DTYPE DESCRIPTION: The precision for training. TYPE: str default: 'float32'
+  DTYPE: bfloat16
+  # USE_FAIRSCALE DESCRIPTION: Use fairscale as the backend of ddp, default False. TYPE: bool default: False
+  USE_FAIRSCALE: False
+  # USE_FSDP DESCRIPTION: Use fsdp as the backend of ddp, default False. TYPE: bool default: False
+  USE_FSDP: True
+  # LOAD_MODEL_ONLY DESCRIPTION: Only load the model rather than the optimizer and schedule, default is False. TYPE: bool default: False
+  LOAD_MODEL_ONLY: False
+  # RESUME_FROM DESCRIPTION: Resume from some state of training! TYPE: str default: ''
+  RESUME_FROM:
+  WORK_DIR: ./cache/save_data/dit_flux_dev_1024_lora
+  LOG_FILE: std_log.txt
+  # EVAL_INTERVAL DESCRIPTION: Eval the model interval. TYPE: int default: 1
+  EVAL_INTERVAL: 100
+  # LOG_TRAIN_NUM DESCRIPTION: The number samples used to log in training phase. TYPE: int default: -1
+  LOG_TRAIN_NUM: 16
+  # FSDP_REDUCE_DTYPE DESCRIPTION: The dtype of reduce in FSDP. TYPE: str default: 'float16'
+  FSDP_REDUCE_DTYPE: float32
+  # FSDP_BUFFER_DTYPE DESCRIPTION: The dtype of buffer in FSDP. TYPE: str default: 'float16'
+  FSDP_BUFFER_DTYPE: float32
+  # FSDP_SHARD_MODULES DESCRIPTION: The modules to be sharded in FSDP. TYPE: list default: ['model']
+  FSDP_SHARD_MODULES: [ 'model', 'cond_stage_model.t5_model' ] #
+  SAVE_MODULES: [ 'model'] #
+  TRAIN_MODULES: ['model']
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  TUNER:
+    - NAME: SwiftLoRA
+      R: 4
+      LORA_ALPHA: 4
+      LORA_DROPOUT: 0.0
+      BIAS: "none"
+      TARGET_MODULES: "(model.double_blocks.*(.qkv|.proj|.img_mod.lin|.txt_mod.lin))|(model.single_blocks.*(.linear1|.linear2|.modulation.lin))$"
+  ##
+  MODEL:
+    NAME: LatentDiffusionFlux
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    DEFAULT_N_PROMPT:
+    USE_EMA: False
+    EVAL_EMA: False
+    DIFFUSION:
+      # NAME DESCRIPTION:  TYPE:  default: 'DiffusionFluxRF'
+      NAME: DiffusionFluxRF
+      PREDICTION_TYPE: raw
+      # NOISE_SCHEDULER DESCRIPTION:  TYPE:  default: ''
+      NOISE_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchSigmaScheduler'
+        NAME: FlowMatchSigmaScheduler
+        # WEIGHTING_SCHEME DESCRIPTION: The weighting scheme for sampling timesteps, choose from ['sigma_sqrt', 'logit_normal', 'mode', 'cosmap', 'none']. TYPE: str default: 'logit_normal'
+        WEIGHTING_SCHEME: logit_normal
+        SHIFT: 3.0
+        # LOGIT_MEAN DESCRIPTION: The mean of the logit distribution for sampling timesteps. TYPE: float default: 0.0
+        LOGIT_MEAN: 0.0
+        # LOGIT_STD DESCRIPTION: The standard deviation of the logit distribution for sampling timesteps. TYPE: float default: 1.0
+        LOGIT_STD: 1.0
+        # MODE_SCALE DESCRIPTION: The scale factor for the mode of the logit distribution for sampling timesteps. TYPE: float default: 1.29
+        MODE_SCALE: 1.29
+      SAMPLER_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchFluxShiftScheduler'
+        NAME: FlowMatchFluxShiftScheduler
+        # SHIFT DESCRIPTION: Use timestamp shift or not, default is True. TYPE: bool default: True
+        SHIFT: False
+        # SIGMOID_SCALE DESCRIPTION: The scale of sigmoid function for sampling timesteps. TYPE: int default: 1
+        SIGMOID_SCALE: 1
+        # BASE_SHIFT DESCRIPTION: The base shift factor for the timestamp. TYPE: float default: 0.5
+        BASE_SHIFT: 0.5
+        # MAX_SHIFT DESCRIPTION: The max shift factor for the timestamp. TYPE: float default: 1.15
+        MAX_SHIFT: 1.15
+        #
+    DIFFUSION_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'Flux'
+      NAME: Flux
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@flux1-dev.safetensors
+      # IN_CHANNELS DESCRIPTION: model's input channels. TYPE: int default: 64
+      IN_CHANNELS: 64
+      # HIDDEN_SIZE DESCRIPTION: model's hidden size. TYPE: int default: 1024
+      HIDDEN_SIZE: 3072
+      # NUM_HEADS DESCRIPTION: number of heads in the transformer. TYPE: int default: 16
+      NUM_HEADS: 24
+      # AXES_DIM DESCRIPTION: dimensions of the axes of the positional encoding. TYPE: list default: [16, 56, 56]
+      AXES_DIM: [ 16, 56, 56 ]
+      # THETA DESCRIPTION: theta for positional encoding. TYPE: int default: 10000
+      THETA: 10000
+      # VEC_IN_DIM DESCRIPTION: dimension of the vector input. TYPE: int default: 768
+      VEC_IN_DIM: 768
+      # GUIDANCE_EMBED DESCRIPTION: whether to use guidance embedding. TYPE: bool default: False
+      GUIDANCE_EMBED: True
+      # CONTEXT_IN_DIM DESCRIPTION: dimension of the context input. TYPE: int default: 4096
+      CONTEXT_IN_DIM: 4096
+      # MLP_RATIO DESCRIPTION: ratio of mlp hidden size to hidden size. TYPE: float default: 4.0
+      MLP_RATIO: 4.0
+      # QKV_BIAS DESCRIPTION: whether to use bias in qkv projection. TYPE: bool default: True
+      QKV_BIAS: True
+      # DEPTH DESCRIPTION: number of transformer blocks. TYPE: int default: 19
+      DEPTH: 19
+      # DEPTH_SINGLE_BLOCKS DESCRIPTION: number of transformer blocks in the single stream block. TYPE: int default: 38
+      DEPTH_SINGLE_BLOCKS: 38
+      USE_GRAD_CHECKPOINT: True
+
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKLFlux
+      EMBED_DIM: 16
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@ae.safetensors
+      IGNORE_KEYS: [ ]
+      BATCH_SIZE: 8
+      USE_CONV: False
+      SCALE_FACTOR: 0.3611
+      SHIFT_FACTOR: 0.1159
+      #
+      ENCODER:
+        NAME: Encoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'T5PlusClipFluxEmbedder'
+      NAME: T5PlusClipFluxEmbedder
+      # T5_MODEL DESCRIPTION:  TYPE:  default: ''
+      T5_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: T5EncoderModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder_2/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: T5Tokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer_2/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 512
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: last_hidden_state
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: False
+        CLEAN: whitespace
+      # CLIP_MODEL DESCRIPTION:  TYPE:  default: ''
+      CLIP_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: CLIPTextModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: CLIPTokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 77
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: pooler_output
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: True
+        CLEAN: whitespace
+        USE_GRAD_CHECKPOINT: True
+  #
+  SAMPLE_ARGS:
+    SAMPLE_STEPS: 50
+    SAMPLER: flow_eluer
+    SEED: 2024
+    IMAGE_SIZE: [ 1024, 1024 ]
+    GUIDE_SCALE: 3.5
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 4e-4
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a boy wearing a jacket", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 2
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      PRIORITY: 0
+    - NAME: BackwardHook
+#      GRADIENT_CLIP: 1.0
+      PRIORITY: 10
+    - NAME: LogHook
+      LOG_INTERVAL: 10
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  EVAL_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      PRIORITY: 0

scepter/methods/examples/generation/dit_flux1.0_schnell_1024_lora.yaml

@@ -0,0 +1,302 @@
+ENV:
+  BACKEND: nccl
+  SEED: 166666
+SOLVER:
+  # NAME DESCRIPTION:  TYPE:  default: 'LatentUfitSolver'
+  NAME: LatentDiffusionSolver
+  # MAX_STEPS DESCRIPTION: The total steps for training. TYPE: int default: 100000
+  MAX_STEPS: 100000
+  # USE_AMP DESCRIPTION: Use amp to surpport mix precision or not, default is False. TYPE: bool default: False
+  USE_AMP: True
+  # DTYPE DESCRIPTION: The precision for training. TYPE: str default: 'float32'
+  DTYPE: bfloat16
+  # USE_FAIRSCALE DESCRIPTION: Use fairscale as the backend of ddp, default False. TYPE: bool default: False
+  USE_FAIRSCALE: False
+  # USE_FSDP DESCRIPTION: Use fsdp as the backend of ddp, default False. TYPE: bool default: False
+  USE_FSDP: True
+  # LOAD_MODEL_ONLY DESCRIPTION: Only load the model rather than the optimizer and schedule, default is False. TYPE: bool default: False
+  LOAD_MODEL_ONLY: False
+  # RESUME_FROM DESCRIPTION: Resume from some state of training! TYPE: str default: ''
+  RESUME_FROM:
+  WORK_DIR: ./cache/save_data/dit_flux_schnell_1024_lora
+  LOG_FILE: std_log.txt
+  # EVAL_INTERVAL DESCRIPTION: Eval the model interval. TYPE: int default: 1
+  EVAL_INTERVAL: 100
+  # LOG_TRAIN_NUM DESCRIPTION: The number samples used to log in training phase. TYPE: int default: -1
+  LOG_TRAIN_NUM: 16
+  # FSDP_REDUCE_DTYPE DESCRIPTION: The dtype of reduce in FSDP. TYPE: str default: 'float16'
+  FSDP_REDUCE_DTYPE: float32
+  # FSDP_BUFFER_DTYPE DESCRIPTION: The dtype of buffer in FSDP. TYPE: str default: 'float16'
+  FSDP_BUFFER_DTYPE: float32
+  # FSDP_SHARD_MODULES DESCRIPTION: The modules to be sharded in FSDP. TYPE: list default: ['model']
+  FSDP_SHARD_MODULES: [ 'model', 'cond_stage_model.t5_model' ] #
+  SAVE_MODULES: [ 'model'] #
+  TRAIN_MODULES: ['model']
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  TUNER:
+    - NAME: SwiftLoRA
+      R: 4
+      LORA_ALPHA: 4
+      LORA_DROPOUT: 0.0
+      BIAS: "none"
+      TARGET_MODULES: "(model.double_blocks.*(.qkv|.proj|.img_mod.lin|.txt_mod.lin))|(model.single_blocks.*(.linear1|.linear2|.modulation.lin))$"
+  ##
+  MODEL:
+    NAME: LatentDiffusionFlux
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    DEFAULT_N_PROMPT:
+    USE_EMA: False
+    EVAL_EMA: False
+    DIFFUSION:
+      # NAME DESCRIPTION:  TYPE:  default: 'DiffusionFluxRF'
+      NAME: DiffusionFluxRF
+      PREDICTION_TYPE: raw
+      # NOISE_SCHEDULER DESCRIPTION:  TYPE:  default: ''
+      NOISE_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchSigmaScheduler'
+        NAME: FlowMatchSigmaScheduler
+        # WEIGHTING_SCHEME DESCRIPTION: The weighting scheme for sampling timesteps, choose from ['sigma_sqrt', 'logit_normal', 'mode', 'cosmap', 'none']. TYPE: str default: 'logit_normal'
+        WEIGHTING_SCHEME: logit_normal
+        SHIFT: 3.0
+        # LOGIT_MEAN DESCRIPTION: The mean of the logit distribution for sampling timesteps. TYPE: float default: 0.0
+        LOGIT_MEAN: 0.0
+        # LOGIT_STD DESCRIPTION: The standard deviation of the logit distribution for sampling timesteps. TYPE: float default: 1.0
+        LOGIT_STD: 1.0
+        # MODE_SCALE DESCRIPTION: The scale factor for the mode of the logit distribution for sampling timesteps. TYPE: float default: 1.29
+        MODE_SCALE: 1.29
+      SAMPLER_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchFluxShiftScheduler'
+        NAME: FlowMatchFluxShiftScheduler
+        # SHIFT DESCRIPTION: Use timestamp shift or not, default is True. TYPE: bool default: True
+        SHIFT: False
+        # SIGMOID_SCALE DESCRIPTION: The scale of sigmoid function for sampling timesteps. TYPE: int default: 1
+        SIGMOID_SCALE: 1
+        # BASE_SHIFT DESCRIPTION: The base shift factor for the timestamp. TYPE: float default: 0.5
+        BASE_SHIFT: 0.5
+        # MAX_SHIFT DESCRIPTION: The max shift factor for the timestamp. TYPE: float default: 1.15
+        MAX_SHIFT: 1.15
+        #
+    DIFFUSION_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'Flux'
+      NAME: Flux
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-schnell@flux1-schnell.safetensors
+      # IN_CHANNELS DESCRIPTION: model's input channels. TYPE: int default: 64
+      IN_CHANNELS: 64
+      # HIDDEN_SIZE DESCRIPTION: model's hidden size. TYPE: int default: 1024
+      HIDDEN_SIZE: 3072
+      # NUM_HEADS DESCRIPTION: number of heads in the transformer. TYPE: int default: 16
+      NUM_HEADS: 24
+      # AXES_DIM DESCRIPTION: dimensions of the axes of the positional encoding. TYPE: list default: [16, 56, 56]
+      AXES_DIM: [ 16, 56, 56 ]
+      # THETA DESCRIPTION: theta for positional encoding. TYPE: int default: 10000
+      THETA: 10000
+      # VEC_IN_DIM DESCRIPTION: dimension of the vector input. TYPE: int default: 768
+      VEC_IN_DIM: 768
+      # GUIDANCE_EMBED DESCRIPTION: whether to use guidance embedding. TYPE: bool default: False
+      GUIDANCE_EMBED: False
+      # CONTEXT_IN_DIM DESCRIPTION: dimension of the context input. TYPE: int default: 4096
+      CONTEXT_IN_DIM: 4096
+      # MLP_RATIO DESCRIPTION: ratio of mlp hidden size to hidden size. TYPE: float default: 4.0
+      MLP_RATIO: 4.0
+      # QKV_BIAS DESCRIPTION: whether to use bias in qkv projection. TYPE: bool default: True
+      QKV_BIAS: True
+      # DEPTH DESCRIPTION: number of transformer blocks. TYPE: int default: 19
+      DEPTH: 19
+      # DEPTH_SINGLE_BLOCKS DESCRIPTION: number of transformer blocks in the single stream block. TYPE: int default: 38
+      DEPTH_SINGLE_BLOCKS: 38
+      USE_GRAD_CHECKPOINT: True
+
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKLFlux
+      EMBED_DIM: 16
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-schnell@ae.safetensors
+      IGNORE_KEYS: [ ]
+      BATCH_SIZE: 8
+      USE_CONV: False
+      SCALE_FACTOR: 0.3611
+      SHIFT_FACTOR: 0.1159
+      #
+      ENCODER:
+        NAME: Encoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'T5PlusClipFluxEmbedder'
+      NAME: T5PlusClipFluxEmbedder
+      # T5_MODEL DESCRIPTION:  TYPE:  default: ''
+      T5_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: T5EncoderModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-schnell@text_encoder_2/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: T5Tokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-schnell@tokenizer_2/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 256
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: last_hidden_state
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: False
+        CLEAN: whitespace
+      # CLIP_MODEL DESCRIPTION:  TYPE:  default: ''
+      CLIP_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: CLIPTextModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-schnell@text_encoder/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: CLIPTokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-schnell@tokenizer/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 77
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: pooler_output
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: True
+        CLEAN: whitespace
+  #
+  SAMPLE_ARGS:
+    SAMPLE_STEPS: 4
+    SAMPLER: flow_eluer
+    SEED: 2024
+    IMAGE_SIZE: [ 1024, 1024 ]
+    GUIDE_SCALE: 3.5
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 4e-4
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a boy wearing a jacket", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 2
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      PRIORITY: 0
+    - NAME: BackwardHook
+#      GRADIENT_CLIP: 1.0
+      PRIORITY: 10
+    - NAME: LogHook
+      LOG_INTERVAL: 10
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  EVAL_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      PRIORITY: 0

scepter/methods/examples/generation/dit_pixart_alpha_1024_lora.yaml

@@ -0,0 +1,223 @@
+ENV:
+  BACKEND: nccl
+#
+SOLVER:
+  NAME: LatentDiffusionSolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 1000
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: 100
+  RESCALE_LR: False
+  #
+  WORK_DIR: ./cache/save_data/dit_pixart_alpha_1024_lora
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  #
+  TUNER:
+    - NAME: SwiftLoRA
+      R: 128
+      LORA_ALPHA: 128
+      LORA_DROPOUT: 0.0
+      BIAS: "none"
+      TARGET_MODULES: "model.*(.q|.k|.v|.o|mlp.fc1|mlp.fc2)$"
+  #
+  MODEL:
+    NAME: LatentDiffusionPixart
+    PARAMETERIZATION: eps
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    SCALE_FACTOR: 0.18215
+    SIZE_FACTOR: 8
+    DECODER_BIAS: 0.5
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "linear"
+      "BETA_MIN": 0.0001
+      "BETA_MAX": 0.02
+    USE_EMA: False
+    LOAD_REFINER: False
+    #
+    DIFFUSION_MODEL:
+      NAME: PixArt
+      PRETRAINED_MODEL: ms://AI-ModelScope/PixArt-alpha@PixArt-XL-2-1024-MS.pth
+      INPUT_SIZE: 128
+      PATCH_SIZE: 2
+      IN_CHANNELS: 4
+      HIDDEN_SIZE: 1152
+      DEPTH: 28
+      NUM_HEADS: 16
+      MLP_RATIO: 4.0
+      CLASS_DROPOUT_PROB: 0.1
+      PRED_SIGMA: True
+      DROP_PATH: 0.0
+      WINDOW_DIZE: 0
+      USE_REL_POS: False
+      CAPTION_CHANNELS: 4096
+      LEWEI_SCALE: 2
+      MODEL_MAX_LENGTH: 120
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-2-base@512-base-ema.safetensors
+      EMBED_DIM: 4
+      IGNORE_KEYS: [ ]
+      BATCH_SIZE: 1
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      NAME: T5EmbedderHF
+      PRETRAINED_MODEL: ms://AI-ModelScope/PixArt-alpha@t5-v1_1-xxl/
+      TOKENIZER_PATH: ms://AI-ModelScope/PixArt-alpha@t5-v1_1-xxl/
+      LENGTH: 120
+      CLEAN: heavy
+      USE_GRAD: False
+    #
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: ddim
+    SAMPLE_STEPS: 20
+    SEED: 2024
+    GUIDE_SCALE: 4.5
+    GUIDE_RESCALE: 0.5
+    DISCRETIZATION: trailing
+    RUN_TRAIN_N: False
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 0.0001
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a boy wearing a jacket", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 0
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 10
+      SHOW_GPU_MEM: True
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  #
+  EVAL_HOOKS:
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/examples/generation/dit_sd3_1024.yaml

