Upload folder using huggingface_hub

.gitattributes

@@ -1,35 +1,61 @@
 *.7z filter=lfs diff=lfs merge=lfs -text
 *.arrow filter=lfs diff=lfs merge=lfs -text
 *.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
 *.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
 *.ftz filter=lfs diff=lfs merge=lfs -text
 *.gz filter=lfs diff=lfs merge=lfs -text
 *.h5 filter=lfs diff=lfs merge=lfs -text
 *.joblib filter=lfs diff=lfs merge=lfs -text
 *.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
 *.model filter=lfs diff=lfs merge=lfs -text
 *.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
 *.onnx filter=lfs diff=lfs merge=lfs -text
 *.ot filter=lfs diff=lfs merge=lfs -text
 *.parquet filter=lfs diff=lfs merge=lfs -text
 *.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
 *.pt filter=lfs diff=lfs merge=lfs -text
 *.pth filter=lfs diff=lfs merge=lfs -text
 *.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
 saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
 *.tflite filter=lfs diff=lfs merge=lfs -text
 *.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*.tfevents* filter=lfs diff=lfs merge=lfs -text
+*.db* filter=lfs diff=lfs merge=lfs -text
+*.ark* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.gguf* filter=lfs diff=lfs merge=lfs -text
+*.ggml filter=lfs diff=lfs merge=lfs -text
+*.llamafile* filter=lfs diff=lfs merge=lfs -text
+*.pt2 filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -textassets/extra_videos/video_1.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/extra_videos/video_2.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/extra_videos/video_3.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/image_2.jpg filter=lfs diff=lfs merge=lfs -text
+assets/image_3.jpg filter=lfs diff=lfs merge=lfs -text
+assets/video_1.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_10.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_2.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_3.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_4.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_5.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_6.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_7.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_8.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/video_9.mp4 filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,317 @@
+---
+license: apache-2.0
+---
+# MiniMax-H3 Templates: Text Embeddings
+
+This repository provides a collection of video effect templates for the [MiniMax-H3](https://modelscope.cn/models/MiniMax/MiniMax-H3) model. These templates encapsulate specific model capabilities as text embeddings, implemented via [Diffusion Templates](https://arxiv.org/abs/2604.24351).
+
+## Showcase
+
+### Effects Gallery
+
+| **art_is_explosion** | **storm_magic** |
+| --- | --- |
+| <video src="assets/video_5.mp4" controls muted loop></video> | <video src="assets/video_6.mp4" controls muted loop></video> |
+| **dark_magic** | **kiss_camera** |
+| <video src="assets/video_1.mp4" controls muted loop></video> | <video src="assets/video_2.mp4" controls muted loop></video> |
+| **bullet_time** | **truman_show** |
+| <video src="assets/video_3.mp4" controls muted loop></video> | <video src="assets/video_4.mp4" controls muted loop></video> |
+| **fire_breath** | **blooming_flowers** |
+| <video src="assets/video_7.mp4" controls muted loop></video> | <video src="assets/video_8.mp4" controls muted loop></video> |
+| **four_seasons** | **spiral_ascent** |
+| <video src="assets/video_9.mp4" controls muted loop></video> | <video src="assets/video_10.mp4" controls muted loop></video> |
+
+### Effect Combinations
+
+Combining effects: `art_is_explosion` + `storm_magic`
+
+| | | |
+| --- | --- | --- |
+| <video src="assets/extra_videos/video_1.mp4" autoplay muted loop></video> | <video src="assets/extra_videos/video_2.mp4" autoplay muted loop></video> | <video src="assets/extra_videos/video_3.mp4" autoplay muted loop></video> |
+
+## How It Works
+
+Each template consists of a single tensor. This tensor can either replace or be combined with the output of the Text Encoder, a mechanism very similar to [Textual Inversion](https://arxiv.org/abs/2208.01618).
+
