hcpdiff 0.9.1__tar.gz → 2.1__tar.gz

hcpdiff-2.1/PKG-INFO

@@ -0,0 +1,285 @@
+Metadata-Version: 2.4
+Name: hcpdiff
+Version: 2.1
+Summary: A universal Diffusion toolbox
+Home-page: https://github.com/IrisRainbowNeko/HCP-Diffusion
+Author: Ziyi Dong
+Author-email: rainbow-neko@outlook.com
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.8
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: rainbowneko
+Requires-Dist: diffusers
+Requires-Dist: matplotlib
+Requires-Dist: pyarrow
+Requires-Dist: transformers>=4.25.1
+Requires-Dist: pytorch-msssim
+Requires-Dist: lmdb
+Dynamic: author
+Dynamic: author-email
+Dynamic: classifier
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: license-file
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+# HCP-Diffusion V2
+
+[![PyPI](https://img.shields.io/pypi/v/hcpdiff)](https://pypi.org/project/hcpdiff/)
+[![GitHub stars](https://img.shields.io/github/stars/7eu7d7/HCP-Diffusion)](https://github.com/7eu7d7/HCP-Diffusion/stargazers)
+[![GitHub license](https://img.shields.io/github/license/7eu7d7/HCP-Diffusion)](https://github.com/7eu7d7/HCP-Diffusion/blob/master/LICENSE)
+[![codecov](https://codecov.io/gh/7eu7d7/HCP-Diffusion/branch/main/graph/badge.svg)](https://codecov.io/gh/7eu7d7/HCP-Diffusion)
+[![open issues](https://isitmaintained.com/badge/open/7eu7d7/HCP-Diffusion.svg)](https://github.com/7eu7d7/HCP-Diffusion/issues)
+
+[📘中文说明](./README_cn.md)
+
+[📘English document](https://hcpdiff.readthedocs.io/en/latest/)
+[📘中文文档](https://hcpdiff.readthedocs.io/zh_CN/latest/)
+
+Old HCP-Diffusion V1 at [main branch](https://github.com/IrisRainbowNeko/HCP-Diffusion/tree/main)
+
+## Introduction
+
+**HCP-Diffusion** is a Diffusion model toolbox built on top of the [🐱 RainbowNeko Engine](https://github.com/IrisRainbowNeko/RainbowNekoEngine).  
+It features a clean code structure and a flexible **Python-based configuration file**, making it easier to conduct and manage complex experiments. It includes a wide variety of training components, and compared to existing frameworks, it's more extensible, flexible, and user-friendly.
+
+HCP-Diffusion allows you to use a single `.py` config file to unify training workflows across popular methods and model architectures, including Prompt-tuning (Textual Inversion), DreamArtist, Fine-tuning, DreamBooth, LoRA, ControlNet, ....  
+Different techniques can also be freely combined.
+
+This framework also implements **DreamArtist++**, an upgraded version of DreamArtist based on LoRA. It enables high generalization and controllability with just a single image for training.  
+Compared to the original DreamArtist, it offers better stability, image quality, controllability, and faster training.
+
+---
+
+## Installation
+
+Install via pip:
+
+```bash
+pip install hcpdiff
+# Initialize configuration
+hcpinit
+```
+
+Install from source:
+
+```bash
+git clone https://github.com/7eu7d7/HCP-Diffusion.git
+cd HCP-Diffusion
+pip install -e .
+# Initialize configuration
+hcpinit
+```
+
+Use xFormers to reduce memory usage and accelerate training:
+
+```bash
+# Choose the appropriate xformers version for your PyTorch version
+pip install xformers==?
+```
+
+## 🚀 Python Configuration Files
+RainbowNeko Engine supports configuration files written in a Python-like syntax. This allows users to call functions and classes directly within the configuration file, with function parameters inheritable from parent configuration files. The framework automatically handles the formatting of these configuration files.
+
+For example, consider the following configuration file:
+```python
+dict(
+    layer=Linear(in_features=4, out_features=4)
+)
+```
+During parsing, this will be automatically compiled into:
+```python
+dict(
+    layer=dict(_target_=Linear, in_features=4, out_features=4)
+)
+```
+After parsing, the framework will instantiate the components accordingly. This means users can write configuration files using familiar Python syntax.
+
+---
+
+## ✨ Features
+
+<details>
+<summary>Features</summary>
+
+### 📦 Model Support
+
+| Model Name                | Status      |
+|--------------------------|-------------|
+| Stable Diffusion 1.5     | ✅ Supported |
+| Stable Diffusion XL (SDXL)| ✅ Supported |
+| PixArt                   | ✅ Supported |
+| FLUX                     | 🚧 In Development |
+| Stable Diffusion 3 (SD3) | 🚧 In Development |
+
+---
+
+### 🧠 Fine-Tuning Capabilities
+
+| Feature                         | Description/Support |