@@ -0,0 +1,233 @@
+ENV:
+  BACKEND: nccl
+#
+SOLVER:
+  NAME: LatentDiffusionSolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 500
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: 50
+  RESCALE_LR: False
+  #
+  WORK_DIR: ./cache/save_data/dit_sd3_1024
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    - NAME: "ModelscopeFs"
+      TEMP_DIR: ./cache/cache_data
+  #
+  MODEL:
+    NAME: LatentDiffusionSD3
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    SCALE_FACTOR: 1.5305
+    SHIFT_FACTOR: 0.0609
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "shifted"
+      "SHIFT": 3
+    USE_EMA: False
+    T_WEIGHT: uniform
+    #
+    DIFFUSION_MODEL:
+      NAME: MMDiT
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+      IGNORE_KEYS: '^first_stage_model.'
+      IN_CHANNELS: 16
+      PATCH_SIZE: 2
+      OUT_CHANNELS: 16
+      DEPTH: 24
+      INPUT_SIZE:
+      ADM_IN_CHANNELS: 2048
+      CONTEXT_EMBEDDER_CONFIG: { 'target': 'torch.nn.Linear', 'params': { 'in_features': 4096, 'out_features': 1536 } }
+      NUM_PATCHES: 36864
+      POS_EMBED_MAX_SIZE: 192
+      POS_EMBED_SCALING_FACTOR:
+      USE_CHECKPOINT: True
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+      EMBED_DIM: 16
+      IGNORE_KEYS: '^model.diffusion_model.'
+      BATCH_SIZE: 1
+      USE_CONV: False
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      NAME: SD3TextEmbedder
+      P_ZERO: 0.0
+      CLIP_L:
+        NAME: FrozenCLIPEmbedder2
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer
+        MAX_LENGTH: 77
+        FREEZE: True
+        LAYER: penultimate
+        RETURN_POOLED: True
+        USE_FINAL_LAYER_NORM: False
+        IS_TRAINABLE: False
+      CLIP_G:
+        NAME: FrozenCLIPEmbedder2
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_2
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_2
+        MAX_LENGTH: 77
+        FREEZE: True
+        LAYER: penultimate
+        RETURN_POOLED: True
+        USE_FINAL_LAYER_NORM: False
+        IS_TRAINABLE: False
+      T5_XXL:
+        NAME: T5EmbedderHF
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_3
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_3
+        LENGTH: 256
+        CLEAN: whitespace
+        USE_GRAD: False
+        T5_DTYPE: float16
+    #
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: euler
+    SAMPLE_STEPS: 28
+    SEED: 1749023094
+    GUIDE_SCALE: 5.0
+    GUIDE_RESCALE: 0.0
+    DISCRETIZATION: trailing
+    RUN_TRAIN_N: False
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 1e-5
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a cat holds a blackboard that writes \"hello world\"", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 10000
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 10
+      SHOW_GPU_MEM: True
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  #
+  EVAL_HOOKS:
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 50
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/examples/generation/dit_sd3_1024_lora.yaml

@@ -0,0 +1,241 @@
+ENV:
+  BACKEND: nccl
+#
+SOLVER:
+  NAME: LatentDiffusionSolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 500
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: 50
+  RESCALE_LR: False
+  #
+  WORK_DIR: ./cache/save_data/dit_sd3_1024_lora
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    - NAME: "ModelscopeFs"
+      TEMP_DIR: ./cache/cache_data
+  #
+  TUNER:
+    - NAME: SwiftLoRA
+      R: 128
+      LORA_ALPHA: 128
+      LORA_DROPOUT: 0.0
+      BIAS: "none"
+      TARGET_MODULES: "model.*(.attn.qkv|.attn.proj|mlp.fc1|mlp.fc2)$"
+  #
+  MODEL:
+    NAME: LatentDiffusionSD3
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    SCALE_FACTOR: 1.5305
+    SHIFT_FACTOR: 0.0609
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "shifted"
+      "SHIFT": 3
+    USE_EMA: False
+    T_WEIGHT: uniform
+    #
+    DIFFUSION_MODEL:
+      NAME: MMDiT
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+      IGNORE_KEYS: '^first_stage_model.'
+      IN_CHANNELS: 16
+      PATCH_SIZE: 2
+      OUT_CHANNELS: 16
+      DEPTH: 24
+      INPUT_SIZE:
+      ADM_IN_CHANNELS: 2048
+      CONTEXT_EMBEDDER_CONFIG: { 'target': 'torch.nn.Linear', 'params': { 'in_features': 4096, 'out_features': 1536 } }
+      NUM_PATCHES: 36864
+      POS_EMBED_MAX_SIZE: 192
+      POS_EMBED_SCALING_FACTOR:
+      USE_CHECKPOINT: True
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+      EMBED_DIM: 16
+      IGNORE_KEYS: '^model.diffusion_model.'
+      BATCH_SIZE: 1
+      USE_CONV: False
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      NAME: SD3TextEmbedder
+      P_ZERO: 0.0
+      CLIP_L:
+        NAME: FrozenCLIPEmbedder2
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer
+        MAX_LENGTH: 77
+        FREEZE: True
+        LAYER: penultimate
+        RETURN_POOLED: True
+        USE_FINAL_LAYER_NORM: False
+        IS_TRAINABLE: False
+      CLIP_G:
+        NAME: FrozenCLIPEmbedder2
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_2
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_2
+        MAX_LENGTH: 77
+        FREEZE: True
+        LAYER: penultimate
+        RETURN_POOLED: True
+        USE_FINAL_LAYER_NORM: False
+        IS_TRAINABLE: False
+      T5_XXL:
+        NAME: T5EmbedderHF
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_3
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_3
+        LENGTH: 256
+        CLEAN: whitespace
+        USE_GRAD: False
+        T5_DTYPE: float16
+    #
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: euler
+    SAMPLE_STEPS: 28
+    SEED: 1749023094
+    GUIDE_SCALE: 5.0
+    GUIDE_RESCALE: 0.0
+    DISCRETIZATION: trailing
+    RUN_TRAIN_N: False
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 5e-5
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a cat holds a blackboard that writes \"hello world\"", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 10000
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 10
+      SHOW_GPU_MEM: True
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  #
+  EVAL_HOOKS:
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 50
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/examples/generation/stable_diffusion_1.5_512.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd15_512_full
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   MODEL:
     NAME: LatentDiffusion
@@ -124,7 +125,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.0064
+    LEARNING_RATE: 0.00001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -190,7 +191,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/examples/generation/stable_diffusion_1.5_512_lora.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd15_512_lora
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   TUNER:
     -
       NAME: SwiftLoRA
@@ -132,7 +133,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -198,7 +199,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/examples/generation/stable_diffusion_1.5_512_textlora.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd15_512_textlora
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   TUNER:
     -
       NAME: SwiftLoRA
@@ -140,7 +141,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -206,7 +207,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/examples/generation/stable_diffusion_2.1_512.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_512_full
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   MODEL:
     NAME: LatentDiffusion
@@ -120,7 +121,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.0064
+    LEARNING_RATE: 0.00001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -186,7 +187,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/examples/generation/stable_diffusion_2.1_512_lora.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_512_lora
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -129,7 +130,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.0064
+    LEARNING_RATE: 0.00001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -195,7 +196,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/examples/generation/stable_diffusion_2.1_768.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_768_full
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   MODEL:
     NAME: LatentDiffusion
@@ -120,7 +121,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.0064
+    LEARNING_RATE: 0.00001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -186,7 +187,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/examples/generation/stable_diffusion_2.1_768_lora.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_768_lora
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -129,7 +130,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -195,7 +196,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/examples/generation/stable_diffusion_xl_1024.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_full
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   MODEL:
     NAME: LatentDiffusionXL
@@ -238,7 +239,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.0064
+    LEARNING_RATE: 0.00001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -306,7 +307,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       - NAME: Select

scepter/methods/examples/generation/stable_diffusion_xl_1024_lora.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_lora
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -247,7 +248,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -315,7 +316,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       - NAME: Select

scepter/methods/examples/generation/stable_diffusion_xl_1024_textlora.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_textlora
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -255,7 +256,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -323,7 +324,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       - NAME: Select

scepter/methods/scedit/ctr/sd15_512_sce_ctr_hed.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd15_512_sce_ctr_hed
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -127,7 +128,7 @@ SOLVER:
         DOWN_RATIO: 1.0
     CONTROL_ANNO:
       NAME: HedAnnotator
-      PRETRAINED_MODEL: ms://damo/scepter_scedit@annotator/ckpts/ControlNetHED.pth
+      PRETRAINED_MODEL: ms://iic/scepter_scedit@annotator/ckpts/ControlNetHED.pth
   #
   SAMPLE_ARGS:
     SAMPLER: ddim
@@ -141,7 +142,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -167,13 +168,13 @@ SOLVER:
         RGB_ORDER: RGB
         BACKEND: pillow
       - NAME: Resize
-        SIZE: 768
+        SIZE: 512
         INTERPOLATION: bilinear
         INPUT_KEY: [ 'img' ]
         OUTPUT_KEY: [ 'img' ]
         BACKEND: pillow
       - NAME: CenterCrop
-        SIZE: 768
+        SIZE: 512
         INPUT_KEY: [ 'img' ]
         OUTPUT_KEY: [ 'img' ]
         BACKEND: pillow
@@ -215,13 +216,13 @@ SOLVER:
         RGB_ORDER: RGB
         BACKEND: pillow
       - NAME: Resize
-        SIZE: 768
+        SIZE: 512
         INTERPOLATION: bilinear
         INPUT_KEY: [ 'img' ]
         OUTPUT_KEY: [ 'img' ]
         BACKEND: pillow
       - NAME: CenterCrop
-        SIZE: 768
+        SIZE: 512
         INPUT_KEY: [ 'img' ]
         OUTPUT_KEY: [ 'img' ]
         BACKEND: pillow

scepter/methods/scedit/ctr/sd21_768_sce_ctr_canny.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_768_sce_ctr_canny
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -140,7 +141,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2

scepter/methods/scedit/ctr/sd21_768_sce_ctr_pose.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_768_sce_ctr_pose
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -125,8 +126,8 @@ SOLVER:
         DOWN_RATIO: 1.0
     CONTROL_ANNO:
       NAME: OpenposeAnnotator
-      BODY_MODEL_PATH: ms://damo/scepter_scedit@annotator/ckpts/body_pose_model.pth
-      HAND_MODEL_PATH: ms://damo/scepter_scedit@annotator/ckpts/hand_pose_model.pth
+      BODY_MODEL_PATH: ms://iic/scepter_scedit@annotator/ckpts/body_pose_model.pth
+      HAND_MODEL_PATH: ms://iic/scepter_scedit@annotator/ckpts/hand_pose_model.pth
   #
   SAMPLE_ARGS:
     SAMPLER: ddim
@@ -140,7 +141,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2

scepter/methods/scedit/ctr/sdxl_1024_sce_ctr_canny.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_sce_ctr_canny
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -254,7 +255,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2

scepter/methods/scedit/ctr/sdxl_1024_sce_ctr_color.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_sce_ctr_color
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -255,7 +256,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2

scepter/methods/scedit/ctr/sdxl_1024_sce_ctr_color_datatxt.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_sce_ctr_color_datatxt
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -255,7 +256,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2

scepter/methods/scedit/ctr/sdxl_1024_sce_ctr_depth.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_sce_ctr_depth
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -241,7 +242,7 @@ SOLVER:
         DOWN_RATIO: 1.0
     CONTROL_ANNO:
       NAME: MidasDetector
-      PRETRAINED_MODEL: ms://damo/scepter_scedit@annotator/ckpts/dpt_hybrid-midas-501f0c75.pt
+      PRETRAINED_MODEL: ms://iic/scepter_scedit@annotator/ckpts/dpt_hybrid-midas-501f0c75.pt
   #
   SAMPLE_ARGS:
     SAMPLER: ddim
@@ -255,7 +256,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2

scepter/methods/scedit/t2i/sd15_512_sce_t2i.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd15_512_sce_t2i
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -134,7 +135,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -200,7 +201,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/scedit/t2i/sd15_512_sce_t2i_swift.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd15_512_sce_t2i_swift
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -132,7 +133,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -198,7 +199,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/scedit/t2i/sd15_512_textsce_t2i_swift.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd15_512_textsce_t2i_swift
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -140,7 +141,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -206,7 +207,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/scedit/t2i/sd21_768_sce_t2i.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_768_sce_t2i
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -130,7 +131,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -196,7 +197,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/scedit/t2i/sd21_768_sce_t2i_swift.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sd21_768_sce_t2i_swift
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -128,7 +129,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -194,7 +195,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       -

scepter/methods/scedit/t2i/sdxl_1024_sce_t2i.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_sce_t2i
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -247,7 +248,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -315,7 +316,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       - NAME: Select

scepter/methods/scedit/t2i/sdxl_1024_sce_t2i_datatxt.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_sce_t2i_datatxt
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
     FREEZE_PART: [ "first_stage_model", "cond_stage_model", "model" ]
@@ -247,7 +248,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2

scepter/methods/scedit/t2i/sdxl_1024_sce_t2i_swift.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_sce_t2i_swift
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -245,7 +246,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -313,7 +314,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       - NAME: Select

scepter/methods/scedit/t2i/sdxl_1024_textsce_t2i_swift.yaml

@@ -14,13 +14,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: 100
+  RESCALE_LR: False
   #
   WORK_DIR: ./cache/save_data/sdxl_1024_textsce_t2i_swift
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   TUNER:
     -
@@ -253,7 +254,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2
@@ -321,7 +322,7 @@ SOLVER:
     NUM_WORKERS: 4
     FILE_SYSTEM:
       NAME: "ModelscopeFs"
-      TEMP_DIR: "./cache/data"
+      TEMP_DIR: "./cache/cache_data"
     #
     TRANSFORMS:
       - NAME: Select

scepter/methods/studio/chatbot/chatbot.yaml

@@ -0,0 +1,25 @@
+WORK_DIR: chatbot
+FILE_SYSTEM:
+  - NAME: LocalFs
+    TEMP_DIR: ./cache/cache_data
+  - NAME: ModelscopeFs
+    TEMP_DIR: ./cache/cache_data
+  - NAME: HuggingfaceFs
+    TEMP_DIR: ./cache/cache_data
+#
+ENABLE_I2V: False
+#
+MODEL:
+  EDIT_MODEL:
+    MODEL_CFG_DIR: scepter/methods/studio/chatbot/models/
+    DEFAULT: ace_0.6b_512
+  I2V:
+    MODEL_NAME: CogVideoX-5b-I2V
+    MODEL_DIR: ms://ZhipuAI/CogVideoX-5b-I2V/
+  CAPTIONER:
+    MODEL_NAME: InternVL2-2B
+    MODEL_DIR: ms://OpenGVLab/InternVL2-2B/
+    PROMPT: '<image>\nThis image is the first frame of a video. Based on this image, please imagine what changes may occur in the next few seconds of the video. Please output brief description, such as "a dog running" or "a person turns to left". No more than 30 words.'
+  ENHANCER:
+    MODEL_NAME: Meta-Llama-3.1-8B-Instruct
+    MODEL_DIR: ms://LLM-Research/Meta-Llama-3.1-8B-Instruct/

scepter/methods/studio/chatbot/models/ace_0.6b_512.yaml

@@ -0,0 +1,127 @@
+NAME: ACE_0.6B_512
+IS_DEFAULT: False
+DEFAULT_PARAS:
+  PARAS:
+  #
+  INPUT:
+    INPUT_IMAGE:
+    INPUT_MASK:
+    TASK:
+    PROMPT: ""
+    NEGATIVE_PROMPT: ""
+    OUTPUT_HEIGHT: 512
+    OUTPUT_WIDTH: 512
+    SAMPLER: ddim
+    SAMPLE_STEPS: 20
+    GUIDE_SCALE: 4.5
+    GUIDE_RESCALE: 0.5
+    SEED: -1
+    TAR_INDEX: 0
+  OUTPUT:
+    LATENT:
+    IMAGES:
+    SEED:
+  MODULES_PARAS:
+    FIRST_STAGE_MODEL:
+      FUNCTION:
+        - NAME: encode
+          DTYPE: float16
+          INPUT: ["IMAGE"]
+        - NAME: decode
+          DTYPE: float16
+          INPUT: ["LATENT"]
+    #
+    DIFFUSION_MODEL:
+      FUNCTION:
+        - NAME: forward
+          DTYPE: float16
+          INPUT: ["SAMPLE_STEPS", "SAMPLE", "GUIDE_SCALE"]
+    #
+    COND_STAGE_MODEL:
+      FUNCTION:
+        - NAME: encode_list
+          DTYPE: bfloat16
+          INPUT: ["PROMPT"]
+#
+MODEL:
+  NAME: LatentDiffusionACE
+  PRETRAINED_MODEL:
+  IGNORE_KEYS: [ ]
+  SCALE_FACTOR: 0.18215
+  SIZE_FACTOR: 8
+  DECODER_BIAS: 0.5
+  DEFAULT_N_PROMPT: ""
+  TEXT_IDENTIFIER: [ '{image}', '{image1}', '{image2}', '{image3}', '{image4}', '{image5}', '{image6}', '{image7}', '{image8}', '{image9}' ]
+  USE_TEXT_POS_EMBEDDINGS: True
+  #
+  DIFFUSION:
+    NAME: BaseDiffusion
+    PREDICTION_TYPE: eps
+    MIN_SNR_GAMMA:
+    NOISE_SCHEDULER:
+      NAME: LinearScheduler
+      NUM_TIMESTEPS: 1000
+      BETA_MIN: 0.0001
+      BETA_MAX: 0.02
+  #
+  DIFFUSION_MODEL:
+    NAME: ACE
+    PRETRAINED_MODEL: ms://iic/ACE-0.6B-512px@models/dit/ace_0.6b_512px.pth
+    IGNORE_KEYS: [ ]
+    PATCH_SIZE: 2
+    IN_CHANNELS: 4
+    HIDDEN_SIZE: 1152
+    DEPTH: 28
+    NUM_HEADS: 16
+    MLP_RATIO: 4.0
+    PRED_SIGMA: True
+    DROP_PATH: 0.0
+    WINDOW_DIZE: 0
+    Y_CHANNELS: 4096
+    MAX_SEQ_LEN: 1024
+    QK_NORM: True
+    USE_GRAD_CHECKPOINT: True
+    ATTENTION_BACKEND: flash_attn
+  #
+  FIRST_STAGE_MODEL:
+    NAME: AutoencoderKL
+    EMBED_DIM: 4
+    PRETRAINED_MODEL: ms://iic/ACE-0.6B-512px@models/vae/vae.bin
+    IGNORE_KEYS: []
+    #
+    ENCODER:
+      NAME: Encoder
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 4
+      DOUBLE_Z: True
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+    #
+    DECODER:
+      NAME: Decoder
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 4
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+      GIVE_PRE_END: False
+      TANH_OUT: False
+  #
+  COND_STAGE_MODEL:
+    NAME: T5EmbedderHF
+    PRETRAINED_MODEL: ms://iic/ACE-0.6B-512px@models/text_encoder/t5-v1_1-xxl/
+    TOKENIZER_PATH: ms://iic/ACE-0.6B-512px@models/tokenizer/t5-v1_1-xxl
+    LENGTH: 120
+    T5_DTYPE: bfloat16
+    ADDED_IDENTIFIER: [ '{image}', '{caption}', '{mask}', '{ref_image}', '{image1}', '{image2}', '{image3}', '{image4}', '{image5}', '{image6}', '{image7}', '{image8}', '{image9}' ]
+    CLEAN: whitespace
+    USE_GRAD: False