+While MiniMax-H3 boasts powerful base capabilities, its massive parameter count makes LoRA training challenging. Text Embeddings offer a lightweight alternative to LoRA with the following advantages:
+
+*   **Modular Capabilities:** Text Embeddings serve as carriers of model functionality. Similar to agent skills, they enable specific capabilities to be saved, reused, and distributed as atomic units through model platforms.
+*   **Flexible Initialization:** Embeddings can be initialized from text prompts, images, or videos. Simply pass the input data through the MiniMax-H3 Text Encoder to quickly generate a usable Text Embedding.
+*   **Trainable:** Like LoRA, Text Embeddings support end-to-end training on video datasets, enabling targeted enhancement of specific generative abilities.
+*   **Composable:** Multiple templates can be combined to jointly influence generation, allowing you to create complex and visually striking effects.
+
+## Inference and Training
+
+### Installation
+
+Install [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio):
+
+```shell
+git clone https://github.com/modelscope/DiffSynth-Studio.git
+cd DiffSynth-Studio
+pip install -e ".[all]"
+```
+
+### Initialize a Text Embedding via Text Encoder
+
+The following code has low VRAM requirements and can run with as little as 6GB of VRAM:
+
+```python
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from safetensors.torch import save_file
+```
+
+```python
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[ModelConfig(
+        model_id="MiniMax/MiniMax-H3",
+        origin_file_pattern="FL2VA/text_encoder/model*.safetensors",
+        offload_dtype="disk",
+        offload_device="disk",
+        onload_dtype="disk",
+        onload_device="disk",
+        preparing_dtype=torch.bfloat16,
+        preparing_device="cuda",
+        computation_dtype=torch.bfloat16,
+        computation_device="cuda",
+    )],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=0,
+)
+prompt = "xxx"
+text_embedding = pipe.export_text_embedding(prompt)
+save_file({"weight": text_embedding}, "model.safetensors")
+```
+
+### Inference with Text Embeddings via Diffusion Templates
+
+You can load text embeddings through [Diffusion Templates](https://arxiv.org/abs/2604.24351) for inference. Optionally, load an acceleration LoRA to speed up inference (note: the text encoder is not required during this stage):
+
+```python
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from diffsynth.diffusion.template import TemplatePipeline
+from diffsynth.utils.data.audio_video import write_video_audio
+from diffsynth.core.data.operators import ImageCropAndResize
+from modelscope import snapshot_download
+from PIL import Image
+```
+
+```python
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="minimax-h3-fl2va-pruned-nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="video_vae_nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="audio_vae_nf4.safetensors", **vram_config),
+    ],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 2,
+)
+pipe.load_lora(
+    pipe.dit,
+    ModelConfig(
+        model_id="lightx2v/Minimax-h3-Turbo",
+        origin_file_pattern="minimax_h3_fl2v_turbo_4step_v1.0_768p_bf16.safetensors",
+    ),
+)
+template = TemplatePipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[ModelConfig(
+        model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings", origin_file_pattern="models/art_is_explosion/",
+    )],
+)
+snapshot_download("DiffSynth-Studio/MiniMax-H3-Text-Embeddings", allow_file_pattern="assets/image_1.jpg", local_dir="data")
+first_frame = ImageCropAndResize(height=1344, width=768)(Image.open("data/assets/image_1.jpg"))
+video, audio = template(
+    pipe,
+    height=1344, width=768, num_frames=56,
+    num_inference_steps=4, seed=0, flow_shift=6,
+    keyframes=[first_frame], keyframe_indices=[0],
+    template_inputs=[{}],
+)
+write_video_audio(
+    video=video, audio=audio,
+    output_path="output.mp4", fps=24, audio_sample_rate=32000,
+)
+```
+
+<details>
+<summary>Inference with Multiple Text Embeddings</summary>
+
+```diff
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from diffsynth.diffusion.template import TemplatePipeline
+from diffsynth.utils.data.audio_video import write_video_audio
+from diffsynth.core.data.operators import ImageCropAndResize
+from modelscope import snapshot_download
+from PIL import Image
+
+```python
