kelseye commited on
Commit
320bbfa
·
verified ·
1 Parent(s): 4527646

Upload model using ModelScope SDK

Browse files
.gitattributes CHANGED
@@ -1,35 +1,57 @@
1
  *.7z filter=lfs diff=lfs merge=lfs -text
2
  *.arrow filter=lfs diff=lfs merge=lfs -text
3
  *.bin filter=lfs diff=lfs merge=lfs -text
 
4
  *.bz2 filter=lfs diff=lfs merge=lfs -text
5
- *.ckpt filter=lfs diff=lfs merge=lfs -text
6
  *.ftz filter=lfs diff=lfs merge=lfs -text
7
  *.gz filter=lfs diff=lfs merge=lfs -text
8
  *.h5 filter=lfs diff=lfs merge=lfs -text
9
  *.joblib filter=lfs diff=lfs merge=lfs -text
10
  *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
- *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
  *.model filter=lfs diff=lfs merge=lfs -text
13
  *.msgpack filter=lfs diff=lfs merge=lfs -text
14
- *.npy filter=lfs diff=lfs merge=lfs -text
15
- *.npz filter=lfs diff=lfs merge=lfs -text
16
  *.onnx filter=lfs diff=lfs merge=lfs -text
17
  *.ot filter=lfs diff=lfs merge=lfs -text
18
  *.parquet filter=lfs diff=lfs merge=lfs -text
19
  *.pb filter=lfs diff=lfs merge=lfs -text
20
- *.pickle filter=lfs diff=lfs merge=lfs -text
21
- *.pkl filter=lfs diff=lfs merge=lfs -text
22
  *.pt filter=lfs diff=lfs merge=lfs -text
23
  *.pth filter=lfs diff=lfs merge=lfs -text
24
  *.rar filter=lfs diff=lfs merge=lfs -text
25
- *.safetensors filter=lfs diff=lfs merge=lfs -text
26
  saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
  *.tar.* filter=lfs diff=lfs merge=lfs -text
28
- *.tar filter=lfs diff=lfs merge=lfs -text
29
  *.tflite filter=lfs diff=lfs merge=lfs -text
30
  *.tgz filter=lfs diff=lfs merge=lfs -text
31
- *.wasm filter=lfs diff=lfs merge=lfs -text
32
  *.xz filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
- *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
1
  *.7z filter=lfs diff=lfs merge=lfs -text
2
  *.arrow filter=lfs diff=lfs merge=lfs -text
3
  *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bin.* filter=lfs diff=lfs merge=lfs -text
5
  *.bz2 filter=lfs diff=lfs merge=lfs -text
 
6
  *.ftz filter=lfs diff=lfs merge=lfs -text
7
  *.gz filter=lfs diff=lfs merge=lfs -text
8
  *.h5 filter=lfs diff=lfs merge=lfs -text
9
  *.joblib filter=lfs diff=lfs merge=lfs -text
10
  *.lfs.* filter=lfs diff=lfs merge=lfs -text
 
11
  *.model filter=lfs diff=lfs merge=lfs -text
12
  *.msgpack filter=lfs diff=lfs merge=lfs -text
 
 
13
  *.onnx filter=lfs diff=lfs merge=lfs -text
14
  *.ot filter=lfs diff=lfs merge=lfs -text
15
  *.parquet filter=lfs diff=lfs merge=lfs -text
16
  *.pb filter=lfs diff=lfs merge=lfs -text
 
 
17
  *.pt filter=lfs diff=lfs merge=lfs -text
18
  *.pth filter=lfs diff=lfs merge=lfs -text
19
  *.rar filter=lfs diff=lfs merge=lfs -text
 
20
  saved_model/**/* filter=lfs diff=lfs merge=lfs -text
21
  *.tar.* filter=lfs diff=lfs merge=lfs -text
 
22
  *.tflite filter=lfs diff=lfs merge=lfs -text
23
  *.tgz filter=lfs diff=lfs merge=lfs -text
 