scepter/methods/studio/extensions/controllers/official_controllers.yaml

@@ -5,59 +5,59 @@ CONTROLLERS:
     DESCRIPTION:
     BASE_MODEL: SD2.1
     TYPE: Canny
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD2.1/canny_control/
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD2.1/canny_control/
   - NAME: openpose
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD2.1
     TYPE: Openpose
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD2.1/pose_control/
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD2.1/pose_control/
   - NAME: color
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD2.1
     TYPE: Color
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD2.1/color_control/
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD2.1/color_control/
   - NAME: hed
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD2.1
     TYPE: Hed
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD2.1/hed_control
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD2.1/hed_control
   - NAME: depth
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD2.1
     TYPE: Midas
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD2.1/depth_control
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD2.1/depth_control
   # SD_XL1.0
   - NAME: canny
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD_XL1.0
     TYPE: Canny
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD_XL1.0/canny_control
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD_XL1.0/canny_control
   - NAME: color
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD_XL1.0
     TYPE: Color
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD_XL1.0/color_control
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD_XL1.0/color_control
   - NAME: depth
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD_XL1.0
     TYPE: Midas
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD_XL1.0/depth_control
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD_XL1.0/depth_control
   - NAME: hed
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD_XL1.0
     TYPE: Hed
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD_XL1.0/hed_control
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD_XL1.0/hed_control
   - NAME: openpose
     NAME_ZH:
     DESCRIPTION:
     BASE_MODEL: SD_XL1.0
     TYPE: Openpose
-    MODEL_PATH: ms://damo/scepter_scedit@controllable_model/SD_XL1.0/pose_control
+    MODEL_PATH: ms://iic/scepter_scedit@controllable_model/SD_XL1.0/pose_control

scepter/methods/studio/extensions/tuners/official_tuners.yaml

@@ -1,11 +1,20 @@
 TUNERS:
+  - NAME: Clay-Style-Editing
+    NAME_ZH: 黏土风
+    SOURCE: scepter
+    DESCRIPTION: None
+    BASE_MODEL: EDIT
+    MODEL_PATH: ms://iic/stylebooth@tuners/clay_style_edit/
+    IMAGE_PATH: ms://iic/stylebooth@tuners/clay_style_edit/image.jpg
+    TUNER_TYPE: LORA
+    PROMPT_EXAMPLE: Convert this image into clay style
   - NAME: Azure-Dragon
     NAME_ZH: 青龙
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/azure_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/azure_dragon/xl_azure_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/azure_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/azure_dragon/xl_azure_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: Azure Dragon, 8K, high quality,Ultra High Detail.One of the Four Divine Creatures in Charge of Water.
   - NAME: Gold-Dragon
@@ -13,8 +22,8 @@ TUNERS:
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/gold_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/gold_dragon/xl_gold_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/gold_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/gold_dragon/xl_gold_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: Chinese Gold Dragon in the clouds. Translucent Texture. Zbrush.  Fuzzy Art. Exquisite Craftsmanship. 3D. 8K. Ultra High Detail
   - NAME: SpringFestival-Dragon
@@ -22,8 +31,8 @@ TUNERS:
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/spring_festival_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/spring_festival_dragon/xl_spring_festival_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/spring_festival_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/spring_festival_dragon/xl_spring_festival_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: Chinese dragon. Spring Festival.Festive.Street.Lanterns.32K.High quality.expressive, dramatic, dreamlike and mysterious, Surrealism
   - NAME: Red-Dragon
@@ -31,8 +40,8 @@ TUNERS:
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/red_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/red_dragon/xl_red_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/red_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/red_dragon/xl_red_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: Traditional Red Dragon of China. Low Water Level. Studio Ghibli Style. Mural Illustration. White Background. High Detail
   - NAME: ChinesePunk-Dragon
@@ -40,8 +49,8 @@ TUNERS:
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/chinese_punk_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/chinese_punk_dragon/xl_chinese_punk_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/chinese_punk_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/chinese_punk_dragon/xl_chinese_punk_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: uhd Image,Dragon,Chinese Dragon, Dunhuang Mural Style, Traditional Maritime Art Style
   - NAME: Cute-Dragon
@@ -49,8 +58,8 @@ TUNERS:
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/cute_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/cute_dragon/xl_kawaii_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/cute_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/cute_dragon/xl_kawaii_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: China Kawaii Dragon. Contest Winner. Minimalist Illustration. White Background. Flat Style. Digital Painting Style. Red. 32k uhd. Fun Comics. Fuzzy Art. Bold. Comic-Inspired Characters
   - NAME: Dragon-Baby
@@ -58,8 +67,8 @@ TUNERS:
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/baby_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/baby_dragon/xl_baby_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/baby_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/baby_dragon/xl_baby_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: Warm Colors, Soft,Chinese Dragon Baby, Felt Style,Dragon Baby, Best Quality, 3D Doll, Macaron Tones, Glittering Big Eyes, Winter,Dragon
   - NAME: Sloppy-Dragon
@@ -67,8 +76,8 @@ TUNERS:
     SOURCE: scepter
     DESCRIPTION: None
     BASE_MODEL: SD_XL1.0
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sloppy_dragon/
-    IMAGE_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sloppy_dragon/xl_sloppy_dragon.png
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sloppy_dragon/
+    IMAGE_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sloppy_dragon/xl_sloppy_dragon.png
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: Messy Chinese Dragon,Cute, Wu Guanzhong, Rough
   -
@@ -77,8 +86,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/894f40ed44b37c3372e6a22b8ae577a4.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/Caricature
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/894f40ed44b37c3372e6a22b8ae577a4.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/Caricature
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -87,8 +96,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/894f40ed44b37c3372e6a22b8ae577a4.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/Caricature
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/894f40ed44b37c3372e6a22b8ae577a4.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/Caricature
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -97,8 +106,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/894f40ed44b37c3372e6a22b8ae577a4.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/Caricature
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/894f40ed44b37c3372e6a22b8ae577a4.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/Caricature
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -107,8 +116,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/80e5b4075c572c04cbb4e48c37b8366b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/ColorFieldPainting
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/80e5b4075c572c04cbb4e48c37b8366b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/ColorFieldPainting
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -117,8 +126,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/80e5b4075c572c04cbb4e48c37b8366b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/ColorFieldPainting
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/80e5b4075c572c04cbb4e48c37b8366b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/ColorFieldPainting
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -127,8 +136,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/80e5b4075c572c04cbb4e48c37b8366b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/ColorFieldPainting
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/80e5b4075c572c04cbb4e48c37b8366b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/ColorFieldPainting
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -137,8 +146,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/9ae235d7f1a7c2a4edab52a5e9f9cbae.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/ColoredPencilArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/9ae235d7f1a7c2a4edab52a5e9f9cbae.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/ColoredPencilArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -147,8 +156,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/9ae235d7f1a7c2a4edab52a5e9f9cbae.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/ColoredPencilArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/9ae235d7f1a7c2a4edab52a5e9f9cbae.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/ColoredPencilArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -157,8 +166,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/9ae235d7f1a7c2a4edab52a5e9f9cbae.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/ColoredPencilArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/9ae235d7f1a7c2a4edab52a5e9f9cbae.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/ColoredPencilArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -167,8 +176,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/3da915da2f5cedaf243e57e08163f35b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/DarkMoodyAtmosphere
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/3da915da2f5cedaf243e57e08163f35b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/DarkMoodyAtmosphere
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -177,8 +186,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/3da915da2f5cedaf243e57e08163f35b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/DarkMoodyAtmosphere
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/3da915da2f5cedaf243e57e08163f35b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/DarkMoodyAtmosphere
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -187,8 +196,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/3da915da2f5cedaf243e57e08163f35b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/DarkMoodyAtmosphere
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/3da915da2f5cedaf243e57e08163f35b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/DarkMoodyAtmosphere
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -197,8 +206,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/69fd81f5983107acc3d334af62915851.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/DrippingPaintSplatterArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/69fd81f5983107acc3d334af62915851.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/DrippingPaintSplatterArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -207,8 +216,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/69fd81f5983107acc3d334af62915851.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/DrippingPaintSplatterArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/69fd81f5983107acc3d334af62915851.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/DrippingPaintSplatterArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -217,8 +226,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/69fd81f5983107acc3d334af62915851.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/DrippingPaintSplatterArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/69fd81f5983107acc3d334af62915851.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/DrippingPaintSplatterArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -227,8 +236,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/f152edb4b3ca6248758b48115258ddfa.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/FadedPolaroidPhoto
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/f152edb4b3ca6248758b48115258ddfa.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/FadedPolaroidPhoto
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -237,8 +246,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/f152edb4b3ca6248758b48115258ddfa.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/FadedPolaroidPhoto
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/f152edb4b3ca6248758b48115258ddfa.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/FadedPolaroidPhoto
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -247,8 +256,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/f152edb4b3ca6248758b48115258ddfa.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/FadedPolaroidPhoto
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/f152edb4b3ca6248758b48115258ddfa.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/FadedPolaroidPhoto
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -257,8 +266,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/940cfd34155634cf051e1b2942cca426.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/Flat2DArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/940cfd34155634cf051e1b2942cca426.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/Flat2DArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -267,8 +276,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/940cfd34155634cf051e1b2942cca426.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/Flat2DArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/940cfd34155634cf051e1b2942cca426.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/Flat2DArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -277,8 +286,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/940cfd34155634cf051e1b2942cca426.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/Flat2DArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/940cfd34155634cf051e1b2942cca426.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/Flat2DArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -287,8 +296,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/57b751b11564cb22cd49ef21f2004a5f.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/GraffitiArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/57b751b11564cb22cd49ef21f2004a5f.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/GraffitiArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -297,8 +306,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/57b751b11564cb22cd49ef21f2004a5f.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/GraffitiArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/57b751b11564cb22cd49ef21f2004a5f.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/GraffitiArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -307,8 +316,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/57b751b11564cb22cd49ef21f2004a5f.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/GraffitiArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/57b751b11564cb22cd49ef21f2004a5f.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/GraffitiArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -317,8 +326,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/0312b673dc6858a9864d7f45f0c5c1fc.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/Impressionism
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/0312b673dc6858a9864d7f45f0c5c1fc.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/Impressionism
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -327,8 +336,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/0312b673dc6858a9864d7f45f0c5c1fc.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/Impressionism
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/0312b673dc6858a9864d7f45f0c5c1fc.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/Impressionism
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -337,8 +346,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/0312b673dc6858a9864d7f45f0c5c1fc.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/Impressionism
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/0312b673dc6858a9864d7f45f0c5c1fc.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/Impressionism
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -347,8 +356,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/9aa040b0c60d289da9610c91ad9b7c7e.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/LogoDesign
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/9aa040b0c60d289da9610c91ad9b7c7e.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/LogoDesign
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -357,8 +366,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/9aa040b0c60d289da9610c91ad9b7c7e.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/LogoDesign
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/9aa040b0c60d289da9610c91ad9b7c7e.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/LogoDesign
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -367,8 +376,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/9aa040b0c60d289da9610c91ad9b7c7e.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/LogoDesign
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/9aa040b0c60d289da9610c91ad9b7c7e.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/LogoDesign
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -377,8 +386,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/a9056e1eac85e5e4fe96a93917d4cce4.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/PencilSketchDrawing
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/a9056e1eac85e5e4fe96a93917d4cce4.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/PencilSketchDrawing
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -387,8 +396,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/a9056e1eac85e5e4fe96a93917d4cce4.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/PencilSketchDrawing
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/a9056e1eac85e5e4fe96a93917d4cce4.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/PencilSketchDrawing
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -397,8 +406,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/a9056e1eac85e5e4fe96a93917d4cce4.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/PencilSketchDrawing
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/a9056e1eac85e5e4fe96a93917d4cce4.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/PencilSketchDrawing
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -407,8 +416,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/568777f447fc02510b618152726d5002.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/SilhouetteArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/568777f447fc02510b618152726d5002.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/SilhouetteArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -417,8 +426,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/568777f447fc02510b618152726d5002.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/SilhouetteArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/568777f447fc02510b618152726d5002.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/SilhouetteArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -427,8 +436,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/568777f447fc02510b618152726d5002.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/SilhouetteArt
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/568777f447fc02510b618152726d5002.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/SilhouetteArt
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -437,8 +446,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/07d7b27cd73f2d43684003563511c15b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/Steampunk2
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/07d7b27cd73f2d43684003563511c15b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/Steampunk2
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -447,8 +456,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/07d7b27cd73f2d43684003563511c15b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/Steampunk2
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/07d7b27cd73f2d43684003563511c15b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/Steampunk2
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -457,8 +466,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/07d7b27cd73f2d43684003563511c15b.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/Steampunk2
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/07d7b27cd73f2d43684003563511c15b.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/Steampunk2
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -467,8 +476,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/2d1e9867058db2c57f2fe47530de3243.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/StickerDesigns
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/2d1e9867058db2c57f2fe47530de3243.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/StickerDesigns
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -477,8 +486,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/2d1e9867058db2c57f2fe47530de3243.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/StickerDesigns
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/2d1e9867058db2c57f2fe47530de3243.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/StickerDesigns
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -487,8 +496,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/2d1e9867058db2c57f2fe47530de3243.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/StickerDesigns
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/2d1e9867058db2c57f2fe47530de3243.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/StickerDesigns
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -497,8 +506,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/8859d532ae5901cc8457d6118fb9b7da.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/Watercolor2
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/8859d532ae5901cc8457d6118fb9b7da.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/Watercolor2
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -507,8 +516,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/8859d532ae5901cc8457d6118fb9b7da.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/Watercolor2
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/8859d532ae5901cc8457d6118fb9b7da.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/Watercolor2
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -517,8 +526,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: diva
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/8859d532ae5901cc8457d6118fb9b7da.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/Watercolor2
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/8859d532ae5901cc8457d6118fb9b7da.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/Watercolor2
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -527,8 +536,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/5895d78cf58c1ca05178991f37cc48ff.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/mre-elemental-art
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/5895d78cf58c1ca05178991f37cc48ff.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/mre-elemental-art
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -537,8 +546,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/5895d78cf58c1ca05178991f37cc48ff.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/mre-elemental-art
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/5895d78cf58c1ca05178991f37cc48ff.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/mre-elemental-art
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -547,8 +556,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/5895d78cf58c1ca05178991f37cc48ff.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/mre-elemental-art
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/5895d78cf58c1ca05178991f37cc48ff.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/mre-elemental-art
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -557,8 +566,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/a08149bc8e50f6bc65c0010d4cd416f8.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/mre-anime
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/a08149bc8e50f6bc65c0010d4cd416f8.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/mre-anime
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -567,8 +576,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/a08149bc8e50f6bc65c0010d4cd416f8.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/mre-anime
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/a08149bc8e50f6bc65c0010d4cd416f8.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/mre-anime
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -577,8 +586,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/a08149bc8e50f6bc65c0010d4cd416f8.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/mre-anime
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/a08149bc8e50f6bc65c0010d4cd416f8.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/mre-anime
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -587,8 +596,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/48c65cebf1fa4284d7b8feb619412e65.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/mre-comic
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/48c65cebf1fa4284d7b8feb619412e65.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/mre-comic
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -597,8 +606,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/48c65cebf1fa4284d7b8feb619412e65.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/mre-comic
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/48c65cebf1fa4284d7b8feb619412e65.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/mre-comic
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -607,8 +616,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: mre
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/48c65cebf1fa4284d7b8feb619412e65.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/mre-comic
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/48c65cebf1fa4284d7b8feb619412e65.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/mre-comic
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -617,8 +626,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/8fc51113f725f27326c4398a7457cd6d.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sai-craftclay
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/8fc51113f725f27326c4398a7457cd6d.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sai-craftclay
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -627,8 +636,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/8fc51113f725f27326c4398a7457cd6d.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/sai-craftclay
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/8fc51113f725f27326c4398a7457cd6d.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/sai-craftclay
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -637,8 +646,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/8fc51113f725f27326c4398a7457cd6d.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/sai-craftclay
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/8fc51113f725f27326c4398a7457cd6d.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/sai-craftclay
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -647,8 +656,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/a6f8d92afcd5803dfb2ebecbc92091b6.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sai-fantasyart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/a6f8d92afcd5803dfb2ebecbc92091b6.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sai-fantasyart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -657,8 +666,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/a6f8d92afcd5803dfb2ebecbc92091b6.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/sai-fantasyart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/a6f8d92afcd5803dfb2ebecbc92091b6.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/sai-fantasyart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -667,8 +676,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/a6f8d92afcd5803dfb2ebecbc92091b6.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/sai-fantasyart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/a6f8d92afcd5803dfb2ebecbc92091b6.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/sai-fantasyart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -677,8 +686,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/034a51b0dd34b018be8859bf45b4f7ed.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sai-lineart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/034a51b0dd34b018be8859bf45b4f7ed.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sai-lineart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -687,8 +696,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/034a51b0dd34b018be8859bf45b4f7ed.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/sai-lineart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/034a51b0dd34b018be8859bf45b4f7ed.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/sai-lineart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -697,8 +706,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/034a51b0dd34b018be8859bf45b4f7ed.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/sai-lineart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/034a51b0dd34b018be8859bf45b4f7ed.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/sai-lineart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -707,8 +716,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/7e9ed25bb34008beb5f417df63c4b2fe.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sai-neonpunk
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/7e9ed25bb34008beb5f417df63c4b2fe.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sai-neonpunk
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -717,8 +726,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/7e9ed25bb34008beb5f417df63c4b2fe.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/sai-neonpunk
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/7e9ed25bb34008beb5f417df63c4b2fe.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/sai-neonpunk
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -727,8 +736,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/7e9ed25bb34008beb5f417df63c4b2fe.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/sai-neonpunk
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/7e9ed25bb34008beb5f417df63c4b2fe.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/sai-neonpunk
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -737,8 +746,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/924f46a8f276011a0953d7988e90ee25.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sai-origami
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/924f46a8f276011a0953d7988e90ee25.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sai-origami
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -747,8 +756,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/924f46a8f276011a0953d7988e90ee25.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/sai-origami
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/924f46a8f276011a0953d7988e90ee25.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/sai-origami
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -757,8 +766,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/924f46a8f276011a0953d7988e90ee25.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/sai-origami
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/924f46a8f276011a0953d7988e90ee25.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/sai-origami
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -767,8 +776,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD_XL1.0
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD_XL1.0/a5ab89c0960be8c1216e65c98d92ae4a.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD_XL1.0/sai-pixelart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD_XL1.0/a5ab89c0960be8c1216e65c98d92ae4a.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD_XL1.0/sai-pixelart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -777,8 +786,8 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD2.1
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD2.1/a5ab89c0960be8c1216e65c98d92ae4a.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD2.1/sai-pixelart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD2.1/a5ab89c0960be8c1216e65c98d92ae4a.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD2.1/sai-pixelart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket
   -
@@ -787,7 +796,7 @@ TUNERS:
     DESCRIPTION:
     SOURCE: sai
     BASE_MODEL: SD1.5
-    IMAGE_PATH: ms://damo/scepter@mantra_images/SD1.5/a5ab89c0960be8c1216e65c98d92ae4a.png
-    MODEL_PATH: ms://damo/scepter_scedit@tuners_model/SD1.5/sai-pixelart
+    IMAGE_PATH: ms://iic/scepter@mantra_images_jpg/SD1.5/a5ab89c0960be8c1216e65c98d92ae4a.jpg
+    MODEL_PATH: ms://iic/scepter_scedit@tuners_model/SD1.5/sai-pixelart
     TUNER_TYPE: SwiftSCE
     PROMPT_EXAMPLE: a boy wearing green jacket