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="minimax-h3-fl2va-pruned-nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="video_vae_nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="audio_vae_nf4.safetensors", **vram_config),
+    ],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 2,
+)
+pipe.load_lora(
+    pipe.dit,
+    ModelConfig(
+        model_id="lightx2v/Minimax-h3-Turbo",
+        origin_file_pattern="minimax_h3_fl2v_turbo_4step_v1.0_768p_bf16.safetensors",
+    ),
+)
+template = TemplatePipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings", origin_file_pattern="models/art_is_explosion/"),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings", origin_file_pattern="models/storm_magic/"),
+    ],
+)
+snapshot_download("DiffSynth-Studio/MiniMax-H3-Text-Embeddings", allow_file_pattern="assets/image_1.jpg", local_dir="data")
+first_frame = ImageCropAndResize(height=1344, width=768)(Image.open("data/assets/image_1.jpg"))
+video, audio = template(
+    pipe,
+    height=1344, width=768, num_frames=56,
+    num_inference_steps=4, seed=0, flow_shift=6,
+    keyframes=[first_frame], keyframe_indices=[0],
+    template_inputs=[{"model_id": 0}, {"model_id": 1}],
+)
+write_video_audio(
+    video=video, audio=audio,
+    output_path="output.mp4", fps=24, audio_sample_rate=32000,
+)
+```
+
+</details>
+
+<details>
+<summary>Inference Using the Native Pipeline</summary>
+
+```python
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from diffsynth.utils.data.audio_video import write_video_audio
+from diffsynth.core.data.operators import ImageCropAndResize
+from diffsynth import load_state_dict
+from modelscope import snapshot_download
+from PIL import Image
+
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="minimax-h3-fl2va-pruned-nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="video_vae_nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="audio_vae_nf4.safetensors", **vram_config),
+    ],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 2,
+)
+pipe.load_lora(
+    pipe.dit,
+    ModelConfig(
+        model_id="lightx2v/Minimax-h3-Turbo",
+        origin_file_pattern="minimax_h3_fl2v_turbo_4step_v1.0_768p_bf16.safetensors",
+    ),
+)
+text_embedding_config = ModelConfig(
+    model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings",
+    origin_file_pattern="models/art_is_explosion/model.safetensors",
+)
+text_embedding_config.download_if_necessary()
+text_embedding = load_state_dict(text_embedding_config.path)["weight"]
+```
+
+```python
+snapshot_download("DiffSynth-Studio/MiniMax-H3-Text-Embeddings", allow_file_pattern="assets/image_1.jpg", local_dir="data")
+first_frame = ImageCropAndResize(height=1344, width=768)(Image.open("data/assets/image_1.jpg"))
+video, audio = pipe(
+    height=1344, width=768, num_frames=56,
+    num_inference_steps=4, seed=0, flow_shift=6,
+    keyframes=[first_frame], keyframe_indices=[0],
+    text_embedding=text_embedding,
+)
+write_video_audio(
+    video=video, audio=audio,
+    output_path="output.mp4", fps=24, audio_sample_rate=32000,
+)
+```
+
+</details>
+
+### Training Text Embeddings
+
+```shell
+modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "minimax_h3/MiniMax-H3-Text-Embeddings/*" --local_dir ./data/diffsynth_example_dataset
+
+accelerate launch examples/minimax_h3/model_training/train.py \
+  --dataset_base_path data/diffsynth_example_dataset/minimax_h3/MiniMax-H3-Text-Embeddings \
+  --dataset_metadata_path data/diffsynth_example_dataset/minimax_h3/MiniMax-H3-Text-Embeddings/metadata.json \
+  --data_file_keys "video,input_audio" \
+  --extra_inputs "input_audio,input_image,template_inputs" \
+  --height 832 \
+  --width 480 \
+  --num_frames 124 \
+  --dataset_repeat 100 \
+  --model_id_with_origin_paths "DiffSynth-Studio/MiniMax-H3-NF4:video_vae_nf4.safetensors,DiffSynth-Studio/MiniMax-H3-NF4:audio_vae_nf4.safetensors,DiffSynth-Studio/MiniMax-H3-NF4:minimax-h3-fl2va-pruned-nf4.safetensors" \
+  --template_model_id_or_path "DiffSynth-Studio/MiniMax-H3-Text-Embeddings:models/art_is_explosion/" \
+  --learning_rate 1e-4 \
+  --num_epochs 2 \
+  --remove_prefix_in_ckpt "pipe.template_model." \
+  --output_path "./models/train/MiniMax-H3-Text-Embeddings-full" \
+  --trainable_models "template_model" \
+  --use_gradient_checkpointing
+```