+|----------------------------------|---------------------|
+| LoRA Layer-wise Configuration   | ✅ Supported (including Conv2d) |
+| Layer-wise Fine-Tuning          | ✅ Supported |
+| Multi-token Prompt-Tuning       | ✅ Supported |
+| Layer-wise Model Merging        | ✅ Supported |
+| Custom Optimizers               | ✅ Supported (Lion, DAdaptation, pytorch-optimizer, etc.) |
+| Custom LR Schedulers            | ✅ Supported |
+
+---
+
+### 🧩 Extension Method Support
+
+| Method                         | Status      |
+|--------------------------------|-------------|
+| ControlNet (including training)| ✅ Supported |
+| DreamArtist / DreamArtist++    | ✅ Supported |
+| Token Attention Adjustment     | ✅ Supported |
+| Max Sentence Length Extension  | ✅ Supported |
+| Textual Inversion (Custom Tokens)| ✅ Supported |
+| CLIP Skip                      | ✅ Supported |
+
+---
+
+### 🚀 Training Acceleration
+
+| Tool/Library                                       | Supported Modules        |
+|---------------------------------------------------|---------------------------|
+| [🤗 Accelerate](https://github.com/huggingface/accelerate)    | ✅ Supported |
+| [Colossal-AI](https://github.com/hpcaitech/ColossalAI)       | ✅ Supported |
+| [xFormers](https://github.com/facebookresearch/xformers)     | ✅ Supported (UNet and text encoder) |
+
+---
+
+### 🗂 Dataset Support
+
+| Feature                         | Description |
+|----------------------------------|-------------|
+| Aspect Ratio Bucket (ARB)       | ✅ Auto-clustering supported |
+| Multi-source / Multi-dataset    | ✅ Supported |
+| LMDB                            | ✅ Supported |
+| webdataset                      | 🚧 In Development |
+| Local Attention Enhancement     | ✅ Supported |
+| Tag Shuffling & Dropout         | ✅ Multiple tag editing strategies |
+
+---
+
+### 📉 Supported Loss Functions
+
+| Loss Type  | Description |
+|------------|-------------|
+| Min-SNR    | ✅ Supported |
+| SSIM       | ✅ Supported |
+| GWLoss     | ✅ Supported |
+
+---
+
+### 🌫 Supported Diffusion Strategies
+
+| Strategy Type   | Status       |
+|------------------|--------------|
+| DDPM             | ✅ Supported |
+| EDM              | ✅ Supported |
+| Flow Matching    | ✅ Supported |
+
+---
+
+### 🧠 Automatic Evaluation (Step Selection Assistant)
+
+| Feature         | Description/Status                       |
+|------------------|------------------------------------------|
+| Image Preview    | ✅ Supported (workflow preview)           |
+| FID              | 🚧 In Development                        |
+| CLIP Score       | 🚧 In Development                        |
+| CCIP Score       | 🚧 In Development                        |
+| Corrupt Score    | 🚧 In Development                        |
+
+</details>
+
+---
+
+## Getting Started
+
+### Training
+
+HCP-Diffusion provides training scripts based on 🤗 Accelerate.
+
+```bash
+# Multi-GPU training, configure GPUs in cfgs/launcher/multi.yaml
+hcp_train --cfg cfgs/train/py/your_config.py
+
+# Single-GPU training, configure GPU in cfgs/launcher/single.yaml
+hcp_train_1gpu --cfg cfgs/train/py/your_config.py
+```
+
+You can also override config items via command line:
+
+```bash
+# Override base model path
+hcp_train --cfg cfgs/train/py/your_config.py model.wrapper.models.ckpt_path=pretrained_model_path
+```
+
+### Image Generation
+
+Use the workflow defined in the Python config to generate images:
+
+```bash
+hcp_run --cfg cfgs/workflow/text2img.py
+```
+
+Or override parameters via command line:
+
+```bash
+hcp_run --cfg cfgs/workflow/text2img_cli.py \
+    pretrained_model=pretrained_model_path \
+    prompt='positive_prompt' \
+    negative_prompt='negative_prompt' \
+    seed=42
+```
+
+### Tutorials
+
+🚧 In Development
+
+---
+
+## Contributing
+
+We welcome contributions to support more models and features.
+
+---
+
+## Team
+
+Maintained by [HCP-Lab at Sun Yat-sen University](https://www.sysu-hcp.net/).
+
+---
+
+## Citation
+
+```bibtex
+@article{DBLP:journals/corr/abs-2211-11337,
+  author    = {Ziyi Dong and
+               Pengxu Wei and
+               Liang Lin},
+  title     = {DreamArtist: Towards Controllable One-Shot Text-to-Image Generation
+               via Positive-Negative Prompt-Tuning},
+  journal   = {CoRR},
+  volume    = {abs/2211.11337},
+  year      = {2022},
+  doi       = {10.48550/arXiv.2211.11337},
+  eprinttype = {arXiv},
+  eprint    = {2211.11337},
+}
+```