scepter/methods/studio/home/home.yaml

@@ -10,7 +10,7 @@ DESC_INFO:
     <table align="center">
       <tr>
       <td>
-        <img src="https://modelscope.cn/api/v1/models/damo/scepter/repo?Revision=master&FilePath=assets/scepter_studio/scepter_studio_banner.jpg">
+        <img src="https://modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets/scepter_studio/scepter_studio_banner.jpg">
         <h3><center>SCEPTER Studio是基于开源基模型和自研微调编辑算法构建的生成定制和编辑工具箱，提供围绕生成、微调、编辑、数据处理等一系列的工具和插件。</center><h3>
       </td>
       </tr>
@@ -22,7 +22,7 @@ DESC_INFO:
     <table align="center">
       <tr>
       <td>
-        <img src="https://modelscope.cn/api/v1/models/damo/scepter/repo?Revision=master&FilePath=assets/scepter_studio/scepter_studio_banner.jpg">
+        <img src="https://modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets/scepter_studio/scepter_studio_banner.jpg">
         <h3><center>SCEPTER Studio is a customized generation and editing toolkit built on the open-source base models and proprietary fine-tuning editing algorithms, offering a range of tools and plugins centered around generation, fine-tuning, editing, and data processing.</center><h3>
       </td>
       </tr>

scepter/methods/studio/inference/dit/flux1.0_dev_pro.yaml

@@ -0,0 +1,201 @@
+NAME: FLUX1.0_DEV
+IS_DEFAULT: False
+DEFAULT_PARAS:
+  PARAS:
+    RESOLUTIONS: [[1024, 1024]]
+  INPUT:
+    IMAGE:
+    ORIGINAL_SIZE_AS_TUPLE: [1024, 1024]
+    TARGET_SIZE_AS_TUPLE: [1024, 1024]
+    PROMPT: ""
+    NEGATIVE_PROMPT:
+      DEFAULT: ""
+      VISIBLE: False
+    PROMPT_PREFIX: ""
+    SAMPLE:
+      VALUES: ["flow_eluer"]
+      DEFAULT: "flow_eluer"
+    SAMPLE_STEPS: 50
+    GUIDE_SCALE: 3.5
+    GUIDE_RESCALE:
+      DEFAULT: 0.0
+      VISIBLE: False
+    DISCRETIZATION:
+      VALUES: []
+      DEFAULT:
+      VISIBLE: False
+  OUTPUT:
+    LATENT:
+    IMAGES:
+    SEED:
+  MODULES_PARAS:
+    FIRST_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: bfloat16
+          INPUT: ["IMAGE"]
+        -
+          NAME: decode
+          DTYPE: bfloat16
+          INPUT: ["LATENT"]
+      PARAS:
+        SCALE_FACTOR: 1.5305
+        SHIFT_FACTOR: 0.0609
+        SIZE_FACTOR: 8
+    DIFFUSION_MODEL:
+      FUNCTION:
+        -
+          NAME: forward
+          DTYPE: bfloat16
+          INPUT: ["SAMPLE_STEPS", "SAMPLE", "GUIDE_SCALE"]
+    COND_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: bfloat16
+          INPUT: ["PROMPT"]
+#
+MODEL:
+  NAME: LatentDiffusionFlux
+  PARAMETERIZATION: rf
+  TIMESTEPS: 1000
+  MIN_SNR_GAMMA:
+  ZERO_TERMINAL_SNR: False
+  PRETRAINED_MODEL:
+  IGNORE_KEYS: [ ]
+  DEFAULT_N_PROMPT:
+  USE_EMA: False
+  EVAL_EMA: False
+  DIFFUSION:
+    # NAME DESCRIPTION:  TYPE:  default: 'DiffusionFluxRF'
+    NAME: DiffusionFluxRF
+    PREDICTION_TYPE: raw
+    # NOISE_SCHEDULER DESCRIPTION:  TYPE:  default: ''
+    NOISE_SCHEDULER:
+      # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchSigmaScheduler'
+      NAME: FlowMatchSigmaScheduler
+      # WEIGHTING_SCHEME DESCRIPTION: The weighting scheme for sampling timesteps, choose from ['sigma_sqrt', 'logit_normal', 'mode', 'cosmap', 'none']. TYPE: str default: 'logit_normal'
+      WEIGHTING_SCHEME: logit_normal
+      SHIFT: 3.0
+      # LOGIT_MEAN DESCRIPTION: The mean of the logit distribution for sampling timesteps. TYPE: float default: 0.0
+      LOGIT_MEAN: 0.0
+      # LOGIT_STD DESCRIPTION: The standard deviation of the logit distribution for sampling timesteps. TYPE: float default: 1.0
+      LOGIT_STD: 1.0
+      # MODE_SCALE DESCRIPTION: The scale factor for the mode of the logit distribution for sampling timesteps. TYPE: float default: 1.29
+      MODE_SCALE: 1.29
+      #
+  DIFFUSION_MODEL:
+    # NAME DESCRIPTION:  TYPE:  default: 'Flux'
+    NAME: Flux
+    PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@flux1-dev.safetensors
+    # IN_CHANNELS DESCRIPTION: model's input channels. TYPE: int default: 64
+    IN_CHANNELS: 64
+    # HIDDEN_SIZE DESCRIPTION: model's hidden size. TYPE: int default: 1024
+    HIDDEN_SIZE: 3072
+    # NUM_HEADS DESCRIPTION: number of heads in the transformer. TYPE: int default: 16
+    NUM_HEADS: 24
+    # AXES_DIM DESCRIPTION: dimensions of the axes of the positional encoding. TYPE: list default: [16, 56, 56]
+    AXES_DIM: [ 16, 56, 56 ]
+    # THETA DESCRIPTION: theta for positional encoding. TYPE: int default: 10000
+    THETA: 10000
+    # VEC_IN_DIM DESCRIPTION: dimension of the vector input. TYPE: int default: 768
+    VEC_IN_DIM: 768
+    # GUIDANCE_EMBED DESCRIPTION: whether to use guidance embedding. TYPE: bool default: False
+    GUIDANCE_EMBED: True
+    # CONTEXT_IN_DIM DESCRIPTION: dimension of the context input. TYPE: int default: 4096
+    CONTEXT_IN_DIM: 4096
+    # MLP_RATIO DESCRIPTION: ratio of mlp hidden size to hidden size. TYPE: float default: 4.0
+    MLP_RATIO: 4.0
+    # QKV_BIAS DESCRIPTION: whether to use bias in qkv projection. TYPE: bool default: True
+    QKV_BIAS: True
+    # DEPTH DESCRIPTION: number of transformer blocks. TYPE: int default: 19
+    DEPTH: 19
+    # DEPTH_SINGLE_BLOCKS DESCRIPTION: number of transformer blocks in the single stream block. TYPE: int default: 38
+    DEPTH_SINGLE_BLOCKS: 38
+
+  #
+  FIRST_STAGE_MODEL:
+    NAME: AutoencoderKLFlux
+    EMBED_DIM: 16
+    PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@ae.safetensors
+    IGNORE_KEYS: [ ]
+    BATCH_SIZE: 8
+    USE_CONV: False
+    SCALE_FACTOR: 0.3611
+    SHIFT_FACTOR: 0.1159
+    #
+    ENCODER:
+      NAME: Encoder
+      USE_CHECKPOINT: True
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 16
+      DOUBLE_Z: True
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+    #
+    DECODER:
+      NAME: Decoder
+      USE_CHECKPOINT: True
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 16
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+      GIVE_PRE_END: False
+      TANH_OUT: False
+  #
+  COND_STAGE_MODEL:
+    # NAME DESCRIPTION:  TYPE:  default: 'T5PlusClipFluxEmbedder'
+    NAME: T5PlusClipFluxEmbedder
+    # T5_MODEL DESCRIPTION:  TYPE:  default: ''
+    T5_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+      NAME: HFEmbedder
+      # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_MODEL_CLS: T5EncoderModel
+      # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+      MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder_2/
+      # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_TOKENIZER_CLS: T5Tokenizer
+      # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+      TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer_2/
+      # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+      MAX_LENGTH: 512
+      # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+      OUTPUT_KEY: last_hidden_state
+      # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+      D_TYPE: bfloat16
+      # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+      BATCH_INFER: False
+      CLEAN: whitespace
+    # CLIP_MODEL DESCRIPTION:  TYPE:  default: ''
+    CLIP_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+      NAME: HFEmbedder
+      # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_MODEL_CLS: CLIPTextModel
+      # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+      MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder/
+      # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_TOKENIZER_CLS: CLIPTokenizer
+      # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+      TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer/
+      # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+      MAX_LENGTH: 77
+      # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+      OUTPUT_KEY: pooler_output
+      # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+      D_TYPE: bfloat16
+      # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+      BATCH_INFER: True
+      CLEAN: whitespace

scepter/methods/studio/inference/dit/flux1.0_schnell_pro.yaml

@@ -0,0 +1,212 @@
+NAME: FLUX1.0_SCHNELL
+IS_DEFAULT: False
+DEFAULT_PARAS:
+  PARAS:
+    RESOLUTIONS: [[1024, 1024]]
+  INPUT:
+    IMAGE:
+    ORIGINAL_SIZE_AS_TUPLE: [1024, 1024]
+    TARGET_SIZE_AS_TUPLE: [1024, 1024]
+    PROMPT: ""
+    NEGATIVE_PROMPT:
+      DEFAULT: ""
+      VISIBLE: False
+    PROMPT_PREFIX: ""
+    SAMPLE:
+      VALUES: ["flow_eluer"]
+      DEFAULT: "flow_eluer"
+    SAMPLE_STEPS: 4
+    GUIDE_SCALE: 3.5
+    GUIDE_RESCALE:
+      DEFAULT: 0.0
+      VISIBLE: False
+    DISCRETIZATION:
+      VALUES: []
+      DEFAULT:
+      VISIBLE: False
+  OUTPUT:
+    LATENT:
+    IMAGES:
+    SEED:
+  MODULES_PARAS:
+    FIRST_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: bfloat16
+          INPUT: ["IMAGE"]
+        -
+          NAME: decode
+          DTYPE: bfloat16
+          INPUT: ["LATENT"]
+      PARAS:
+        SCALE_FACTOR: 1.5305
+        SHIFT_FACTOR: 0.0609
+        SIZE_FACTOR: 8
+    DIFFUSION_MODEL:
+      FUNCTION:
+        -
+          NAME: forward
+          DTYPE: bfloat16
+          INPUT: ["SAMPLE_STEPS", "SAMPLE", "GUIDE_SCALE"]
+    COND_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: bfloat16
+          INPUT: ["PROMPT"]
+#
+MODEL:
+  NAME: LatentDiffusionFlux
+  PARAMETERIZATION: rf
+  TIMESTEPS: 1000
+  MIN_SNR_GAMMA:
+  ZERO_TERMINAL_SNR: False
+  PRETRAINED_MODEL:
+  IGNORE_KEYS: [ ]
+  DEFAULT_N_PROMPT:
+  USE_EMA: False
+  EVAL_EMA: False
+  DIFFUSION:
+    # NAME DESCRIPTION:  TYPE:  default: 'DiffusionFluxRF'
+    NAME: DiffusionFluxRF
+    PREDICTION_TYPE: raw
+    # NOISE_SCHEDULER DESCRIPTION:  TYPE:  default: ''
+    NOISE_SCHEDULER:
+      # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchSigmaScheduler'
+      NAME: FlowMatchSigmaScheduler
+      # WEIGHTING_SCHEME DESCRIPTION: The weighting scheme for sampling timesteps, choose from ['sigma_sqrt', 'logit_normal', 'mode', 'cosmap', 'none']. TYPE: str default: 'logit_normal'
+      WEIGHTING_SCHEME: logit_normal
+      SHIFT: 3.0
+      # LOGIT_MEAN DESCRIPTION: The mean of the logit distribution for sampling timesteps. TYPE: float default: 0.0
+      LOGIT_MEAN: 0.0
+      # LOGIT_STD DESCRIPTION: The standard deviation of the logit distribution for sampling timesteps. TYPE: float default: 1.0
+      LOGIT_STD: 1.0
+      # MODE_SCALE DESCRIPTION: The scale factor for the mode of the logit distribution for sampling timesteps. TYPE: float default: 1.29
+      MODE_SCALE: 1.29
+    SAMPLER_SCHEDULER:
+      # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchFluxShiftScheduler'
+      NAME: FlowMatchFluxShiftScheduler
+      # SHIFT DESCRIPTION: Use timestamp shift or not, default is True. TYPE: bool default: True
+      SHIFT: False
+      # SIGMOID_SCALE DESCRIPTION: The scale of sigmoid function for sampling timesteps. TYPE: int default: 1
+      SIGMOID_SCALE: 1
+      # BASE_SHIFT DESCRIPTION: The base shift factor for the timestamp. TYPE: float default: 0.5
+      BASE_SHIFT: 0.5
+      # MAX_SHIFT DESCRIPTION: The max shift factor for the timestamp. TYPE: float default: 1.15
+      MAX_SHIFT: 1.15
+      #
+  DIFFUSION_MODEL:
+    # NAME DESCRIPTION:  TYPE:  default: 'Flux'
+    NAME: Flux
+    PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-schnell@flux1-schnell.safetensors
+    # IN_CHANNELS DESCRIPTION: model's input channels. TYPE: int default: 64
+    IN_CHANNELS: 64
+    # HIDDEN_SIZE DESCRIPTION: model's hidden size. TYPE: int default: 1024
+    HIDDEN_SIZE: 3072
+    # NUM_HEADS DESCRIPTION: number of heads in the transformer. TYPE: int default: 16
+    NUM_HEADS: 24
+    # AXES_DIM DESCRIPTION: dimensions of the axes of the positional encoding. TYPE: list default: [16, 56, 56]
+    AXES_DIM: [ 16, 56, 56 ]
+    # THETA DESCRIPTION: theta for positional encoding. TYPE: int default: 10000
+    THETA: 10000
+    # VEC_IN_DIM DESCRIPTION: dimension of the vector input. TYPE: int default: 768
+    VEC_IN_DIM: 768
+    # GUIDANCE_EMBED DESCRIPTION: whether to use guidance embedding. TYPE: bool default: False
+    GUIDANCE_EMBED: False
+    # CONTEXT_IN_DIM DESCRIPTION: dimension of the context input. TYPE: int default: 4096
+    CONTEXT_IN_DIM: 4096
+    # MLP_RATIO DESCRIPTION: ratio of mlp hidden size to hidden size. TYPE: float default: 4.0
+    MLP_RATIO: 4.0
+    # QKV_BIAS DESCRIPTION: whether to use bias in qkv projection. TYPE: bool default: True
+    QKV_BIAS: True
+    # DEPTH DESCRIPTION: number of transformer blocks. TYPE: int default: 19
+    DEPTH: 19
+    # DEPTH_SINGLE_BLOCKS DESCRIPTION: number of transformer blocks in the single stream block. TYPE: int default: 38
+    DEPTH_SINGLE_BLOCKS: 38
+
+  #
+  FIRST_STAGE_MODEL:
+    NAME: AutoencoderKLFlux
+    EMBED_DIM: 16
+    PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-schnell@ae.safetensors
+    IGNORE_KEYS: [ ]
+    BATCH_SIZE: 8
+    USE_CONV: False
+    SCALE_FACTOR: 0.3611
+    SHIFT_FACTOR: 0.1159
+    #
+    ENCODER:
+      NAME: Encoder
+      USE_CHECKPOINT: True
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 16
+      DOUBLE_Z: True
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+    #
+    DECODER:
+      NAME: Decoder
+      USE_CHECKPOINT: True
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 16
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+      GIVE_PRE_END: False
+      TANH_OUT: False
+  #
+  COND_STAGE_MODEL:
+    # NAME DESCRIPTION:  TYPE:  default: 'T5PlusClipFluxEmbedder'
+    NAME: T5PlusClipFluxEmbedder
+    # T5_MODEL DESCRIPTION:  TYPE:  default: ''
+    T5_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+      NAME: HFEmbedder
+      # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_MODEL_CLS: T5EncoderModel
+      # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+      MODEL_PATH: ms://AI-ModelScope/FLUX.1-schnell@text_encoder_2/
+      # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_TOKENIZER_CLS: T5Tokenizer
+      # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+      TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-schnell@tokenizer_2/
+      # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+      MAX_LENGTH: 256
+      # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+      OUTPUT_KEY: last_hidden_state
+      # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+      D_TYPE: bfloat16
+      # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+      BATCH_INFER: False
+      CLEAN: whitespace
+    # CLIP_MODEL DESCRIPTION:  TYPE:  default: ''
+    CLIP_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+      NAME: HFEmbedder
+      # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_MODEL_CLS: CLIPTextModel
+      # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+      MODEL_PATH: ms://AI-ModelScope/FLUX.1-schnell@text_encoder/
+      # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+      HF_TOKENIZER_CLS: CLIPTokenizer
+      # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+      TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-schnell@tokenizer/
+      # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+      MAX_LENGTH: 77
+      # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+      OUTPUT_KEY: pooler_output
+      # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+      D_TYPE: bfloat16
+      # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+      BATCH_INFER: True
+      CLEAN: whitespace