README_from_modelscope.md

@@ -0,0 +1,317 @@
+---
+base_model:
+- ""
+frameworks:
+- ""
+license: Apache License 2.0
+tags: []
+tasks:
+- text-to-video-synthesis
+---
+# MiniMax-H3 Templates: Text Embeddings
+
+This repository provides a collection of video effect templates for the [MiniMax-H3](https://modelscope.cn/models/MiniMax/MiniMax-H3) model. These templates encapsulate specific model capabilities as text embeddings, implemented via [Diffusion Templates](https://arxiv.org/abs/2604.24351).
+
+## Showcase
+
+### Effects Gallery
+
+| **art_is_explosion** | **storm_magic** |
+| --- | --- |
+| <video src="assets/video_5.mp4" controls muted loop></video> | <video src="assets/video_6.mp4" controls muted loop></video> |
+| **dark_magic** | **kiss_camera** |
+| <video src="assets/video_1.mp4" controls muted loop></video> | <video src="assets/video_2.mp4" controls muted loop></video> |
+| **bullet_time** | **truman_show** |
+| <video src="assets/video_3.mp4" controls muted loop></video> | <video src="assets/video_4.mp4" controls muted loop></video> |
+| **fire_breath** | **blooming_flowers** |
+| <video src="assets/video_7.mp4" controls muted loop></video> | <video src="assets/video_8.mp4" controls muted loop></video> |
+| **four_seasons** | **spiral_ascent** |
+| <video src="assets/video_9.mp4" controls muted loop></video> | <video src="assets/video_10.mp4" controls muted loop></video> |
+
+### Effect Combinations
+
+Combining effects: `art_is_explosion` + `storm_magic`
+
+| | | |
+| --- | --- | --- |
+| <video src="assets/extra_videos/video_1.mp4" autoplay muted loop></video> | <video src="assets/extra_videos/video_2.mp4" autoplay muted loop></video> | <video src="assets/extra_videos/video_3.mp4" autoplay muted loop></video> |
+
+## How It Works
+
+Each template consists of a single tensor. This tensor can either replace or be combined with the output of the Text Encoder, a mechanism very similar to [Textual Inversion](https://arxiv.org/abs/2208.01618).
+
+While MiniMax-H3 boasts powerful base capabilities, its massive parameter count makes LoRA training challenging. Text Embeddings offer a lightweight alternative to LoRA with the following advantages:
+
+*   **Modular Capabilities:** Text Embeddings act as carriers of model functionality. Like agent skills, they allow specific capabilities to be saved, reused, and distributed as atomic units via model platforms.
+*   **Flexible Initialization:** Embeddings can be initialized from text prompts, images, or videos. Simply process the input data through the MiniMax-H3 Text Encoder to quickly generate a usable Text Embedding.
+*   **Trainable:** Like LoRA, Text Embeddings support end-to-end training on video datasets, enabling targeted enhancement of specific generative capabilities.
+*   **Composable:** Multiple templates can be combined to jointly influence generation, allowing you to create complex and stunning visual effects.
+
+## Inference and Training
+
+### Installation
+
+Install [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio):
+
+```shell
+git clone https://github.com/modelscope/DiffSynth-Studio.git
+cd DiffSynth-Studio
+pip install -e ".[all]"
+```
+
+### Initialize a Text Embedding via Text Encoder
+
+The following code has minimal VRAM requirements and can run with as little as 6GB of VRAM:
+
+```python
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from safetensors.torch import save_file
+
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[ModelConfig(
+        model_id="MiniMax/MiniMax-H3",
+        origin_file_pattern="FL2VA/text_encoder/model*.safetensors",
+        offload_dtype="disk",
+        offload_device="disk",
+        onload_dtype="disk",
+        onload_device="disk",