hcpdiff-2.1/README.md

@@ -0,0 +1,248 @@
+# HCP-Diffusion V2
+
+[![PyPI](https://img.shields.io/pypi/v/hcpdiff)](https://pypi.org/project/hcpdiff/)
+[![GitHub stars](https://img.shields.io/github/stars/7eu7d7/HCP-Diffusion)](https://github.com/7eu7d7/HCP-Diffusion/stargazers)
+[![GitHub license](https://img.shields.io/github/license/7eu7d7/HCP-Diffusion)](https://github.com/7eu7d7/HCP-Diffusion/blob/master/LICENSE)
+[![codecov](https://codecov.io/gh/7eu7d7/HCP-Diffusion/branch/main/graph/badge.svg)](https://codecov.io/gh/7eu7d7/HCP-Diffusion)
+[![open issues](https://isitmaintained.com/badge/open/7eu7d7/HCP-Diffusion.svg)](https://github.com/7eu7d7/HCP-Diffusion/issues)
+
+[📘中文说明](./README_cn.md)
+
+[📘English document](https://hcpdiff.readthedocs.io/en/latest/)
+[📘中文文档](https://hcpdiff.readthedocs.io/zh_CN/latest/)
+
+Old HCP-Diffusion V1 at [main branch](https://github.com/IrisRainbowNeko/HCP-Diffusion/tree/main)
+
+## Introduction
+
+**HCP-Diffusion** is a Diffusion model toolbox built on top of the [🐱 RainbowNeko Engine](https://github.com/IrisRainbowNeko/RainbowNekoEngine).  
+It features a clean code structure and a flexible **Python-based configuration file**, making it easier to conduct and manage complex experiments. It includes a wide variety of training components, and compared to existing frameworks, it's more extensible, flexible, and user-friendly.
+
+HCP-Diffusion allows you to use a single `.py` config file to unify training workflows across popular methods and model architectures, including Prompt-tuning (Textual Inversion), DreamArtist, Fine-tuning, DreamBooth, LoRA, ControlNet, ....  
+Different techniques can also be freely combined.
+
+This framework also implements **DreamArtist++**, an upgraded version of DreamArtist based on LoRA. It enables high generalization and controllability with just a single image for training.  
+Compared to the original DreamArtist, it offers better stability, image quality, controllability, and faster training.
+
+---
+
+## Installation
+
+Install via pip:
+
+```bash
+pip install hcpdiff
+# Initialize configuration
+hcpinit
+```
+
+Install from source:
+
+```bash
+git clone https://github.com/7eu7d7/HCP-Diffusion.git
+cd HCP-Diffusion
+pip install -e .
+# Initialize configuration
+hcpinit
+```
+
+Use xFormers to reduce memory usage and accelerate training:
+
+```bash
+# Choose the appropriate xformers version for your PyTorch version
+pip install xformers==?
+```
+
+## 🚀 Python Configuration Files
+RainbowNeko Engine supports configuration files written in a Python-like syntax. This allows users to call functions and classes directly within the configuration file, with function parameters inheritable from parent configuration files. The framework automatically handles the formatting of these configuration files.
+
+For example, consider the following configuration file:
+```python
+dict(
+    layer=Linear(in_features=4, out_features=4)
+)
+```
+During parsing, this will be automatically compiled into:
+```python
+dict(
+    layer=dict(_target_=Linear, in_features=4, out_features=4)
+)
+```
+After parsing, the framework will instantiate the components accordingly. This means users can write configuration files using familiar Python syntax.
+
+---
+
+## ✨ Features
+
+<details>
+<summary>Features</summary>
+
+### 📦 Model Support
+
+| Model Name                | Status      |
+|--------------------------|-------------|
+| Stable Diffusion 1.5     | ✅ Supported |
+| Stable Diffusion XL (SDXL)| ✅ Supported |
+| PixArt                   | ✅ Supported |
+| FLUX                     | 🚧 In Development |
+| Stable Diffusion 3 (SD3) | 🚧 In Development |
+
+---
+
+### 🧠 Fine-Tuning Capabilities
+
+| Feature                         | Description/Support |
+|----------------------------------|---------------------|
+| LoRA Layer-wise Configuration   | ✅ Supported (including Conv2d) |
+| Layer-wise Fine-Tuning          | ✅ Supported |
+| Multi-token Prompt-Tuning       | ✅ Supported |
+| Layer-wise Model Merging        | ✅ Supported |
+| Custom Optimizers               | ✅ Supported (Lion, DAdaptation, pytorch-optimizer, etc.) |