scepter/methods/studio/inference/dit/pixart_alpha_pro.yaml

@@ -0,0 +1,120 @@
+NAME: PIXART_ALPHA
+IS_DEFAULT: False
+DEFAULT_PARAS:
+  PARAS:
+    RESOLUTIONS: [[1024, 1024]]
+  INPUT:
+    IMAGE:
+    ORIGINAL_SIZE_AS_TUPLE: [1024, 1024]
+    TARGET_SIZE_AS_TUPLE: [1024, 1024]
+    PROMPT: ""
+    NEGATIVE_PROMPT: ""
+    PROMPT_PREFIX: ""
+    SAMPLE: ddim
+    SAMPLE_STEPS: 20
+    GUIDE_SCALE: 4.5
+    GUIDE_RESCALE: 0.5
+    DISCRETIZATION: trailing
+  OUTPUT:
+    LATENT:
+    IMAGES:
+    SEED:
+  MODULES_PARAS:
+    FIRST_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: float32
+          INPUT: ["IMAGE"]
+        -
+          NAME: decode
+          DTYPE: float32
+          INPUT: ["LATENT"]
+      PARAS:
+        SCALE_FACTOR: 0.18215
+        SIZE_FACTOR: 8
+    DIFFUSION_MODEL:
+      FUNCTION:
+        -
+          NAME: forward
+          DTYPE: float16
+          INPUT: ["SAMPLE_STEPS", "SAMPLE", "GUIDE_SCALE", "GUIDE_RESCALE", "DISCRETIZATION"]
+    COND_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: float32
+          INPUT: ["PROMPT"]
+#
+MODEL:
+  PRETRAINED_MODEL:
+  DECODER_BIAS: 0.5
+  SCHEDULE:
+    PARAMETERIZATION: "eps"
+    TIMESTEPS: 1000
+    ZERO_TERMINAL_SNR: False
+    SCHEDULE_ARGS:
+      "NAME": "linear"
+      "BETA_MIN": 0.0001
+      "BETA_MAX": 0.02
+  #
+  DIFFUSION_MODEL:
+    NAME: PixArt
+    PRETRAINED_MODEL: ms://AI-ModelScope/PixArt-alpha@PixArt-XL-2-1024-MS.pth
+    INPUT_SIZE: 128
+    PATCH_SIZE: 2
+    IN_CHANNELS: 4
+    HIDDEN_SIZE: 1152
+    DEPTH: 28
+    NUM_HEADS: 16
+    MLP_RATIO: 4.0
+    CLASS_DROPOUT_PROB: 0.1
+    PRED_SIGMA: True
+    DROP_PATH: 0.0
+    WINDOW_DIZE: 0
+    USE_REL_POS: False
+    CAPTION_CHANNELS: 4096
+    LEWEI_SCALE: 2
+    MODEL_MAX_LENGTH: 120
+  #
+  FIRST_STAGE_MODEL:
+    NAME: AutoencoderKL
+    PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-2-base@512-base-ema.safetensors
+    EMBED_DIM: 4
+    IGNORE_KEYS: [ ]
+    BATCH_SIZE: 1
+    #
+    ENCODER:
+      NAME: Encoder
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 4
+      DOUBLE_Z: True
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+    #
+    DECODER:
+      NAME: Decoder
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 4
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+      GIVE_PRE_END: False
+      TANH_OUT: False
+  #
+  COND_STAGE_MODEL:
+    NAME: T5EmbedderHF
+    PRETRAINED_MODEL: ms://AI-ModelScope/PixArt-alpha@t5-v1_1-xxl/
+    TOKENIZER_PATH: ms://AI-ModelScope/PixArt-alpha@t5-v1_1-xxl/
+    LENGTH: 120
+    CLEAN: heavy
+    USE_GRAD: False

scepter/methods/studio/inference/dit/sd3_pro.yaml

@@ -0,0 +1,148 @@
+NAME: SD3
+IS_DEFAULT: False
+DEFAULT_PARAS:
+  PARAS:
+    RESOLUTIONS: [[1024, 1024]]
+  INPUT:
+    IMAGE:
+    ORIGINAL_SIZE_AS_TUPLE: [1024, 1024]
+    TARGET_SIZE_AS_TUPLE: [1024, 1024]
+    PROMPT: ""
+    NEGATIVE_PROMPT: ""
+    PROMPT_PREFIX: ""
+    SAMPLE: euler
+    SAMPLE_STEPS: 28
+    GUIDE_SCALE: 5.0
+    GUIDE_RESCALE: 0.0
+    DISCRETIZATION: trailing
+  OUTPUT:
+    LATENT:
+    IMAGES:
+    SEED:
+  MODULES_PARAS:
+    FIRST_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: float32
+          INPUT: ["IMAGE"]
+        -
+          NAME: decode
+          DTYPE: float32
+          INPUT: ["LATENT"]
+      PARAS:
+        SCALE_FACTOR: 1.5305
+        SHIFT_FACTOR: 0.0609
+        SIZE_FACTOR: 8
+    DIFFUSION_MODEL:
+      FUNCTION:
+        -
+          NAME: forward
+          DTYPE: float16
+          INPUT: ["SAMPLE_STEPS", "SAMPLE", "GUIDE_SCALE", "GUIDE_RESCALE", "DISCRETIZATION"]
+    COND_STAGE_MODEL:
+      FUNCTION:
+        -
+          NAME: encode
+          DTYPE: float32
+          INPUT: ["PROMPT"]
+#
+MODEL:
+  PRETRAINED_MODEL:
+  SCHEDULE:
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    SCALE_FACTOR: 1.5305
+    SHIFT_FACTOR: 0.0609
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "shifted"
+      "SHIFT": 3
+    T_WEIGHT: uniform
+  #
+  DIFFUSION_MODEL:
+    NAME: MMDiT
+    PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+    IGNORE_KEYS: '^first_stage_model.'
+    IN_CHANNELS: 16
+    PATCH_SIZE: 2
+    OUT_CHANNELS: 16
+    DEPTH: 24
+    INPUT_SIZE:
+    ADM_IN_CHANNELS: 2048
+    CONTEXT_EMBEDDER_CONFIG: { 'target': 'torch.nn.Linear', 'params': { 'in_features': 4096, 'out_features': 1536 } }
+    NUM_PATCHES: 36864
+    POS_EMBED_MAX_SIZE: 192
+    POS_EMBED_SCALING_FACTOR:
+    USE_CHECKPOINT: True
+  #
+  FIRST_STAGE_MODEL:
+    NAME: AutoencoderKL
+    PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+    EMBED_DIM: 16
+    IGNORE_KEYS: '^model.diffusion_model.'
+    BATCH_SIZE: 1
+    USE_CONV: False
+    #
+    ENCODER:
+      NAME: Encoder
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 16
+      DOUBLE_Z: True
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+    #
+    DECODER:
+      NAME: Decoder
+      CH: 128
+      OUT_CH: 3
+      NUM_RES_BLOCKS: 2
+      IN_CHANNELS: 3
+      ATTN_RESOLUTIONS: [ ]
+      CH_MULT: [ 1, 2, 4, 4 ]
+      Z_CHANNELS: 16
+      DROPOUT: 0.0
+      RESAMP_WITH_CONV: True
+      GIVE_PRE_END: False
+      TANH_OUT: False
+  #
+  COND_STAGE_MODEL:
+    NAME: SD3TextEmbedder
+    P_ZERO: 0.0
+    CLIP_L:
+      NAME: FrozenCLIPEmbedder2
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder
+      TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer
+      MAX_LENGTH: 77
+      FREEZE: True
+      LAYER: penultimate
+      RETURN_POOLED: True
+      USE_FINAL_LAYER_NORM: False
+      IS_TRAINABLE: False
+    CLIP_G:
+      NAME: FrozenCLIPEmbedder2
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_2
+      TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_2
+      MAX_LENGTH: 77
+      FREEZE: True
+      LAYER: penultimate
+      RETURN_POOLED: True
+      USE_FINAL_LAYER_NORM: False
+      IS_TRAINABLE: False
+    T5_XXL:
+      NAME: T5EmbedderHF
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_3
+      TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_3
+      LENGTH: 256
+      CLEAN: whitespace
+      USE_GRAD: False
+      T5_DTYPE: float16

scepter/methods/studio/inference/edit/stylebooth_tb_pro.yaml

@@ -2,7 +2,7 @@ NAME: EDIT
 IS_DEFAULT: False
 DEFAULT_PARAS:
   PARAS:
-    RESOLUTIONS: [[1024, 1024]]
+    RESOLUTIONS: [[512, 512], [1024, 1024]]
   INPUT:
     IMAGE:
     PROMPT: ""
@@ -49,7 +49,7 @@ DEFAULT_PARAS:
           INPUT: ["PROMPT", "NEGATIVE_PROMPT"]
 
 MODEL:
-  PRETRAINED_MODEL: ms://damo/stylebooth@models/stylebooth-tb-5000-0.bin
+  PRETRAINED_MODEL: ms://iic/stylebooth@models/stylebooth-tb-5000-0.bin
   SCHEDULE:
     PARAMETERIZATION: "eps"
     TIMESTEPS: 1000

scepter/methods/studio/inference/inference.yaml

@@ -6,64 +6,81 @@ DIFFUSION_PARAS:
              'dpm2_karras', 'dpm2_ancestral_karras', 'dpmpp_2s_ancestral_karras', 'dpmpp_2m_karras',
              'dpmpp_sde_karras', 'dpmpp_2m_sde_karras']
     DEFAULT: 'dpmpp_2s_ancestral'
+    VISIBLE: True
   NEGATIVE_PROMPT:
     DEFAULT:
+    VISIBLE: True
   PROMPT_PREFIX:
     DEFAULT:
+    VISIBLE: True
   SAMPLES:
     MIN: 1
     MAX: 4
     DEFAULT: 1
+    VISIBLE: True
   SAMPLE_STEPS:
     MIN: 1
     MAX: 100
     DEFAULT: 30
+    VISIBLE: True
   GUIDE_SCALE:
     MIN: 0
     MAX: 10
     DEFAULT: 5.0
+    VISIBLE: True
   GUIDE_RESCALE:
     MIN: 0
     MAX: 1.0
     DEFAULT: 0.5
+    VISIBLE: True
   DISCRETIZATION:
     VALUES: ["trailing", "leading", "linspace"]
     DEFAULT: "linspace"
+    VISIBLE: True
   REFINE_SAMPLERS:
     VALUES: [ 'ddim', 'euler', 'euler_ancestral', 'heun', 'dpm2',
               'dpm2_ancestral', 'dpmpp_2m', 'dpmpp_sde', 'dpmpp_2m_sde', 'dpmpp_2s_ancestral',
               'dpm2_karras', 'dpm2_ancestral_karras', 'dpmpp_2s_ancestral_karras', 'dpmpp_2m_karras',
               'dpmpp_sde_karras', 'dpmpp_2m_sde_karras' ]
     DEFAULT: 'dpmpp_2s_ancestral'
+    VISIBLE: True
   REFINE_SAMPLE_STEPS:
     MIN: 0
     MAX: 100
     DEFAULT: 30
+    VISIBLE: True
   REFINE_GUIDE_SCALE:
     MIN: 0
     MAX: 10
     DEFAULT: 5.0
+    VISIBLE: True
   REFINE_GUIDE_RESCALE:
     MIN: 0
     MAX: 1.0
     DEFAULT: 0.5
+    VISIBLE: True
   REFINE_DISCRETIZATION:
     VALUES: [ "trailing", "leading", "linspace" ]
     DEFAULT: "linspace"
+    VISIBLE: True
   AESTHETIC_SCORE:
     MIN: 0.0
     MAX: 10.0
     DEFAULT: 6.0
+    VISIBLE: True
   NEGATIVE_AESTHETIC_SCORE:
     MIN: 0.0
     MAX: 10.0
     DEFAULT: 2.5
+    VISIBLE: True
   REFINE_STRENGTH:
     MIN: 0
     MAX: 1.0
     DEFAULT: 0.15
   RESOLUTIONS:
-    VALUES: [[704, 1408], [704, 1344], [768, 1344],
+    VALUES: [
+             [512, 512], [768, 768],
+             [704, 1408], [704, 1344], [768, 1344],
              [720, 1280],
              [768, 1280], [832, 1216], [832, 1152],
              [896, 1152], [896, 1088], [960, 1088],
@@ -73,8 +90,13 @@ DIFFUSION_PARAS:
              [1280, 768],
              [1344, 768], [1344, 704], [1408, 704],
              [1472, 704], [1536, 640], [1600, 640],
-             [1664, 576], [1728, 576]]
+             [1664, 576], [1728, 576],
+             [2048, 2048], [2048, 1920], [1920, 2048],
+             [1536, 2560], [2560, 1536], [2560, 1440],
+             [2560, 1440]
+            ]
     DEFAULT: [1024, 1024]
+    VISIBLE: True
 EXTENSION_PARAS:
   MANTRA_BOOK: scepter/methods/studio/extensions/mantra_book/mantra_book.yaml
   OFFICIAL_TUNERS: scepter/methods/studio/extensions/tuners/official_tuners.yaml
@@ -87,18 +109,18 @@ CONTROLABLE_ANNOTATORS:
     IS_DEFAULT: True
   -
     NAME: "HedAnnotator"
-    PRETRAINED_MODEL: "ms://damo/scepter_scedit@annotator/ckpts/ControlNetHED.pth"
+    PRETRAINED_MODEL: "ms://iic/scepter_scedit@annotator/ckpts/ControlNetHED.pth"
     TYPE: Hed
     IS_DEFAULT: False
   -
     NAME: "OpenposeAnnotator"
-    BODY_MODEL_PATH: "ms://damo/scepter_scedit@annotator/ckpts/body_pose_model.pth"
-    HAND_MODEL_PATH: "ms://damo/scepter_scedit@annotator/ckpts/hand_pose_model.pth"
+    BODY_MODEL_PATH: "ms://iic/scepter_scedit@annotator/ckpts/body_pose_model.pth"
+    HAND_MODEL_PATH: "ms://iic/scepter_scedit@annotator/ckpts/hand_pose_model.pth"
     TYPE: Openpose
     IS_DEFAULT: False
   -
     NAME: "MidasDetector"
-    PRETRAINED_MODEL: "ms://damo/scepter_scedit@annotator/ckpts/dpt_hybrid-midas-501f0c75.pt"
+    PRETRAINED_MODEL: "ms://iic/scepter_scedit@annotator/ckpts/dpt_hybrid-midas-501f0c75.pt"
     TYPE: Midas
     IS_DEFAULT: False
   -

scepter/methods/studio/inference/largen/largen_pro.yaml

@@ -68,7 +68,7 @@ DEFAULT_PARAS:
           INPUT: ["ORIGINAL_SIZE_AS_TUPLE", "AESTHETIC_SCORE", "NEGATIVE_AESTHETIC_SCORE", "CROP_COORDS_TOP_LEFT", "PROMPT", "NEGATIVE_PROMPT"]
 