+        preparing_dtype=torch.bfloat16,
+        preparing_device="cuda",
+        computation_dtype=torch.bfloat16,
+        computation_device="cuda",
+    )],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=0,
+)
+prompt = "xxx"
+text_embedding = pipe.export_text_embedding(prompt)
+save_file({"weight": text_embedding}, "model.safetensors")
+```
+
+### Inference with Text Embeddings via Diffusion Templates
+
+You can load Text Embeddings through [Diffusion Templates](https://arxiv.org/abs/2604.24351) for inference. Optionally, load an acceleration LoRA to speed up inference (note: the Text Encoder is not required during this stage):
+
+```python
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from diffsynth.diffusion.template import TemplatePipeline
+from diffsynth.utils.data.audio_video import write_video_audio
+from diffsynth.core.data.operators import ImageCropAndResize
+from modelscope import snapshot_download
+from PIL import Image
+
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="minimax-h3-fl2va-pruned-nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="video_vae_nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="audio_vae_nf4.safetensors", **vram_config),
+    ],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 2,
+)
+pipe.load_lora(
+    pipe.dit,
+    ModelConfig(
+        model_id="lightx2v/Minimax-h3-Turbo",
+        origin_file_pattern="minimax_h3_fl2v_turbo_4step_v1.0_768p_bf16.safetensors",
+    ),
+)
+template = TemplatePipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[ModelConfig(
+        model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings", origin_file_pattern="models/art_is_explosion/",
+    )],
+)
+snapshot_download("DiffSynth-Studio/MiniMax-H3-Text-Embeddings", allow_file_pattern="assets/image_1.jpg", local_dir="data")
+first_frame = ImageCropAndResize(height=1344, width=768)(Image.open("data/assets/image_1.jpg"))
+video, audio = template(
+    pipe,
+    height=1344, width=768, num_frames=56,
+    num_inference_steps=4, seed=0, flow_shift=6,
+    keyframes=[first_frame], keyframe_indices=[0],
+    template_inputs=[{}],
+)
+write_video_audio(
+    video=video, audio=audio,
+    output_path="output.mp4", fps=24, audio_sample_rate=32000,
+)
+```
+
+<details>
+<summary>Inference with Multiple Text Embeddings</summary>
+
+```diff
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from diffsynth.diffusion.template import TemplatePipeline
+from diffsynth.utils.data.audio_video import write_video_audio
+from diffsynth.core.data.operators import ImageCropAndResize
+from modelscope import snapshot_download
+from PIL import Image
+
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="minimax-h3-fl2va-pruned-nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="video_vae_nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="audio_vae_nf4.safetensors", **vram_config),
+    ],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 2,
+)
+pipe.load_lora(
+    pipe.dit,
+    ModelConfig(
+        model_id="lightx2v/Minimax-h3-Turbo",
+        origin_file_pattern="minimax_h3_fl2v_turbo_4step_v1.0_768p_bf16.safetensors",
+    ),
+)
+template = TemplatePipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
++       ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings", origin_file_pattern="models/art_is_explosion/"),
++       ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings", origin_file_pattern="models/storm_magic/"),
+    ],
+)
+snapshot_download("DiffSynth-Studio/MiniMax-H3-Text-Embeddings", allow_file_pattern="assets/image_1.jpg", local_dir="data")
+first_frame = ImageCropAndResize(height=1344, width=768)(Image.open("data/assets/image_1.jpg"))
+video, audio = template(