+| Custom LR Schedulers            | ✅ Supported |
+
+---
+
+### 🧩 Extension Method Support
+
+| Method                         | Status      |
+|--------------------------------|-------------|
+| ControlNet (including training)| ✅ Supported |
+| DreamArtist / DreamArtist++    | ✅ Supported |
+| Token Attention Adjustment     | ✅ Supported |
+| Max Sentence Length Extension  | ✅ Supported |
+| Textual Inversion (Custom Tokens)| ✅ Supported |
+| CLIP Skip                      | ✅ Supported |
+
+---
+
+### 🚀 Training Acceleration
+
+| Tool/Library                                       | Supported Modules        |
+|---------------------------------------------------|---------------------------|
+| [🤗 Accelerate](https://github.com/huggingface/accelerate)    | ✅ Supported |
+| [Colossal-AI](https://github.com/hpcaitech/ColossalAI)       | ✅ Supported |
+| [xFormers](https://github.com/facebookresearch/xformers)     | ✅ Supported (UNet and text encoder) |
+
+---
+
+### 🗂 Dataset Support
+
+| Feature                         | Description |
+|----------------------------------|-------------|
+| Aspect Ratio Bucket (ARB)       | ✅ Auto-clustering supported |
+| Multi-source / Multi-dataset    | ✅ Supported |
+| LMDB                            | ✅ Supported |
+| webdataset                      | 🚧 In Development |
+| Local Attention Enhancement     | ✅ Supported |
+| Tag Shuffling & Dropout         | ✅ Multiple tag editing strategies |
+
+---
+
+### 📉 Supported Loss Functions
+
+| Loss Type  | Description |
+|------------|-------------|
+| Min-SNR    | ✅ Supported |
+| SSIM       | ✅ Supported |
+| GWLoss     | ✅ Supported |
+
+---
+
+### 🌫 Supported Diffusion Strategies
+
+| Strategy Type   | Status       |
+|------------------|--------------|
+| DDPM             | ✅ Supported |
+| EDM              | ✅ Supported |
+| Flow Matching    | ✅ Supported |
+
+---
+
+### 🧠 Automatic Evaluation (Step Selection Assistant)
+
+| Feature         | Description/Status                       |
+|------------------|------------------------------------------|
+| Image Preview    | ✅ Supported (workflow preview)           |
+| FID              | 🚧 In Development                        |
+| CLIP Score       | 🚧 In Development                        |
+| CCIP Score       | 🚧 In Development                        |
+| Corrupt Score    | 🚧 In Development                        |
+
+</details>
+
+---
+
+## Getting Started
+
+### Training
+
+HCP-Diffusion provides training scripts based on 🤗 Accelerate.
+
+```bash
+# Multi-GPU training, configure GPUs in cfgs/launcher/multi.yaml
+hcp_train --cfg cfgs/train/py/your_config.py
+
+# Single-GPU training, configure GPU in cfgs/launcher/single.yaml
+hcp_train_1gpu --cfg cfgs/train/py/your_config.py
+```
+
+You can also override config items via command line:
+
+```bash
+# Override base model path
+hcp_train --cfg cfgs/train/py/your_config.py model.wrapper.models.ckpt_path=pretrained_model_path
+```
+
+### Image Generation
+
+Use the workflow defined in the Python config to generate images:
+
+```bash
+hcp_run --cfg cfgs/workflow/text2img.py
+```
+
+Or override parameters via command line:
+
+```bash
+hcp_run --cfg cfgs/workflow/text2img_cli.py \
+    pretrained_model=pretrained_model_path \
+    prompt='positive_prompt' \
+    negative_prompt='negative_prompt' \
+    seed=42
+```
+
+### Tutorials
+
+🚧 In Development
+
+---
+
+## Contributing
+
+We welcome contributions to support more models and features.
+
+---
+
+## Team
+
+Maintained by [HCP-Lab at Sun Yat-sen University](https://www.sysu-hcp.net/).
+
+---
+
+## Citation
+
+```bibtex
+@article{DBLP:journals/corr/abs-2211-11337,
+  author    = {Ziyi Dong and
+               Pengxu Wei and
+               Liang Lin},
+  title     = {DreamArtist: Towards Controllable One-Shot Text-to-Image Generation
+               via Positive-Negative Prompt-Tuning},
+  journal   = {CoRR},
+  volume    = {abs/2211.11337},
+  year      = {2022},
+  doi       = {10.48550/arXiv.2211.11337},
+  eprinttype = {arXiv},
+  eprint    = {2211.11337},
+}
+```