 MODEL:
-  PRETRAINED_MODEL: ms://damo/LARGEN@models/largen_ckpt_s22k.pth
+  PRETRAINED_MODEL: ms://iic/LARGEN@models/largen_ckpt_s22k.pth
   # SCHEDULE_ARGS DESCRIPTION:  TYPE:  default: ''
   SCHEDULE:
     PARAMETERIZATION: "eps"
@@ -245,8 +245,8 @@ MODEL:
         LEGACY_UCG_VALUE:
       -
         NAME: IPAdapterPlusEmbedder
-        CLIP_DIR: ms://damo/LARGEN@models/clip_encoder/
-        PRETRAINED_MODEL: ms://damo/LARGEN@models/ip-adapter-plus_sdxl_vit-h.bin
+        CLIP_DIR: ms://iic/LARGEN@models/clip_encoder/
+        PRETRAINED_MODEL: ms://iic/LARGEN@models/ip-adapter-plus_sdxl_vit-h.bin
         INPUT_KEYS: [ "ref_ip", "ref_detail" ]
         IN_DIM: 1280
         HEADS: 20

scepter/methods/studio/preprocess/preprocess.yaml

@@ -2,11 +2,10 @@ WORK_DIR: datasets
 EXPORT_DIR: export_datasets
 FILE_SYSTEM:
   -
-    # NAME DESCRIPTION:  TYPE:  default: ''
     NAME: LocalFs
     AUTO_CLEAN: False
-
 PROCESSORS:
+  # Caption processor
   - NAME: BlipImageBase
     TYPE: caption
     MODEL_PATH: ms://cubeai/blip-image-captioning-base
@@ -15,6 +14,81 @@ PROCESSORS:
     PARAS:
       - LANGUAGE_NAME: English
         LANGUAGE_ZH_NAME: 英语
+  - NAME: InternVL15
+    TYPE: caption
+    MODEL_PATH: ms://AI-ModelScope/InternVL-Chat-V1-5
+    DEVICE: "gpu"
+    MEMORY: 49968
+    PARAS:
+      - PROMPT: 用中文描述这张图片
+        LANGUAGE_NAME: Chinese
+        LANGUAGE_ZH_NAME: 中文
+      - PROMPT: Generate the caption in English
+        LANGUAGE_NAME: English
+        LANGUAGE_ZH_NAME: 英语
+  - NAME: QWVLQuantize
+    TYPE: caption
+    DEVICE: "gpu"
+    MEMORY: 7885
+    MODEL_PATH: ms://qwen/Qwen-VL:v1.0.3
+    PARAS:
+      - PROMPT: 用中文描述这张图片
+        LANGUAGE_NAME: Chinese
+        LANGUAGE_ZH_NAME: 中文
+        MAX_NEW_TOKENS:
+          VALUE: 1024
+          MAX: 2048
+          STEP: 128
+          MIN: 256
+        MIN_NEW_TOKENS:
+          VALUE: 16
+          MAX: 1024
+          STEP: 16
+          MIN: 0
+        NUM_BEAMS:
+          VALUE: 1
+          MAX: 12
+          STEP: 1
+          MIN: 1
+        REPETITION_PENALTY:
+          VALUE: 1.0
+          MAX: 100.0
+          STEP: 1.0
+          MIN: 1.0
+        TEMPERATURE:
+          VALUE: 1.0
+          MAX: 100.0
+          STEP: 1.0
+          MIN: 1.0
+      - PROMPT: Generate the caption in English
+        LANGUAGE_NAME: English
+        LANGUAGE_ZH_NAME: 英语
+        MAX_NEW_TOKENS:
+          VALUE: 1024
+          MAX: 2048
+          STEP: 128
+          MIN: 256
+        MIN_NEW_TOKENS:
+          VALUE: 16
+          MAX: 1024
+          STEP: 16
+          MIN: 0
+        NUM_BEAMS:
+          VALUE: 1
+          MAX: 12
+          STEP: 1
+          MIN: 1
+        REPETITION_PENALTY:
+          VALUE: 1.0
+          MAX: 100.0
+          STEP: 1.0
+          MIN: 1.0
+        TEMPERATURE:
+          VALUE: 1.0
+          MAX: 100.0
+          STEP: 1.0
+          MIN: 1.0
+
   - NAME: QWVL
     TYPE: caption
     MODEL_PATH: ms://qwen/Qwen-VL:v1.0.3
@@ -77,75 +151,14 @@ PROCESSORS:
           MAX: 100.0
           STEP: 1.0
           MIN: 1.0
-  -
-    NAME: QWVLQuantize
-    TYPE: caption
-    DEVICE: "gpu"
-    MEMORY: 7885
-    MODEL_PATH: ms://qwen/Qwen-VL:v1.0.3
-    PARAS:
-      - PROMPT: 用中文描述这张图片
-        LANGUAGE_NAME: Chinese
-        LANGUAGE_ZH_NAME: 中文
-        MAX_NEW_TOKENS:
-          VALUE: 1024
-          MAX: 2048
-          STEP: 128
-          MIN: 256
-        MIN_NEW_TOKENS:
-          VALUE: 16
-          MAX: 1024
-          STEP: 16
-          MIN: 0
-        NUM_BEAMS:
-          VALUE: 1
-          MAX: 12
-          STEP: 1
-          MIN: 1
-        REPETITION_PENALTY:
-          VALUE: 1.0
-          MAX: 100.0
-          STEP: 1.0
-          MIN: 1.0
-        TEMPERATURE:
-          VALUE: 1.0
-          MAX: 100.0
-          STEP: 1.0
-          MIN: 1.0
-      - PROMPT: Generate the caption in English
-        LANGUAGE_NAME: English
-        LANGUAGE_ZH_NAME: 英语
-        MAX_NEW_TOKENS:
-          VALUE: 1024
-          MAX: 2048
-          STEP: 128
-          MIN: 256
-        MIN_NEW_TOKENS:
-          VALUE: 16
-          MAX: 1024
-          STEP: 16
-          MIN: 0
-        NUM_BEAMS:
-          VALUE: 1
-          MAX: 12
-          STEP: 1
-          MIN: 1
-        REPETITION_PENALTY:
-          VALUE: 1.0
-          MAX: 100.0
-          STEP: 1.0
-          MIN: 1.0
-        TEMPERATURE:
-          VALUE: 1.0
-          MAX: 100.0
-          STEP: 1.0
-          MIN: 1.0
-  -
-    NAME: CenterCrop
+
+  # Simple processor
+  - NAME: CenterCrop
     TYPE: image
     DEVICE: "cpu"
     MEMORY: 10
     PARAS:
+      CAPTION_INTERACTIVE: False
       HEIGHT_RATIO:
         VALUE: 1
         MAX: 20
@@ -156,18 +169,284 @@ PROCESSORS:
         MAX: 20
         STEP: 1
         MIN: 1
-#  - NAME: PaddingCrop
-#    TYPE: image
-#    DEVICE: "cpu"
-#    MEMORY: 10
-#    PARAS:
-#      HEIGHT_RATIO:
-#        VALUE: 3
-#        MAX: 25
-#        STEP: 1
-#        MIN: 1
-#      WIDTH_RATIO:
-#        VALUE: 4
-#        MAX: 20
-#        STEP: 1
-#        MIN: 1
+  - NAME: ChangeSample
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      CAPTION_INTERACTIVE: False
+      SRC_IMAGE_INTERACTIVE: True
+      SRC_IMAGE_MASK_INTERACTIVE: True
+      TARGET_IMAGE_INTERACTIVE: True
+      PREVIEW_BTN_VISIBLE: False
+  - NAME: MaskEditSample
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      SRC_IMAGE_INTERACTIVE: True
+      SRC_IMAGE_TOOL: sketch
+      CAPTION_INTERACTIVE: False
+  - NAME: SourceMaskSample
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      SRC_IMAGE_INTERACTIVE: True
+      SRC_IMAGE_TOOL: sketch
+      CAPTION_INTERACTIVE: False
+  - NAME: MaskSwapEditSample
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      PREVIEW_BTN_VISIBLE: False
+      SRC_IMAGE_INTERACTIVE: True
+      SRC_IMAGE_TOOL: sketch
+      CAPTION_INTERACTIVE: False
+  - NAME: SwapMaskSwapEditSample
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      PREVIEW_BTN_VISIBLE: False
+      SRC_IMAGE_INTERACTIVE: True
+      SRC_IMAGE_TOOL: sketch
+      CAPTION_INTERACTIVE: False
+  - NAME: SwapSample
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      PREVIEW_BTN_VISIBLE: True
+      SRC_IMAGE_INTERACTIVE: True
+      TARGET_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: DefaultMaskSample
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      PREVIEW_BTN_VISIBLE: True
+      SRC_IMAGE_INTERACTIVE: False
+      TARGET_IMAGE_INTERACTIVE: False
+      CAPTION_INTERACTIVE: False
+  #  - NAME: PaddingCrop
+  #    TYPE: image
+  #    DEVICE: "cpu"
+  #    MEMORY: 10
+  #    PARAS:
+  #      HEIGHT_RATIO:
+  #        VALUE: 3
+  #        MAX: 25
+  #        STEP: 1
+  #        MIN: 1
+  #      WIDTH_RATIO:
+  #        VALUE: 4
+  #        MAX: 20
+  #        STEP: 1
+  #        MIN: 1
+  # Annotator processor
+  - NAME: CannyExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "CannyAnnotator"
+      LOW_THRESHOLD: 100
+      HIGH_THRESHOLD: 200
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: ColorExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "ColorAnnotator"
+      RATIO: 64
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: InfoDrawContourExtractor
+    TYPE: image
+    DEVICE: "gpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "InfoDrawContourAnnotator"
+      INPUT_NC: 3
+      OUTPUT_NC: 1
+      N_RESIDUAL_BLOCKS: 3
+      SIGMOID: True
+      PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/informative_drawing_contour_style.pth"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: DegradationExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "DegradationAnnotator"
+      RANDOM_DEGRADATION: True
+      PARAMS:
+        gaussian_noise: { }
+        resize: { 'scale': [ 0.4, 0.8 ] }
+        jpeg: { 'jpeg_level': [ 25, 75 ] }
+        gaussian_blur: { 'kernel_size': [ 7, 9, 11, 13, 15 ], 'sigma': [ 0.9, 1.8 ] }
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: MidasExtractor
+    TYPE: image
+    DEVICE: "gpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "MidasDetector"
+      PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/dpt_hybrid-midas-501f0c75.pt"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: DoodleExtractor
+    TYPE: image
+    DEVICE: "gpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "DoodleAnnotator"
+      PROCESSOR_TYPE: "pidinet_sketch"
+      PROCESSOR_CFG:
+        - NAME: "PiDiAnnotator"
+          PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/table5_pidinet.pth"
+        - NAME: "SketchAnnotator"
+          PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/sketch_simplification_gan.pth"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: GrayExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "GrayAnnotator"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: InpaintingExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "InpaintingAnnotator"
+      RETURN_MASK: False
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+
+  - NAME: InpaintingSourceExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: OpenposeExtractor
+    TYPE: image
+    DEVICE: "gpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "OpenposeAnnotator"
+      BODY_MODEL_PATH: "ms://iic/scepter_annotator@annotator/ckpts/body_pose_model.pth"
+      HAND_MODEL_PATH: "ms://iic/scepter_annotator@annotator/ckpts/hand_pose_model.pth"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: OutpaintingExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "OutpaintingAnnotator"
+      RETURN_MASK: False
+      KEEP_PADDING_RATIO: 1
+      RANDOM_CFG:
+        DIRECTION_RANGE: [ 'left', 'right', 'up', 'down' ]
+        RATIO_RANGE: [ 0.1, 0.7 ]
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+      USE_MASK_VISIBLE: True
+      PREVIEW_BTN_VISIBLE: False
+  - NAME: OutpaintingResize
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "OutpaintingResize"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+      PREVIEW_BTN_VISIBLE: False
+  - NAME: InfoDrawAnimeAnnotator
+    TYPE: image
+    DEVICE: "gpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "InfoDrawAnimeAnnotator"
+      INPUT_NC: 3
+      OUTPUT_NC: 1
+      N_RESIDUAL_BLOCKS: 3
+      SIGMOID: True
+      PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/informative_drawing_anime_style.pth"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: ESAMExtractor
+    TYPE: image
+    DEVICE: "gpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "ESAMAnnotator"
+      PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/efficient_sam_vits.pt"
+      SAVE_MODE: 'P'
+      GRID_SIZE: 32
+      USE_DOMINANT_COLOR: True
+      RETURN_MASK: False
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+  - NAME: InvertExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "InvertAnnotator"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False
+
+  - NAME: LamaExtractor
+    TYPE: image
+    DEVICE: "cpu"
+    MEMORY: 10
+    MODEL:
+      NAME: "LamaAnnotator"
+      PRETRAINED_MODEL: "ms:///iic/cv_fft_inpainting_lama/"
+    PARAS:
+      SRC_IMAGE_TOOL: sketch
+      SRC_IMAGE_INTERACTIVE: True
+      CAPTION_INTERACTIVE: False

scepter/methods/studio/scepter_ui.yaml

@@ -48,11 +48,11 @@ BANNER: |
       </div>
       <div class="qr-codes">
           <div class="qr-code-container">
-              <img src="https://modelscope.cn/api/v1/models/damo/scepter/repo?Revision=master&FilePath=assets/scepter_studio/ms_scepter_studio_qr.png" alt="ms_scepter_studio_qr">
+              <img src="https://modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets/scepter_studio/ms_scepter_studio_qr.png" alt="ms_scepter_studio_qr">
               <div class="caption"><a href="https://www.modelscope.cn/studios/iic/scepter_studio">Modelscope Studio</a></div>
           </div>
           <div class="qr-code-container">
-              <img src="https://modelscope.cn/api/v1/models/damo/scepter/repo?Revision=master&FilePath=assets/scepter_studio/scepter_github_qr.png" alt="scepter_github_qr">
+              <img src="https://modelscope.cn/api/v1/models/iic/scepter/repo?Revision=master&FilePath=assets/scepter_studio/scepter_github_qr.png" alt="scepter_github_qr">
               <div class="caption"><a href="https://github.com/modelscope/scepter">Github</a></div>
           </div>
       </div>
@@ -87,3 +87,7 @@ INTERFACE:
     NAME_EN: Inference
     IFID: inference
     CONFIG: scepter/methods/studio/inference/inference.yaml
+  - NAME: 对话式编辑
+    NAME_EN: ChatBot
+    IFID: ChatBot
+    CONFIG: scepter/methods/studio/chatbot/chatbot.yaml

scepter/methods/studio/self_train/dit/flux1.0_dv_pro.yaml

@@ -0,0 +1,343 @@
+ENV:
+  BACKEND: nccl
+META:
+  VERSION: 'FLUX1.0_DEV'
+  DESCRIPTION: "flux 1.0 dev"
+  IS_DEFAULT: False
+  IS_SHARE: True
+  INFERENCE_PARAS:
+    INFERENCE_BATCH_SIZE: 1
+    INFERENCE_PREFIX: ""
+    DEFAULT_SAMPLER: "flow_euler"
+    DEFAULT_SAMPLE_STEPS: 50
+    INFERENCE_N_PROMPT: ""
+    RESOLUTION: [1024, 1024]
+  PARAS:
+    -
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 89000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 4e-4
+      IS_DEFAULT: False
+      TUNER: FULL
+    -
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 89000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 4e-4
+      IS_DEFAULT: True
+      TUNER: LORA
+  #
+  TUNERS:
+    LORA:
+      -
+        NAME: SwiftLoRA
+        R: 4
+        LORA_ALPHA: 4
+        LORA_DROPOUT: 0.0
+        BIAS: "none"
+        TARGET_MODULES: "(model.double_blocks.*(.qkv|.proj|.img_mod.lin|.txt_mod.lin))|(model.single_blocks.*(.linear1|.linear2|.modulation.lin))$"
+#
+SOLVER:
+  NAME: LatentDiffusionSolver
+  # MAX_STEPS DESCRIPTION: The total steps for training. TYPE: int default: 100000
+  MAX_STEPS: 100000
+  # USE_AMP DESCRIPTION: Use amp to surpport mix precision or not, default is False. TYPE: bool default: False
+  USE_AMP: True
+  # DTYPE DESCRIPTION: The precision for training. TYPE: str default: 'float32'
+  DTYPE: bfloat16
+  # USE_FAIRSCALE DESCRIPTION: Use fairscale as the backend of ddp, default False. TYPE: bool default: False
+  USE_FAIRSCALE: False
+  # USE_FSDP DESCRIPTION: Use fsdp as the backend of ddp, default False. TYPE: bool default: False
+  USE_FSDP: True
+  # LOAD_MODEL_ONLY DESCRIPTION: Only load the model rather than the optimizer and schedule, default is False. TYPE: bool default: False
+  LOAD_MODEL_ONLY: False
+  # RESUME_FROM DESCRIPTION: Resume from some state of training! TYPE: str default: ''
+  RESUME_FROM:
+  WORK_DIR: ./cache/save_data/dit_flux_dev_1024_lora
+  LOG_FILE: std_log.txt
+  # EVAL_INTERVAL DESCRIPTION: Eval the model interval. TYPE: int default: 1
+  EVAL_INTERVAL: 100
+  # LOG_TRAIN_NUM DESCRIPTION: The number samples used to log in training phase. TYPE: int default: -1
+  LOG_TRAIN_NUM: 16
+  # FSDP_REDUCE_DTYPE DESCRIPTION: The dtype of reduce in FSDP. TYPE: str default: 'float16'
+  FSDP_REDUCE_DTYPE: float32
+  # FSDP_BUFFER_DTYPE DESCRIPTION: The dtype of buffer in FSDP. TYPE: str default: 'float16'
+  FSDP_BUFFER_DTYPE: float32
+  # FSDP_SHARD_MODULES DESCRIPTION: The modules to be sharded in FSDP. TYPE: list default: ['model']
+  FSDP_SHARD_MODULES: [ 'model', 'cond_stage_model.t5_model' ] #
+  SAVE_MODULES: [ 'model'] #
+  TRAIN_MODULES: ['model']
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  #
+  TUNER:
+  #
+  MODEL:
+    NAME: LatentDiffusionFlux
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    DEFAULT_N_PROMPT:
+    USE_EMA: False
+    EVAL_EMA: False
+    DIFFUSION:
+      # NAME DESCRIPTION:  TYPE:  default: 'DiffusionFluxRF'
+      NAME: DiffusionFluxRF
+      PREDICTION_TYPE: raw
+      # NOISE_SCHEDULER DESCRIPTION:  TYPE:  default: ''
+      NOISE_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchSigmaScheduler'
+        NAME: FlowMatchSigmaScheduler
+        # WEIGHTING_SCHEME DESCRIPTION: The weighting scheme for sampling timesteps, choose from ['sigma_sqrt', 'logit_normal', 'mode', 'cosmap', 'none']. TYPE: str default: 'logit_normal'
+        WEIGHTING_SCHEME: logit_normal
+        SHIFT: 3.0
+        # LOGIT_MEAN DESCRIPTION: The mean of the logit distribution for sampling timesteps. TYPE: float default: 0.0
+        LOGIT_MEAN: 0.0
+        # LOGIT_STD DESCRIPTION: The standard deviation of the logit distribution for sampling timesteps. TYPE: float default: 1.0
+        LOGIT_STD: 1.0
+        # MODE_SCALE DESCRIPTION: The scale factor for the mode of the logit distribution for sampling timesteps. TYPE: float default: 1.29
+        MODE_SCALE: 1.29
+      SAMPLER_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchFluxShiftScheduler'
+        NAME: FlowMatchFluxShiftScheduler
+        # SHIFT DESCRIPTION: Use timestamp shift or not, default is True. TYPE: bool default: True
+        SHIFT: False
+        # SIGMOID_SCALE DESCRIPTION: The scale of sigmoid function for sampling timesteps. TYPE: int default: 1
+        SIGMOID_SCALE: 1
+        # BASE_SHIFT DESCRIPTION: The base shift factor for the timestamp. TYPE: float default: 0.5
+        BASE_SHIFT: 0.5
+        # MAX_SHIFT DESCRIPTION: The max shift factor for the timestamp. TYPE: float default: 1.15
+        MAX_SHIFT: 1.15
+        #
+    DIFFUSION_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'Flux'
+      NAME: Flux
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@flux1-dev.safetensors
+      # IN_CHANNELS DESCRIPTION: model's input channels. TYPE: int default: 64
+      IN_CHANNELS: 64
+      # HIDDEN_SIZE DESCRIPTION: model's hidden size. TYPE: int default: 1024
+      HIDDEN_SIZE: 3072
+      # NUM_HEADS DESCRIPTION: number of heads in the transformer. TYPE: int default: 16
+      NUM_HEADS: 24
+      # AXES_DIM DESCRIPTION: dimensions of the axes of the positional encoding. TYPE: list default: [16, 56, 56]
+      AXES_DIM: [ 16, 56, 56 ]
+      # THETA DESCRIPTION: theta for positional encoding. TYPE: int default: 10000
+      THETA: 10000
+      # VEC_IN_DIM DESCRIPTION: dimension of the vector input. TYPE: int default: 768
+      VEC_IN_DIM: 768
+      # GUIDANCE_EMBED DESCRIPTION: whether to use guidance embedding. TYPE: bool default: False
+      GUIDANCE_EMBED: False
+      # CONTEXT_IN_DIM DESCRIPTION: dimension of the context input. TYPE: int default: 4096
+      CONTEXT_IN_DIM: 4096
+      # MLP_RATIO DESCRIPTION: ratio of mlp hidden size to hidden size. TYPE: float default: 4.0
+      MLP_RATIO: 4.0
+      # QKV_BIAS DESCRIPTION: whether to use bias in qkv projection. TYPE: bool default: True
+      QKV_BIAS: True
+      # DEPTH DESCRIPTION: number of transformer blocks. TYPE: int default: 19
+      DEPTH: 19
+      # DEPTH_SINGLE_BLOCKS DESCRIPTION: number of transformer blocks in the single stream block. TYPE: int default: 38
+      DEPTH_SINGLE_BLOCKS: 38
+      USE_GRAD_CHECKPOINT: True
+
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKLFlux
+      EMBED_DIM: 16
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@ae.safetensors
+      IGNORE_KEYS: [ ]
+      BATCH_SIZE: 8
+      USE_CONV: False
+      SCALE_FACTOR: 0.3611
+      SHIFT_FACTOR: 0.1159
+      #
+      ENCODER:
+        NAME: Encoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'T5PlusClipFluxEmbedder'
+      NAME: T5PlusClipFluxEmbedder
+      # T5_MODEL DESCRIPTION:  TYPE:  default: ''
+      T5_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: T5EncoderModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder_2/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: T5Tokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer_2/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 512
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: last_hidden_state
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: False
+        CLEAN: whitespace
+      # CLIP_MODEL DESCRIPTION:  TYPE:  default: ''
+      CLIP_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: CLIPTextModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: CLIPTokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 77
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: pooler_output
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: True
+        CLEAN: whitespace
+  #
+  SAMPLE_ARGS:
+    SAMPLE_STEPS: 50
+    SAMPLER: flow_eluer
+    SEED: 2024
+    IMAGE_SIZE: [ 1024, 1024 ]
+    SHIFT: True
+    GUIDE_SCALE: 3.5
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 4e-4
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a cat holds a blackboard that writes \"hello world\"", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 2
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      PRIORITY: 0
+    - NAME: BackwardHook
+#      GRADIENT_CLIP: 1.0
+      PRIORITY: 10
+    - NAME: LogHook
+      LOG_INTERVAL: 10
+    - NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  EVAL_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/studio/self_train/dit/flux1.0_schnell_pro.yaml