+    pipe,
+    height=1344, width=768, num_frames=56,
+    num_inference_steps=4, seed=0, flow_shift=6,
+    keyframes=[first_frame], keyframe_indices=[0],
++   template_inputs=[{"model_id": 0}, {"model_id": 1}],
+)
+write_video_audio(
+    video=video, audio=audio,
+    output_path="output.mp4", fps=24, audio_sample_rate=32000,
+)
+```
+
+</details>
+
+<details>
+<summary>Inference Using the Native Pipeline</summary>
+
+```python
+import torch
+from diffsynth.pipelines.minimax_h3_audio_video import MiniMaxH3Pipeline, ModelConfig
+from diffsynth.utils.data.audio_video import write_video_audio
+from diffsynth.core.data.operators import ImageCropAndResize
+from diffsynth import load_state_dict
+from modelscope import snapshot_download
+from PIL import Image
+
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": "disk",
+    "onload_device": "disk",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+pipe = MiniMaxH3Pipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="minimax-h3-fl2va-pruned-nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="video_vae_nf4.safetensors", **vram_config),
+        ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-NF4", origin_file_pattern="audio_vae_nf4.safetensors", **vram_config),
+    ],
+    processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3", origin_file_pattern="FL2VA/processor/"),
+    vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 2,
+)
+pipe.load_lora(
+    pipe.dit,
+    ModelConfig(
+        model_id="lightx2v/Minimax-h3-Turbo",
+        origin_file_pattern="minimax_h3_fl2v_turbo_4step_v1.0_768p_bf16.safetensors",
+    ),
+)
+text_embedding_config = ModelConfig(
+    model_id="DiffSynth-Studio/MiniMax-H3-Text-Embeddings",
+    origin_file_pattern="models/art_is_explosion/model.safetensors",
+)
+text_embedding_config.download_if_necessary()
+text_embedding = load_state_dict(text_embedding_config.path)["weight"]
+
+snapshot_download("DiffSynth-Studio/MiniMax-H3-Text-Embeddings", allow_file_pattern="assets/image_1.jpg", local_dir="data")
+first_frame = ImageCropAndResize(height=1344, width=768)(Image.open("data/assets/image_1.jpg"))
+video, audio = pipe(
+    height=1344, width=768, num_frames=56,
+    num_inference_steps=4, seed=0, flow_shift=6,
+    keyframes=[first_frame], keyframe_indices=[0],
+    text_embedding=text_embedding,
+)
+write_video_audio(
+    video=video, audio=audio,
+    output_path="output.mp4", fps=24, audio_sample_rate=32000,
+)
+```
+
+</details>
+
+### Training Text Embeddings
+
+```shell
+modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "minimax_h3/MiniMax-H3-Text-Embeddings/*" --local_dir ./data/diffsynth_example_dataset
+
+accelerate launch examples/minimax_h3/model_training/train.py \
+  --dataset_base_path data/diffsynth_example_dataset/minimax_h3/MiniMax-H3-Text-Embeddings \
+  --dataset_metadata_path data/diffsynth_example_dataset/minimax_h3/MiniMax-H3-Text-Embeddings/metadata.json \
+  --data_file_keys "video,input_audio" \
+  --extra_inputs "input_audio,input_image,template_inputs" \
+  --height 832 \
+  --width 480 \
+  --num_frames 124 \
+  --dataset_repeat 100 \
+  --model_id_with_origin_paths "DiffSynth-Studio/MiniMax-H3-NF4:video_vae_nf4.safetensors,DiffSynth-Studio/MiniMax-H3-NF4:audio_vae_nf4.safetensors,DiffSynth-Studio/MiniMax-H3-NF4:minimax-h3-fl2va-pruned-nf4.safetensors" \
+  --template_model_id_or_path "DiffSynth-Studio/MiniMax-H3-Text-Embeddings:models/art_is_explosion/" \
+  --learning_rate 1e-4 \
+  --num_epochs 2 \
+  --remove_prefix_in_ckpt "pipe.template_model." \
+  --output_path "./models/train/MiniMax-H3-Text-Embeddings-full" \
+  --trainable_models "template_model" \
+  --use_gradient_checkpointing
+```