24
  *.xz filter=lfs diff=lfs merge=lfs -text
25
  *.zip filter=lfs diff=lfs merge=lfs -text
26
+ *.zstandard filter=lfs diff=lfs merge=lfs -text
27
+ *.tfevents* filter=lfs diff=lfs merge=lfs -text
28
+ *.db* filter=lfs diff=lfs merge=lfs -text
29
+ *.ark* filter=lfs diff=lfs merge=lfs -text
30
+ **/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
31
+ **/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
32
+ **/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
33
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
34
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
35
+ *.gguf* filter=lfs diff=lfs merge=lfs -text
36
+ *.ggml filter=lfs diff=lfs merge=lfs -text
37
+ *.llamafile* filter=lfs diff=lfs merge=lfs -text
38
+ *.pt2 filter=lfs diff=lfs merge=lfs -text
39
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
40
+ *.npy filter=lfs diff=lfs merge=lfs -text
41
+ *.npz filter=lfs diff=lfs merge=lfs -text
42
+ *.pickle filter=lfs diff=lfs merge=lfs -text
43
+ *.pkl filter=lfs diff=lfs merge=lfs -text
44
+ *.tar filter=lfs diff=lfs merge=lfs -text
45
+ *.wasm filter=lfs diff=lfs merge=lfs -text
46
  *.zst filter=lfs diff=lfs merge=lfs -text
47
+ *tfevents* filter=lfs diff=lfs merge=lfs -textassets/bf16.jpg filter=lfs diff=lfs merge=lfs -text
48
+ assets/bf16_pruned.jpg filter=lfs diff=lfs merge=lfs -text
49
+ assets/dit_4p5bit.jpg filter=lfs diff=lfs merge=lfs -text
50
+ assets/dit_4p5bit_int8.jpg filter=lfs diff=lfs merge=lfs -text
51
+ assets/dit_6bit.jpg filter=lfs diff=lfs merge=lfs -text
52
+ assets/dit_6bit_int8.jpg filter=lfs diff=lfs merge=lfs -text
53
+ assets/dit_7bit.jpg filter=lfs diff=lfs merge=lfs -text
54
+ assets/dit_7bit_int8.jpg filter=lfs diff=lfs merge=lfs -text
55
+ assets/dit_8bit.jpg filter=lfs diff=lfs merge=lfs -text
56
+ assets/dit_8bit_int8.jpg filter=lfs diff=lfs merge=lfs -text
57
+ assets/dit_extreme_3bpp.jpg filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,216 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: MiniMax/MiniMax-H3-Image
3
+ base_model_relation: quantized
4
+ ---
5
+
6
+ # MiniMax-H3-Image (AdaLN-pruned) — EntroPack Pre-Quantized Models
7
+
8
+ We provide pre-quantized weight packages produced by [**EntroPack**](https://github.com/modelscope/entropack) for the AdaLN-pruned **MiniMax-H3-Image** transformer ([MiniMax/MiniMax-H3-Image](https://www.modelscope.cn/models/MiniMax/MiniMax-H3-Image)), which can be used with the quantization features of [**DiffSynth-Studio**](https://github.com/modelscope/DiffSynth-Studio). Use them with ComfyUI through [DiffSynth-ComfyUI](https://github.com/modelscope/DiffSynth-ComfyUI). A ready-to-run ComfyUI workflow is included in [`comfyui/featured_minimax_h3_image_entropack.json`](comfyui/featured_minimax_h3_image_entropack.json).
9
+
10
+ The pruned transformer replaces every AdaLN projection and the timestep embedder with lookup tables for the six distilled sampling steps: 20.1B parameters instead of 33.1B, and 40.2 GB instead of 66.3 GB in bfloat16. **This replacement is completely lossless: every lookup-table row is precomputed from the original weights, so the pruned transformer produces exactly the same output as the unpruned checkpoint at the same prompt and seed. It is inference-only and applies only to the distilled six-step schedule.**
11
+
12
+ The packages run with as little as 24 GB of VRAM.
13
+
14
+ > **On a 24 GB GPU, use `dit_7bit` or `dit_7bit_int8`; on a 12 GB or smaller GPU, use `dit_4p5bit` or `dit_4p5bit_int8`.**
15
+
16
+ The DiT is provided at five bitrate levels:
17
+
18
+ * **`8bit` / `7bit` / `6bit`** — uniform entropy coding, with the quantization-sensitive layers held at 8.5 bpp;
19
+ * **`4.5bit`** — the same mixed allocation, with the main blocks at the bitrate derived from the 4.5 bpp weighted target;
20