@@ -0,0 +1,342 @@
+ENV:
+  BACKEND: nccl
+META:
+  VERSION: 'FLUX1.0_SCHNELL'
+  DESCRIPTION: "flux 1.0 schnell"
+  IS_DEFAULT: False
+  IS_SHARE: True
+  INFERENCE_PARAS:
+    INFERENCE_BATCH_SIZE: 1
+    INFERENCE_PREFIX: ""
+    DEFAULT_SAMPLER: "flow_euler"
+    DEFAULT_SAMPLE_STEPS: 4
+    INFERENCE_N_PROMPT: ""
+    RESOLUTION: [1024, 1024]
+  PARAS:
+    -
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 89000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 4e-4
+      IS_DEFAULT: False
+      TUNER: FULL
+    -
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 89000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 4e-4
+      IS_DEFAULT: True
+      TUNER: LORA
+  #
+  TUNERS:
+    LORA:
+      -
+        NAME: SwiftLoRA
+        R: 4
+        LORA_ALPHA: 4
+        LORA_DROPOUT: 0.0
+        BIAS: "none"
+        TARGET_MODULES: "(model.double_blocks.*(.qkv|.proj|.img_mod.lin|.txt_mod.lin))|(model.single_blocks.*(.linear1|.linear2|.modulation.lin))$"
+#
+SOLVER:
+  NAME: LatentDiffusionSolver
+  # MAX_STEPS DESCRIPTION: The total steps for training. TYPE: int default: 100000
+  MAX_STEPS: 100000
+  # USE_AMP DESCRIPTION: Use amp to surpport mix precision or not, default is False. TYPE: bool default: False
+  USE_AMP: True
+  # DTYPE DESCRIPTION: The precision for training. TYPE: str default: 'float32'
+  DTYPE: bfloat16
+  # USE_FAIRSCALE DESCRIPTION: Use fairscale as the backend of ddp, default False. TYPE: bool default: False
+  USE_FAIRSCALE: False
+  # USE_FSDP DESCRIPTION: Use fsdp as the backend of ddp, default False. TYPE: bool default: False
+  USE_FSDP: True
+  # LOAD_MODEL_ONLY DESCRIPTION: Only load the model rather than the optimizer and schedule, default is False. TYPE: bool default: False
+  LOAD_MODEL_ONLY: False
+  # RESUME_FROM DESCRIPTION: Resume from some state of training! TYPE: str default: ''
+  RESUME_FROM:
+  WORK_DIR: ./cache/save_data/dit_flux_schnell_1024_lora
+  LOG_FILE: std_log.txt
+  # EVAL_INTERVAL DESCRIPTION: Eval the model interval. TYPE: int default: 1
+  EVAL_INTERVAL: 100
+  # LOG_TRAIN_NUM DESCRIPTION: The number samples used to log in training phase. TYPE: int default: -1
+  LOG_TRAIN_NUM: 16
+  # FSDP_REDUCE_DTYPE DESCRIPTION: The dtype of reduce in FSDP. TYPE: str default: 'float16'
+  FSDP_REDUCE_DTYPE: float32
+  # FSDP_BUFFER_DTYPE DESCRIPTION: The dtype of buffer in FSDP. TYPE: str default: 'float16'
+  FSDP_BUFFER_DTYPE: float32
+  # FSDP_SHARD_MODULES DESCRIPTION: The modules to be sharded in FSDP. TYPE: list default: ['model']
+  FSDP_SHARD_MODULES: [ 'model', 'cond_stage_model.t5_model' ] #
+  SAVE_MODULES: [ 'model'] #
+  TRAIN_MODULES: ['model']
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  #
+  TUNER:
+  #
+  MODEL:
+    NAME: LatentDiffusionFlux
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    DEFAULT_N_PROMPT:
+    USE_EMA: False
+    EVAL_EMA: False
+    DIFFUSION:
+      # NAME DESCRIPTION:  TYPE:  default: 'DiffusionFluxRF'
+      NAME: DiffusionFluxRF
+      PREDICTION_TYPE: raw
+      # NOISE_SCHEDULER DESCRIPTION:  TYPE:  default: ''
+      NOISE_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchSigmaScheduler'
+        NAME: FlowMatchSigmaScheduler
+        # WEIGHTING_SCHEME DESCRIPTION: The weighting scheme for sampling timesteps, choose from ['sigma_sqrt', 'logit_normal', 'mode', 'cosmap', 'none']. TYPE: str default: 'logit_normal'
+        WEIGHTING_SCHEME: logit_normal
+        SHIFT: 3.0
+        # LOGIT_MEAN DESCRIPTION: The mean of the logit distribution for sampling timesteps. TYPE: float default: 0.0
+        LOGIT_MEAN: 0.0
+        # LOGIT_STD DESCRIPTION: The standard deviation of the logit distribution for sampling timesteps. TYPE: float default: 1.0
+        LOGIT_STD: 1.0
+        # MODE_SCALE DESCRIPTION: The scale factor for the mode of the logit distribution for sampling timesteps. TYPE: float default: 1.29
+        MODE_SCALE: 1.29
+      SAMPLER_SCHEDULER:
+        # NAME DESCRIPTION:  TYPE:  default: 'FlowMatchFluxShiftScheduler'
+        NAME: FlowMatchFluxShiftScheduler
+        # SHIFT DESCRIPTION: Use timestamp shift or not, default is True. TYPE: bool default: True
+        SHIFT: False
+        # SIGMOID_SCALE DESCRIPTION: The scale of sigmoid function for sampling timesteps. TYPE: int default: 1
+        SIGMOID_SCALE: 1
+        # BASE_SHIFT DESCRIPTION: The base shift factor for the timestamp. TYPE: float default: 0.5
+        BASE_SHIFT: 0.5
+        # MAX_SHIFT DESCRIPTION: The max shift factor for the timestamp. TYPE: float default: 1.15
+        MAX_SHIFT: 1.15
+        #
+    DIFFUSION_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'Flux'
+      NAME: Flux
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-schnell@flux1-schnell.safetensors
+      # IN_CHANNELS DESCRIPTION: model's input channels. TYPE: int default: 64
+      IN_CHANNELS: 64
+      # HIDDEN_SIZE DESCRIPTION: model's hidden size. TYPE: int default: 1024
+      HIDDEN_SIZE: 3072
+      # NUM_HEADS DESCRIPTION: number of heads in the transformer. TYPE: int default: 16
+      NUM_HEADS: 24
+      # AXES_DIM DESCRIPTION: dimensions of the axes of the positional encoding. TYPE: list default: [16, 56, 56]
+      AXES_DIM: [ 16, 56, 56 ]
+      # THETA DESCRIPTION: theta for positional encoding. TYPE: int default: 10000
+      THETA: 10000
+      # VEC_IN_DIM DESCRIPTION: dimension of the vector input. TYPE: int default: 768
+      VEC_IN_DIM: 768
+      # GUIDANCE_EMBED DESCRIPTION: whether to use guidance embedding. TYPE: bool default: False
+      GUIDANCE_EMBED: False
+      # CONTEXT_IN_DIM DESCRIPTION: dimension of the context input. TYPE: int default: 4096
+      CONTEXT_IN_DIM: 4096
+      # MLP_RATIO DESCRIPTION: ratio of mlp hidden size to hidden size. TYPE: float default: 4.0
+      MLP_RATIO: 4.0
+      # QKV_BIAS DESCRIPTION: whether to use bias in qkv projection. TYPE: bool default: True
+      QKV_BIAS: True
+      # DEPTH DESCRIPTION: number of transformer blocks. TYPE: int default: 19
+      DEPTH: 19
+      # DEPTH_SINGLE_BLOCKS DESCRIPTION: number of transformer blocks in the single stream block. TYPE: int default: 38
+      DEPTH_SINGLE_BLOCKS: 38
+      USE_GRAD_CHECKPOINT: True
+
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKLFlux
+      EMBED_DIM: 16
+      PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-schnell@ae.safetensors
+      IGNORE_KEYS: [ ]
+      BATCH_SIZE: 8
+      USE_CONV: False
+      SCALE_FACTOR: 0.3611
+      SHIFT_FACTOR: 0.1159
+      #
+      ENCODER:
+        NAME: Encoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        USE_CHECKPOINT: True
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      # NAME DESCRIPTION:  TYPE:  default: 'T5PlusClipFluxEmbedder'
+      NAME: T5PlusClipFluxEmbedder
+      # T5_MODEL DESCRIPTION:  TYPE:  default: ''
+      T5_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: T5EncoderModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-schnell@text_encoder_2/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: T5Tokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-schnell@tokenizer_2/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 256
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: last_hidden_state
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: False
+        CLEAN: whitespace
+      # CLIP_MODEL DESCRIPTION:  TYPE:  default: ''
+      CLIP_MODEL:
+        # NAME DESCRIPTION:  TYPE:  default: 'HFEmbedder'
+        NAME: HFEmbedder
+        # HF_MODEL_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_MODEL_CLS: CLIPTextModel
+        # MODEL_PATH DESCRIPTION: model folder path TYPE: NoneType default: None
+        MODEL_PATH: ms://AI-ModelScope/FLUX.1-schnell@text_encoder/
+        # HF_TOKENIZER_CLS DESCRIPTION: huggingface cls in transfomer TYPE: NoneType default: None
+        HF_TOKENIZER_CLS: CLIPTokenizer
+        # TOKENIZER_PATH DESCRIPTION: tokenizer folder path TYPE: NoneType default: None
+        TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-schnell@tokenizer/
+        # MAX_LENGTH DESCRIPTION: max length of input TYPE: int default: 77
+        MAX_LENGTH: 77
+        # OUTPUT_KEY DESCRIPTION: output key TYPE: str default: 'last_hidden_state'
+        OUTPUT_KEY: pooler_output
+        # D_TYPE DESCRIPTION: dtype TYPE: str default: 'bfloat16'
+        D_TYPE: bfloat16
+        # BATCH_INFER DESCRIPTION: batch infer TYPE: bool default: False
+        BATCH_INFER: True
+        CLEAN: whitespace
+  #
+  SAMPLE_ARGS:
+    SAMPLE_STEPS: 4
+    SAMPLER: flow_eluer
+    SEED: 2024
+    IMAGE_SIZE: [ 1024, 1024 ]
+    GUIDE_SCALE: 3.5
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 4e-4
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a cat holds a blackboard that writes \"hello world\"", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 2
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      PRIORITY: 0
+    - NAME: BackwardHook
+#      GRADIENT_CLIP: 1.0
+      PRIORITY: 10
+    - NAME: LogHook
+      LOG_INTERVAL: 10
+    - NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  EVAL_HOOKS:
+    - NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/studio/self_train/dit/pixart_alpha_pro.yaml

@@ -0,0 +1,264 @@
+ENV:
+  BACKEND: nccl
+META:
+  VERSION: 'PIXART_ALPHA'
+  DESCRIPTION: "PIXART ALPHA"
+  IS_DEFAULT: False
+  IS_SHARE: True
+  INFERENCE_PARAS:
+    INFERENCE_BATCH_SIZE: 1
+    INFERENCE_PREFIX: ""
+    DEFAULT_SAMPLER: "ddim"
+    DEFAULT_SAMPLE_STEPS: 20
+    INFERENCE_N_PROMPT: ""
+    RESOLUTION: [1024, 1024]
+  PARAS:
+    -
+      TRAIN_BATCH_SIZE: 2
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 0.0001
+      IS_DEFAULT: False
+      TUNER: FULL
+    -
+      TRAIN_BATCH_SIZE: 2
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 0.0001
+      IS_DEFAULT: False
+      TUNER: LORA
+  #
+  TUNERS:
+    LORA:
+      -
+        NAME: SwiftLoRA
+        R: 256
+        LORA_ALPHA: 256
+        LORA_DROPOUT: 0.0
+        BIAS: "none"
+        TARGET_MODULES: "model.*(.q|.k|.v|.o|mlp.fc1|mlp.fc2)$"
+#
+SOLVER:
+  NAME: LatentDiffusionSolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 1000
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: -1
+  RESCALE_LR: False
+  #
+  WORK_DIR: cache/scepter_ui/self_train/dit/pixart_alpha_pro
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  #
+  TUNER:
+  #
+  MODEL:
+    NAME: LatentDiffusionPixart
+    PARAMETERIZATION: eps
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    SCALE_FACTOR: 0.18215
+    SIZE_FACTOR: 8
+    DECODER_BIAS: 0.5
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "linear"
+      "BETA_MIN": 0.0001
+      "BETA_MAX": 0.02
+    USE_EMA: False
+    LOAD_REFINER: False
+    #
+    DIFFUSION_MODEL:
+      NAME: PixArt
+      PRETRAINED_MODEL: ms://AI-ModelScope/PixArt-alpha@PixArt-XL-2-1024-MS.pth
+      INPUT_SIZE: 128
+      PATCH_SIZE: 2
+      IN_CHANNELS: 4
+      HIDDEN_SIZE: 1152
+      DEPTH: 28
+      NUM_HEADS: 16
+      MLP_RATIO: 4.0
+      CLASS_DROPOUT_PROB: 0.1
+      PRED_SIGMA: True
+      DROP_PATH: 0.0
+      WINDOW_DIZE: 0
+      USE_REL_POS: False
+      CAPTION_CHANNELS: 4096
+      LEWEI_SCALE: 2
+      MODEL_MAX_LENGTH: 120
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-2-base@512-base-ema.safetensors
+      EMBED_DIM: 4
+      IGNORE_KEYS: [ ]
+      BATCH_SIZE: 1
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      NAME: T5EmbedderHF
+      PRETRAINED_MODEL: ms://AI-ModelScope/PixArt-alpha@t5-v1_1-xxl/
+      TOKENIZER_PATH: ms://AI-ModelScope/PixArt-alpha@t5-v1_1-xxl/
+      LENGTH: 120
+      CLEAN: heavy
+      USE_GRAD: False
+    #
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: ddim
+    SAMPLE_STEPS: 20
+    SEED: 2024
+    GUIDE_SCALE: 4.5
+    GUIDE_RESCALE: 0.5
+    DISCRETIZATION: trailing
+    RUN_TRAIN_N: False
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 0.0001
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a boy wearing a jacket", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 0
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 10
+      SHOW_GPU_MEM: True
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  #
+  EVAL_HOOKS:
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/studio/self_train/dit/sd3_pro.yaml