assets/extra_videos/video_1.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acd448bac5d7d5157dbcd5b09ebdc16757e684dfc8846e6a35eab1927338cd69
+size 4859895

assets/extra_videos/video_2.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68965c8e2e61841334e736ca5ae53ed9d33481d4adf3320063f3fc9b294a5723
+size 6404799

assets/extra_videos/video_3.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:908fce3f04397db745ee507c20b01a69033df200335b72d143b94cdc102085c4
+size 4654019

assets/prompts.txt

@@ -0,0 +1,10 @@
+* 黑暗魔法：画面中的生物闭上眼睛，再次猛然睁开眼睛，眼睛已变成血红色，神情变得狰狞，生物面向镜头，张开双臂，掌心出现红色能量光球，镜头拉远，背后出现顺时针旋转的红色炫光六芒星魔法阵
+* 亲吻镜头：镜头固定不动，画面中的生物缓步向镜头靠近，眼神温柔缱绻、满含爱意，嘴角带着浅浅的笑意；行至镜头前时微微倾身，双唇轻柔地贴上屏幕表面，做出亲吻的动作，镜头随即切换为嘴唇极致特写，清晰呈现唇瓣的纹理、水润光泽与亲吻时的细微形变，背景完全虚化，仅保留唇部与屏幕接触的瞬间细节
+* 子弹时间：画面进入子弹时间效果，所有动态元素瞬间凝固静止，所有物体定格于空中；镜头以画面主体为中心，保持恒定高度与距离，迅速向右匀速平稳地旋转一圈，360度展现主体在时间静止状态下的立体形态与空间关系，背景中的物体保持不变，画面主体位于镜头中心，光影随镜头移动在主体表面流转，强化凝固瞬间的张力与细节质感
+* 楚门的世界：镜头从画面中央的表演区域缓缓匀速拉远，视野逐渐扩展，依次显露出环绕的摄影机、轨道车、灯光支架与绿幕边界；随着景别扩大，剧组工作人员纷纷入画，有人调整设备参数、有人整理道具线缆、有人低声沟通走位，所有人各司其职、动作紧凑有序，现场呈现出真实而繁忙的拍摄氛围，最终全景展现完整片场的工作生态
+* 艺术就是爆炸：画面背景中一颗原子弹爆炸，一朵巨大的蘑菇云伴随着火光升起，光芒耀眼，云层被染成昏黄色。爆炸产生的气浪掀起身后的物体。主角面带微笑，面对镜头，丝毫察觉不到背后的爆炸
+* 暴风魔法：画面中的主角神情凝肃，双手于胸前缓缓合十，指尖相触的刹那泛起青白色微光；随着掌心向外轻推，无形气流骤然凝聚成可见的风旋，自其周身呈环形爆发扩散；狂风呼啸席卷四野，草木齐齐伏倒、尘土飞扬如雾，衣袍猎猎翻飞、发丝狂舞不息，空气中浮现出半透明的风刃轨迹与涡流纹路；镜头保持中景稳定跟拍，清晰捕捉风魔法释放的瞬间张力与环境剧烈响应，营造出磅礴而灵动的自然之力氛围
+* 火焰魔法：画面中的人物深吸一口气，胸腔微微鼓起，随即双唇张开，一股炽热火焰自口中喷涌而出，火舌呈锥形向前延展，边缘泛着金红光泽、核心透出亮白高温；火焰随呼吸节奏起伏波动，照亮人物面部轮廓并投下跃动光影，周围空气因热浪产生轻微扭曲；镜头保持正面中景稳定拍摄，清晰捕捉喷火瞬间的口型变化、火焰形态及环境光效反馈，确保安全距离内呈现真实而富有张力的视觉冲击
+* 百花齐放：主角脚下渐变为鲜嫩绿草地，无数花苞自草根处破土而出，以肉眼可见的速度抽茎、展叶、膨大；顷刻间繁花竞相盛放，红、黄、蓝、紫、粉等各色花瓣层层舒展，花蕊沾着晨露熠熠生辉，色彩饱和而和谐；镜头保持中远景固定视角，完整记录花朵从萌发到怒放的全过程，光影随花影摇曳流转，营造出生机盎然、如梦似幻的自然奇景，主角面向镜头微笑
+* 四季流传：画面中的场景以加速流转的方式经历季节更迭：春日的嫩绿迅速褪为夏末的浓荫，继而染上秋霜的金黄与赭红，最终在数秒内被凛冽寒风席卷，草木凋零、枝干枯寂；漫天飞雪自灰白云层倾泻而下，积雪层层覆盖大地、屋檐与远山，水面凝结成镜，万物裹上素白冰壳；镜头保持全景稳定俯拍，清晰呈现四季压缩转换的节奏感，最终定格于银装素裹、万籁俱寂的冬日旷野，寒气仿佛穿透屏幕弥漫开来
+* 螺旋升天：画面中的主角双臂平展向两侧伸直，躯干与手臂构成标准的T字形姿态，神情沉静而坚定；随即身体以纵轴为心逆时针缓缓旋转，双脚离地轻盈升空，衣袂与发丝随气流向上飘拂；旋转速度由缓渐稳，高度持续攀升，穿过云层间隙，阳光自上方洒落勾勒出其轮廓光晕；镜头采用仰角跟随拍摄，保持主角始终居于画面中央，完整记录从地面起升至高空翱翔的动态轨迹，营造出庄严而自由的升腾仪式感