+ * **`extreme_3bpp`** — extreme quantization: the same mixed allocation with the main blocks at 2.91 bpp, the two input projections at 8.5 bpp and the token-refiner blocks at 5.0 bpp;
21
+ * **`int8_*`** — weights are first **INT8 W8A8-quantized** with convrot, then stored by EntroPack; the `int8_8bit` package stores the INT8 codes verbatim, the others entropy-code them to the level's bitrate. The forward pass uses **INT8 GEMM**, which delivers a faster inference speed than the BF16 GEMM used by the other levels. The INT8 GEMM path requires Triton and a CUDA GPU with compute capability 8.0 or newer.
22
+
23
+ The text encoder is not re-quantized for this repository: reuse `models/MiniMax-H3/text_encoder_3bpp.safetensors` from [DiffSynth-Studio/EntroPackPreQuants](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants), the package shared by every MiniMax-H3 model (video, audio and image).
24
+
25
+ The image VAE is provided as an INT8 package (`models/vae_int8.safetensors`): the ViT decoder's linear layers use INT8 GEMM with convrot, while the convolutional encoder and the normalization layers stay in bfloat16.
26
+
27
+ Every `*.safetensors` ships with a same-named `*.json` that records the quantization config. `actual bpp = 8 × Σ storage bytes / Σ numel` is measured on the package artifacts.
28
+
29
+ ## Preview
30
+
31
+ All previews are generated with the same prompt, seed 42 and size 4096×4096; `bf16` is the released transformer and `bf16_pruned` the AdaLN-pruned transformer, both in full precision.
32
+
33
+ <table width="100%">
34
+ <tr>
35
+ <td width="50%" align="center">bf16 (reference)</td>
36
+ <td width="50%" align="center">bf16 pruned</td>
37
+ </tr>
38
+ <tr>
39
+ <td align="center"><img width="100%" src="assets/bf16.jpg"></td>
40
+ <td align="center"><img width="100%" src="assets/bf16_pruned.jpg"></td>
41
+ </tr>
42
+ <tr>
43
+ <td width="50%" align="center">dit_8bit</td>
44
+ <td width="50%" align="center">dit_8bit_int8</td>
45
+ </tr>
46
+ <tr>
47
+ <td align="center"><img width="100%" src="assets/dit_8bit.jpg"></td>
48
+ <td align="center"><img width="100%" src="assets/dit_8bit_int8.jpg"></td>
49
+ </tr>
50
+ <tr>
51
+ <td width="50%" align="center">dit_7bit</td>
52
+ <td width="50%" align="center">dit_7bit_int8</td>
53
+ </tr>
54
+ <tr>
55
+ <td align="center"><img width="100%" src="assets/dit_7bit.jpg"></td>
56
+ <td align="center"><img width="100%" src="assets/dit_7bit_int8.jpg"></td>
57
+ </tr>
58
+ <tr>
59
+ <td width="50%" align="center">dit_6bit</td>
60
+ <td width="50%" align="center">dit_6bit_int8</td>
61
+ </tr>
62
+ <tr>
63
+ <td align="center"><img width="100%" src="assets/dit_6bit.jpg"></td>
64
+ <td align="center"><img width="100%" src="assets/dit_6bit_int8.jpg"></td>
65
+ </tr>
66
+ <tr>
67
+ <td width="50%" align="center">dit_4p5bit</td>
68
+ <td width="50%" align="center">dit_4p5bit_int8</td>
69
+ </tr>
70
+ <tr>
71
+ <td align="center"><img width="100%" src="assets/dit_4p5bit.jpg"></td>
72
+ <td align="center"><img width="100%" src="assets/dit_4p5bit_int8.jpg"></td>
73
+ </tr>
74
+ <tr>
75
+ <td width="50%" align="center">dit_extreme_3bpp (extreme quantization)</td>
76
+ <td width="50%"></td>
77
+ </tr>
78
+ <tr>
79
+ <td align="center"><img width="100%" src="assets/dit_extreme_3bpp.jpg"></td>
80
+ <td width="50%"></td>
81
+ </tr>
82
+ </table>
83
+
84
+ The two full-precision rows are identical.
85
+
86
+ ## DiT
87
+
88
+ | name | file | actual bpp | size |
89
+ |---|---|---|---|
90
+ | bf16_pruned (full precision) | [`models/bf16_table_pruned.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fbf16_table_pruned.safetensors) | 16.028 | 40.20 GB |
91
+ | dit_8bit | [`models/dit_8bit.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_8bit.safetensors) | 8.107 | 20.34 GB |
92
+ | dit_7bit | [`models/dit_7bit.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_7bit.safetensors) | 7.134 | 17.89 GB |
93
+ | dit_6bit | [`models/dit_6bit.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_6bit.safetensors) | 6.119 | 15.35 GB |