hcpdiff-2.1/hcpdiff/__init__.py

@@ -0,0 +1,4 @@
+#from .train_ac_old import Trainer
+#from .train_ac_single import TrainerSingleCard
+# from .visualizer import Visualizer
+# from .visualizer_reloadable import VisualizerReloadable

hcpdiff-2.1/hcpdiff/ckpt_manager/__init__.py

@@ -0,0 +1,4 @@
+from .format import EmbFormat, DiffusersSD15Format, DiffusersModelFormat, DiffusersSDXLFormat, DiffusersPixArtFormat, OfficialSDXLFormat, \
+    OfficialSD15Format
+from .ckpt import EmbSaver, easy_emb_saver
+from .loader import HCPLoraLoader

hcpdiff-2.1/hcpdiff/ckpt_manager/ckpt.py

@@ -0,0 +1,24 @@
+from rainbowneko.ckpt_manager import NekoSaver, CkptFormat, LocalCkptSource, PKLFormat
+from torch import nn
+from typing import Dict, Any
+
+class EmbSaver(NekoSaver):
+    def __init__(self, format: CkptFormat, source: LocalCkptSource, target_key='embs', prefix=None):
+        super().__init__(format, source)
+        self.target_key = target_key
+        self.prefix = prefix
+
+    def save_to(self, name, model: nn.Module, plugin_groups: Dict[str, Any], model_ema=None, exclude_key=None,
+                name_template=None):
+        train_pts = plugin_groups[self.target_key]
+        for pt_name, pt in train_pts.items():
+            self.save(pt_name, (pt_name, pt), prefix=self.prefix)
+            if name_template is not None:
+                pt_name = name_template.format(pt_name)
+                self.save(pt_name, (pt_name, pt), prefix=self.prefix)
+
+def easy_emb_saver():
+    return EmbSaver(
+        format=PKLFormat(),
+        source=LocalCkptSource(),
+    )

hcpdiff-2.1/hcpdiff/ckpt_manager/format/__init__.py

@@ -0,0 +1,4 @@
+from .emb import EmbFormat
+from .diffusers import DiffusersSD15Format, DiffusersModelFormat, DiffusersSDXLFormat, DiffusersPixArtFormat
+from .sd_single import OfficialSD15Format, OfficialSDXLFormat
+from .lora_webui import LoraWebuiFormat