assets/video_1.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7542a035ef04cf9fead1cce3462cb21399289186fbadd1208e9d88d1d3f3492f
+size 2219174

assets/video_10.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92f7beeb3c946dedf584564089aafb093683f14936627daeeb7c8b81766c4456
+size 2495318

assets/video_2.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df9dd87b5684fc82242bcc5f0b0096c9cf791d0ac77d85c771053ce25891aa60
+size 1702337

assets/video_3.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e70a0a2b7268c185489f2ce0bc05bda84e0be4df72d8d02e28ef54386bbdb8d
+size 2104382

assets/video_4.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:895aa6da18aeaf82f2125111a499296dd65146fa0e0934ce72a1cfda352968fc
+size 2817731

assets/video_5.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38a77f47457b69690910a511e605a16d5a6a35d874c429b029b15de73c60418f
+size 1688071

assets/video_6.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dece0ae35b490b45dec83ac4893242e7bbb584c2e46811d1d1357ce03d0468c0
+size 2128251

assets/video_7.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c32cc8de566541f8a9319fe7c9cee57392059af9bba3991db96b5f21b6994966
+size 1427509

assets/video_8.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcb7b4777b581e8c38e1cabfc4561cf766e0161d261533218deb407246c3afe9
+size 1754576

assets/video_9.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bf0a3451df79eced41d191fc739178e5133ed168ddd7f7954e39c613604d938
+size 1608223

configuration.json

@@ -0,0 +1 @@
+{"task":"text-to-video-synthesis"}

models/art_is_explosion/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/art_is_explosion/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d42cf66406894c257bdf087e16ff0d70976a5336329b3d20fcb9138a6805971
+size 512080

models/blooming_flowers/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/blooming_flowers/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a8420ee3a580530d203f0d960b432654a44195aeb11f908cfcf75f5d1ccd7ce
+size 1259608

models/bullet_time/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/bullet_time/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6b3e0d4a3c906e610b73b34a511714af868d064df495022059e3e9985e810c4
+size 962640

models/dark_magic/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/dark_magic/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92fa5866e07904d26c6ccb1e61aa27e095f9169fabc7e6ca13f7638c17c912f1
+size 604240

models/fire_breath/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/fire_breath/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39d1d9668c9e88c45cf6ab583be55204fa7ec8b3de5a44613685a9e73daf1ecb
+size 1208408

models/four_seasons/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/four_seasons/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c470ad894f635c679800651df47327b926864126da936c6e5a5ee5da20107871
+size 1454168

models/kiss_camera/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/kiss_camera/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b001199c4734a61e995230d176021d7590aad97e59eebe8e3b69dac1e1e2c6d
+size 993360

models/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/spiral_ascent/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/spiral_ascent/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bc34d411da2839b9218b3eb6775b87f2c4bb1e62a17b644d3dba6fee3263469
+size 1341528

models/storm_magic/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/storm_magic/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ed04e01376a04a8328335e6653e18226cdbc7861462a2318a0920335fcee004
+size 1402968

models/truman_show/model.py

@@ -0,0 +1,28 @@
+import torch, os
+from safetensors import safe_open
+
+
+class EmbeddingModel(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        model_path = os.path.join(os.path.dirname(__file__), TEMPLATE_MODEL_PATH)
+        with safe_open(model_path, framework="pt", device="cpu") as f:
+            shape = f.get_tensor("weight").shape
+        self.weight = torch.nn.Parameter(torch.empty(shape))
+        
+    @torch.no_grad()
+    def process_inputs(self, **kwargs):
+        return {}
+
+    def forward(self, **kwargs):
+        return {"text_embedding": self.weight}
+
+
+class DataProcessor:
+    def __call__(self, **kwargs):
+        return kwargs
+
+
+TEMPLATE_MODEL = EmbeddingModel
+TEMPLATE_MODEL_PATH = "model.safetensors"
+TEMPLATE_DATA_PROCESSOR = DataProcessor

models/truman_show/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:263e4493d0b5b133f2bad558821bcb827a01bb42d0252df39dbf54ab35bf316f
+size 921680