@@ -0,0 +1,284 @@
+ENV:
+  BACKEND: nccl
+META:
+  VERSION: 'SD3'
+  DESCRIPTION: "SD3 ALPHA"
+  IS_DEFAULT: False
+  IS_SHARE: True
+  INFERENCE_PARAS:
+    INFERENCE_BATCH_SIZE: 1
+    INFERENCE_PREFIX: ""
+    DEFAULT_SAMPLER: "euler"
+    DEFAULT_SAMPLE_STEPS: 28
+    INFERENCE_N_PROMPT: ""
+    RESOLUTION: [1024, 1024]
+  PARAS:
+    -
+      TRAIN_BATCH_SIZE: 2
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 1e-5
+      IS_DEFAULT: False
+      TUNER: FULL
+    -
+      TRAIN_BATCH_SIZE: 2
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [1024, 1024]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 5e-5
+      IS_DEFAULT: True
+      TUNER: LORA
+  #
+  TUNERS:
+    LORA:
+      -
+        NAME: SwiftLoRA
+        R: 128
+        LORA_ALPHA: 128
+        LORA_DROPOUT: 0.0
+        BIAS: "none"
+        TARGET_MODULES: "model.*(.attn.qkv|.attn.proj|mlp.fc1|mlp.fc2)$"
+#
+SOLVER:
+  NAME: LatentDiffusionSolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 1000
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: -1
+  RESCALE_LR: False
+  #
+  WORK_DIR: cache/scepter_ui/self_train/dit/sd3
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    - NAME: "ModelscopeFs"
+      TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  #
+  TUNER:
+  #
+  MODEL:
+    NAME: LatentDiffusionSD3
+    PARAMETERIZATION: rf
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL:
+    IGNORE_KEYS: [ ]
+    SCALE_FACTOR: 1.5305
+    SHIFT_FACTOR: 0.0609
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "shifted"
+      "SHIFT": 3
+    USE_EMA: False
+    T_WEIGHT: uniform
+    #
+    DIFFUSION_MODEL:
+      NAME: MMDiT
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+      IGNORE_KEYS: '^first_stage_model.'
+      IN_CHANNELS: 16
+      PATCH_SIZE: 2
+      OUT_CHANNELS: 16
+      DEPTH: 24
+      INPUT_SIZE:
+      ADM_IN_CHANNELS: 2048
+      CONTEXT_EMBEDDER_CONFIG: { 'target': 'torch.nn.Linear', 'params': { 'in_features': 4096, 'out_features': 1536 } }
+      NUM_PATCHES: 36864
+      POS_EMBED_MAX_SIZE: 192
+      POS_EMBED_SCALING_FACTOR:
+      USE_CHECKPOINT: True
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium@sd3_medium.safetensors
+      EMBED_DIM: 16
+      IGNORE_KEYS: '^model.diffusion_model.'
+      BATCH_SIZE: 1
+      USE_CONV: False
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 16
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    COND_STAGE_MODEL:
+      NAME: SD3TextEmbedder
+      P_ZERO: 0.0
+      CLIP_L:
+        NAME: FrozenCLIPEmbedder2
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer
+        MAX_LENGTH: 77
+        FREEZE: True
+        LAYER: penultimate
+        RETURN_POOLED: True
+        USE_FINAL_LAYER_NORM: False
+        IS_TRAINABLE: False
+      CLIP_G:
+        NAME: FrozenCLIPEmbedder2
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_2
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_2
+        MAX_LENGTH: 77
+        FREEZE: True
+        LAYER: penultimate
+        RETURN_POOLED: True
+        USE_FINAL_LAYER_NORM: False
+        IS_TRAINABLE: False
+      T5_XXL:
+        NAME: T5EmbedderHF
+        PRETRAINED_MODEL: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@text_encoder_3
+        TOKENIZER_PATH: ms://AI-ModelScope/stable-diffusion-3-medium-diffusers@tokenizer_3
+        LENGTH: 256
+        CLEAN: whitespace
+        USE_GRAD: False
+        T5_DTYPE: float16
+    #
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: euler
+    SAMPLE_STEPS: 28
+    SEED: 1749023094
+    GUIDE_SCALE: 5.0
+    GUIDE_RESCALE: 0.0
+    DISCRETIZATION: trailing
+    RUN_TRAIN_N: False
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 5e-5
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: style_custom_dataset
+    MS_DATASET_NAMESPACE: damo
+    MS_DATASET_SUBNAME: 3D
+    PROMPT_PREFIX: ""
+    MS_DATASET_SPLIT: train
+    MS_REMAP_KEYS: { 'Image:FILE': 'Target:FILE' }
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFile
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 1024, 1024 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'img' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img' ]
+        OUTPUT_KEY: [ 'image' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [ "a cat holds a blackboard that writes \"hello world\"", "a dog running on the lawn" ]
+    IMAGE_SIZE: [ 1024, 1024 ]
+    FIELDS: [ "prompt" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: Select
+        KEYS: [ 'index', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 1
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 10
+      SHOW_GPU_MEM: True
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  #
+  EVAL_HOOKS:
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/studio/self_train/edit/stylebooth_tb_pro.yaml

@@ -0,0 +1,377 @@
+ENV:
+  BACKEND: nccl
+META:
+  VERSION: 'EDIT'
+  DESCRIPTION: "EDIT"
+  IS_DEFAULT: False
+  IS_SHARE: True
+  INFERENCE_PARAS:
+    INFERENCE_BATCH_SIZE: 1
+    INFERENCE_PREFIX: ""
+    DEFAULT_SAMPLER: "ddim"
+    DEFAULT_SAMPLE_STEPS: 40
+    INFERENCE_N_PROMPT: ""
+    RESOLUTION: [512, 512]
+  PARAS:
+    -
+      TASK: 'Image Editing'
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [512, 512]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 0.0001
+      IS_DEFAULT: False
+      TUNER: FULL
+    -
+      TASK: 'Image Editing'
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [512, 512]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 0.0001
+      IS_DEFAULT: True
+      TUNER: LORA
+    -
+      TASK: 'Image Editing'
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [512, 512]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 0.0001
+      IS_DEFAULT: False
+      TUNER: SCE
+
+    -
+      TASK: 'Image Editing'
+      TRAIN_BATCH_SIZE: 1
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [512, 512]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 0.0001
+      IS_DEFAULT: False
+      TUNER: TEXT_SCE
+
+    -
+      TASK: 'Image Editing'
+      TRAIN_BATCH_SIZE: 4
+      TRAIN_PREFIX: ""
+      TRAIN_N_PROMPT: ""
+      RESOLUTION: [512, 512]
+      MEMORY: 29000
+      EPOCHS: 50
+      SAVE_INTERVAL: 25
+      EPSEC: 0.818
+      LEARNING_RATE: 0.0001
+      IS_DEFAULT: False
+      TUNER: TEXT_LORA
+
+  TUNERS:
+    LORA:
+      -
+        NAME: SwiftLoRA
+        R: 256
+        LORA_ALPHA: 256
+        LORA_DROPOUT: 0.0
+        BIAS: "none"
+        TARGET_MODULES: "model.*(to_q|to_k|to_v|to_out.0|net.0.proj|net.2)$"
+    TEXT_LORA:
+      -
+        NAME: SwiftLoRA
+        R: 256
+        LORA_ALPHA: 256
+        LORA_DROPOUT: 0.0
+        BIAS: "none"
+        TARGET_MODULES: "(cond_stage_model.*(q_proj|k_proj|v_proj|out_proj|mlp.fc1|mlp.fc2))|(model.*(to_q|to_k|to_v|to_out.0|net.0.proj|net.2))$"
+    SCE:
+      -
+        NAME: SwiftSCETuning
+        DIMS: [1280, 1280, 1280, 1280, 1280, 640, 640, 640, 320, 320, 320, 320]
+        DOWN_RATIO: 1.0
+        TARGET_MODULES: model.lsc_identity\.\d+$
+        TUNER_MODE: identity
+    TEXT_SCE:
+      -
+        NAME: SwiftSCETuning
+        DIMS: [ 1280, 1280, 1280, 1280, 1280, 640, 640, 640, 320, 320, 320, 320 ]
+        DOWN_RATIO: 1.0
+        TARGET_MODULES: model.lsc_identity\.\d+$
+        TUNER_MODE: identity
+      -
+        NAME: SwiftLoRA
+        R: 256
+        LORA_ALPHA: 256
+        LORA_DROPOUT: 0.0
+        BIAS: "none"
+        TARGET_MODULES: "cond_stage_model.*(q_proj|k_proj|v_proj|out_proj|mlp.fc1|mlp.fc2)$"
+
+  MODIFY_PARAS:
+    TEXT_LORA:
+      TRAIN:
+        SOLVER.MODEL.COND_STAGE_MODEL.USE_GRAD: True
+    TEXT_SCE:
+      TRAIN:
+        SOLVER.MODEL.COND_STAGE_MODEL.USE_GRAD: True
+
+SOLVER:
+  NAME: LatentDiffusionSolver
+  RESUME_FROM:
+  LOAD_MODEL_ONLY: True
+  USE_FSDP: False
+  SHARDING_STRATEGY:
+  USE_AMP: True
+  DTYPE: float16
+  CHANNELS_LAST: True
+  MAX_STEPS: 1000
+  MAX_EPOCHS: -1
+  NUM_FOLDS: 1
+  ACCU_STEP: 1
+  EVAL_INTERVAL: -1
+  RESCALE_LR: False
+  #
+  WORK_DIR:
+  LOG_FILE: std_log.txt
+  #
+  FILE_SYSTEM:
+    NAME: "ModelscopeFs"
+    TEMP_DIR: "./cache/cache_data"
+  #
+  FREEZE:
+  #
+  TUNER:
+  #
+  MODEL:
+    NAME: LatentDiffusionEdit
+    PARAMETERIZATION: eps
+    TIMESTEPS: 1000
+    MIN_SNR_GAMMA:
+    ZERO_TERMINAL_SNR: False
+    PRETRAINED_MODEL: ms://iic/stylebooth@models/stylebooth-tb-5000-0.bin
+    IGNORE_KEYS: [ ]
+    CONCAT_NO_SCALE_FACTOR: True
+    SCALE_FACTOR: 0.18215
+    SIZE_FACTOR: 8
+    # DEFAULT_N_PROMPT: 'lowres, error, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, username, watermark, signature'
+    DEFAULT_N_PROMPT:
+    SCHEDULE_ARGS:
+      "NAME": "scaled_linear"
+      "BETA_MIN": 0.00085
+      "BETA_MAX": 0.012
+    USE_EMA: False
+    #
+    DIFFUSION_MODEL:
+      NAME: DiffusionUNet
+      IN_CHANNELS: 8
+      OUT_CHANNELS: 4
+      MODEL_CHANNELS: 320
+      NUM_HEADS: 8
+      NUM_RES_BLOCKS: 2
+      ATTENTION_RESOLUTIONS: [ 4, 2, 1 ]
+      CHANNEL_MULT: [ 1, 2, 4, 4 ]
+      CONV_RESAMPLE: True
+      DIMS: 2
+      USE_CHECKPOINT: False
+      USE_SCALE_SHIFT_NORM: False
+      RESBLOCK_UPDOWN: False
+      USE_SPATIAL_TRANSFORMER: True
+      TRANSFORMER_DEPTH: 1
+      CONTEXT_DIM: 768
+      DISABLE_MIDDLE_SELF_ATTN: False
+      USE_LINEAR_IN_TRANSFORMER: False
+      IGNORE_KEYS: []
+    #
+    FIRST_STAGE_MODEL:
+      NAME: AutoencoderKL
+      EMBED_DIM: 4
+      IGNORE_KEYS: []
+      BATCH_SIZE: 4
+      #
+      ENCODER:
+        NAME: Encoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DOUBLE_Z: True
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+      #
+      DECODER:
+        NAME: Decoder
+        CH: 128
+        OUT_CH: 3
+        NUM_RES_BLOCKS: 2
+        IN_CHANNELS: 3
+        ATTN_RESOLUTIONS: [ ]
+        CH_MULT: [ 1, 2, 4, 4 ]
+        Z_CHANNELS: 4
+        DROPOUT: 0.0
+        RESAMP_WITH_CONV: True
+        GIVE_PRE_END: False
+        TANH_OUT: False
+    #
+    TOKENIZER:
+      NAME: ClipTokenizer
+      PRETRAINED_PATH: ms://AI-ModelScope/clip-vit-large-patch14
+      LENGTH: 77
+      CLEAN: True
+    #
+    COND_STAGE_MODEL:
+      NAME: FrozenCLIPEmbedder
+      FREEZE: True
+      USE_GRAD: False
+      LAYER: last
+      PRETRAINED_MODEL: ms://AI-ModelScope/clip-vit-large-patch14
+    #
+    LOSS:
+      NAME: ReconstructLoss
+      LOSS_TYPE: l2
+  #
+  SAMPLE_ARGS:
+    SAMPLER: ddim
+    SAMPLE_STEPS: 50
+    SEED: 2023
+    GUIDE_SCALE: #7.5
+      image: 1.5
+      text: 7.5
+    GUIDE_RESCALE: 0.5
+    DISCRETIZATION: trailing
+    IMAGE_SIZE: [512, 512]
+    RUN_TRAIN_N: False
+  #
+  OPTIMIZER:
+    NAME: AdamW
+    LEARNING_RATE: 0.0001
+    BETAS: [ 0.9, 0.999 ]
+    EPS: 1e-8
+    WEIGHT_DECAY: 1e-2
+    AMSGRAD: False
+  #
+  TRAIN_DATA:
+    NAME: ImageTextPairMSDataset
+    MODE: train
+    MS_DATASET_NAME: ./cache/cache_data/delogo/
+    MS_DATASET_NAMESPACE: ""
+    MS_DATASET_SPLIT: "train"
+    MS_DATASET_SUBNAME: ""
+    PROMPT_PREFIX: ""
+    REPLACE_STYLE: False
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    SAMPLER:
+      NAME: LoopSampler
+    TRANSFORMS:
+      - NAME: LoadImageFromFileList
+        FILE_KEYS: ['img_path', 'src_path']
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'img', 'src' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'img', 'src' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'img', 'src' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'img', 'src' ]
+        OUTPUT_KEY: [ 'image', 'condition_cat' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'image', 'condition_cat', 'prompt' ]
+        META_KEYS: [ 'data_key' ]
+  #
+  TRAIN_HOOKS:
+    -
+      NAME: BackwardHook
+      PRIORITY: 0
+    -
+      NAME: LogHook
+      LOG_INTERVAL: 10
+      SHOW_GPU_MEM: True
+    -
+      NAME: TensorboardLogHook
+    -
+      NAME: CheckpointHook
+      INTERVAL: 10000
+      PRIORITY: 200
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      DISABLE_SNAPSHOT: True
+  #
+  EVAL_DATA:
+    NAME: Text2ImageDataset
+    MODE: eval
+    PROMPT_FILE:
+    PROMPT_DATA: [  ]
+    IMAGE_SIZE: [ 512, 512 ]
+    FIELDS: [ "prompt", "src_path" ]
+    DELIMITER: '#;#'
+    PROMPT_PREFIX: ''
+    PIN_MEMORY: True
+    BATCH_SIZE: 1
+    NUM_WORKERS: 4
+    TRANSFORMS:
+      - NAME: LoadImageFromFileList
+        FILE_KEYS: [ 'src_path' ]
+        RGB_ORDER: RGB
+        BACKEND: pillow
+      - NAME: FlexibleResize
+        INTERPOLATION: bilinear
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'src' ]
+        BACKEND: pillow
+      - NAME: FlexibleCenterCrop
+        SIZE: [ 512, 512 ]
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'src' ]
+        BACKEND: pillow
+      - NAME: ImageToTensor
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'src' ]
+        BACKEND: pillow
+      - NAME: Normalize
+        MEAN: [ 0.5,  0.5,  0.5 ]
+        STD: [ 0.5,  0.5,  0.5 ]
+        INPUT_KEY: [ 'src' ]
+        OUTPUT_KEY: [ 'condition_cat' ]
+        BACKEND: torchvision
+      - NAME: Select
+        KEYS: [ 'condition_cat', 'prompt' ]
+        META_KEYS: [ 'image_size' ]
+  EVAL_HOOKS:
+    -
+      NAME: ProbeDataHook
+      PROB_INTERVAL: 100
+      SAVE_LAST: True
+      SAVE_NAME_PREFIX: 'step'
+      SAVE_PROBE_PREFIX: 'image'

scepter/methods/studio/self_train/sd_xl/sdxl_pro.yaml

@@ -147,6 +147,7 @@ SOLVER:
   MAX_EPOCHS: -1
   # NUM_FOLDS DESCRIPTION: Num folds for training. TYPE: int default: 1
   NUM_FOLDS: 1
+  RESCALE_LR: False
   #
   EVAL_INTERVAL: -1
   # WORK_DIR DESCRIPTION: Save dir of the training log or model. TYPE: str default: ''
@@ -155,7 +156,7 @@ SOLVER:
   LOG_FILE: std_log.txt
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   TUNER:
   # MODEL DESCRIPTION:  TYPE:  default: ''
   MODEL:
@@ -538,7 +539,7 @@ SOLVER:
   OPTIMIZER:
     # NAME DESCRIPTION:  TYPE:  default: ''
     NAME: AdamW
-    LEARNING_RATE: 0.0064
+    LEARNING_RATE: 0.00001
     EPS: 1e-8
     AMSGRAD: False
   #

scepter/methods/studio/self_train/stable_diffusion/sd15_pro.yaml

@@ -137,13 +137,14 @@ SOLVER:
   NUM_FOLDS: 1
   ACCU_STEP: 1
   EVAL_INTERVAL: -1
+  RESCALE_LR: False
   #
   WORK_DIR:
   LOG_FILE: std_log.txt
   #
   FILE_SYSTEM:
     NAME: "ModelscopeFs"
-    TEMP_DIR: "./cache/data"
+    TEMP_DIR: "./cache/cache_data"
   #
   FREEZE:
   #
@@ -249,7 +250,7 @@ SOLVER:
   #
   OPTIMIZER:
     NAME: AdamW
-    LEARNING_RATE: 0.064
+    LEARNING_RATE: 0.0001
     BETAS: [ 0.9, 0.999 ]
     EPS: 1e-8
     WEIGHT_DECAY: 1e-2