hcpdiff-2.1/hcpdiff/ckpt_manager/format/diffusers.py

@@ -0,0 +1,59 @@
+import torch
+from diffusers import ModelMixin, AutoencoderKL, UNet2DConditionModel, PixArtTransformer2DModel
+from rainbowneko.ckpt_manager.format import CkptFormat
+from transformers import CLIPTextModel, AutoTokenizer, T5EncoderModel
+
+from hcpdiff.diffusion.sampler import DDPMSampler, DDPMDiscreteSigmaScheduler
+from hcpdiff.models.compose import SDXLTokenizer, SDXLTextEncoder
+
+class DiffusersModelFormat(CkptFormat):
+    def __init__(self, builder: ModelMixin):
+        self.builder = builder
+
+    def save_ckpt(self, sd_model: ModelMixin, save_f: str, **kwargs):
+        sd_model.save_pretrained(save_f)
+
+    def load_ckpt(self, ckpt_f: str, map_location="cpu", **kwargs):
+        self.builder.from_pretrained(ckpt_f, **kwargs)
+
+class DiffusersSD15Format(CkptFormat):
+    def load_ckpt(self, pretrained_model: str, map_location="cpu", denoiser=None, TE=None, vae: AutoencoderKL = None, noise_sampler=None,
+                  tokenizer=None, revision=None, dtype=torch.float32, **kwargs):
+        denoiser = denoiser or UNet2DConditionModel.from_pretrained(
+            pretrained_model, subfolder="unet", revision=revision, torch_dtype=dtype
+        )
+        vae = vae or AutoencoderKL.from_pretrained(pretrained_model, subfolder="vae", revision=revision, torch_dtype=dtype)
+        noise_sampler = noise_sampler or DDPMSampler(DDPMDiscreteSigmaScheduler())
+
+        TE = TE or CLIPTextModel.from_pretrained(pretrained_model, subfolder="text_encoder", revision=revision, torch_dtype=dtype)
+        tokenizer = tokenizer or AutoTokenizer.from_pretrained(pretrained_model, subfolder="tokenizer", revision=revision, use_fast=False)
+
+        return dict(denoiser=denoiser, TE=TE, vae=vae, noise_sampler=noise_sampler, tokenizer=tokenizer)
+
+class DiffusersSDXLFormat(CkptFormat):
+    def load_ckpt(self, pretrained_model: str, map_location="cpu", denoiser=None, TE=None, vae: AutoencoderKL = None, noise_sampler=None,
+                  tokenizer=None, revision=None, dtype=torch.float32, **kwargs):
+        denoiser = denoiser or UNet2DConditionModel.from_pretrained(
+            pretrained_model, subfolder="unet", revision=revision, torch_dtype=dtype
+        )
+        vae = vae or AutoencoderKL.from_pretrained(pretrained_model, subfolder="vae", revision=revision, torch_dtype=dtype)
+        noise_sampler = noise_sampler or DDPMSampler(DDPMDiscreteSigmaScheduler())
+
+        TE = TE or SDXLTextEncoder.from_pretrained(pretrained_model, subfolder="text_encoder", revision=revision, torch_dtype=dtype)
+        tokenizer = tokenizer or SDXLTokenizer.from_pretrained(pretrained_model, subfolder="tokenizer", revision=revision, use_fast=False)
+
+        return dict(denoiser=denoiser, TE=TE, vae=vae, noise_sampler=noise_sampler, tokenizer=tokenizer)
+
+class DiffusersPixArtFormat(CkptFormat):
+    def load_ckpt(self, pretrained_model: str, map_location="cpu", denoiser=None, TE=None, vae: AutoencoderKL = None, noise_sampler=None,
+                  tokenizer=None, revision=None, dtype=torch.float32, **kwargs):
+        denoiser = denoiser or PixArtTransformer2DModel.from_pretrained(
+            pretrained_model, subfolder="transformer", revision=revision, torch_dtype=dtype
+        )
+        vae = vae or AutoencoderKL.from_pretrained(pretrained_model, subfolder="vae", revision=revision, torch_dtype=dtype)
+        noise_sampler = noise_sampler or DDPMSampler(DDPMDiscreteSigmaScheduler())
+
+        TE = TE or T5EncoderModel.from_pretrained(pretrained_model, subfolder="text_encoder", revision=revision, torch_dtype=dtype)
+        tokenizer = tokenizer or AutoTokenizer.from_pretrained(pretrained_model, subfolder="tokenizer", revision=revision, use_fast=False)
+
+        return dict(denoiser=denoiser, TE=TE, vae=vae, noise_sampler=noise_sampler, tokenizer=tokenizer)

hcpdiff-2.1/hcpdiff/ckpt_manager/format/emb.py

@@ -0,0 +1,21 @@
+from typing import Tuple
+
+import torch
+from rainbowneko.ckpt_manager.format import CkptFormat
+from torch.serialization import FILE_LIKE
+
+class EmbFormat(CkptFormat):
+    EXT = 'pt'
+
+    def save_ckpt(self, sd_model: Tuple[str, torch.Tensor], save_f: FILE_LIKE):
+        name, emb = sd_model
+        torch.save({'string_to_param':{'*':emb}, 'name':name}, save_f)
+
+    def load_ckpt(self, ckpt_f: FILE_LIKE, map_location="cpu"):
+        state = torch.load(ckpt_f, map_location=map_location)
+        if 'string_to_param' in state:
+            emb = state['string_to_param']['*']
+        else:
+            emb = state['emb_params']
+        emb.requires_grad_(False)
+        return emb