94
+ | dit_4p5bit | [`models/dit_4p5bit.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_4p5bit.safetensors) | 4.587 | 11.51 GB |
95
+ | dit_8bit_int8 (INT8 GEMM) | [`models/dit_8bit_int8.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_8bit_int8.safetensors) | 8.054 | 20.20 GB |
96
+ | dit_7bit_int8 (INT8 GEMM) | [`models/dit_7bit_int8.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_7bit_int8.safetensors) | 7.076 | 17.75 GB |
97
+ | dit_6bit_int8 (INT8 GEMM) | [`models/dit_6bit_int8.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_6bit_int8.safetensors) | 6.072 | 15.23 GB |
98
+ | dit_4p5bit_int8 (INT8 GEMM) | [`models/dit_4p5bit_int8.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_4p5bit_int8.safetensors) | 4.568 | 11.46 GB |
99
+ | dit_extreme_3bpp (extreme quantization) | [`models/dit_extreme_3bpp.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fdit_extreme_3bpp.safetensors) | 3.057 | 7.67 GB |
100
+
101
+ ## Text Encoder
102
+
103
+ | name | file | actual bpp | size |
104
+ |---|---|---|---|
105
+ | text_encoder_3bpp | [`DiffSynth-Studio/EntroPackPreQuants:models/MiniMax-H3/text_encoder_3bpp.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants/file/view/master/models%2FMiniMax-H3%2Ftext_encoder_3bpp.safetensors) | 3.531 | 11.02 GB |
106
+ | text_encoder_4bpp | [`DiffSynth-Studio/EntroPackPreQuants:models/MiniMax-H3/text_encoder_4bpp.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants/file/view/master/models%2FMiniMax-H3%2Ftext_encoder_4bpp.safetensors) | 4.541 | 14.17 GB |
107
+ | text_encoder_5bpp | [`DiffSynth-Studio/EntroPackPreQuants:models/MiniMax-H3/text_encoder_5bpp.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants/file/view/master/models%2FMiniMax-H3%2Ftext_encoder_5bpp.safetensors) | 5.575 | 17.40 GB |
108
+ | text_encoder_6bpp | [`DiffSynth-Studio/EntroPackPreQuants:models/MiniMax-H3/text_encoder_6bpp.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants/file/view/master/models%2FMiniMax-H3%2Ftext_encoder_6bpp.safetensors) | 6.604 | 20.61 GB |
109
+ | text_encoder_7bpp | [`DiffSynth-Studio/EntroPackPreQuants:models/MiniMax-H3/text_encoder_7bpp.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants/file/view/master/models%2FMiniMax-H3%2Ftext_encoder_7bpp.safetensors) | 7.629 | 23.81 GB |
110
+ | text_encoder_8bpp | [`DiffSynth-Studio/EntroPackPreQuants:models/MiniMax-H3/text_encoder_8bpp.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants/file/view/master/models%2FMiniMax-H3%2Ftext_encoder_8bpp.safetensors) | 8.597 | 26.83 GB |
111
+ | text_encoder_int8_convrot (INT8 GEMM) | [`DiffSynth-Studio/EntroPackPreQuants:models/MiniMax-H3/text_encoder_int8_convrot.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/EntroPackPreQuants/file/view/master/models%2FMiniMax-H3%2Ftext_encoder_int8_convrot.safetensors) | 8.506 | 26.55 GB |
112
+
113
+ ## VAE
114
+
115
+ | name | file | actual bpp | size |
116
+ |---|---|---|---|
117
+ | vae_int8 (INT8 GEMM) | [`models/vae_int8.safetensors`](https://www.modelscope.cn/models/DiffSynth-Studio/MiniMax-H3-Image-EntroPack/file/view/master/models%2Fvae_int8.safetensors) | 8.035 | 0.77 GB |
118
+
119
+ The 98 linear layers of the ViT decoder are INT8 with convrot; the convolutional encoder and the normalization layers stay in bfloat16.
120
+
121
+ ## Usage
122
+
123
+ ### Install
124
+
125
+ ```bash
126
+ # DiffSynth-Studio
127
+ git clone https://github.com/modelscope/DiffSynth-Studio.git
128
+ cd DiffSynth-Studio && pip install -e .
129
+
130
+ # EntroPack (pick the extra matching your CUDA version: cuda13 / cuda12)
131
+ git clone https://github.com/modelscope/entropack.git
132
+ cd entropack && pip install -e '.[cuda13]'
133
+ ```
134
+
135
+ If EntroPack is already installed but its CUDA backend cannot start, the error names the missing piece, for example:
136
+
137
+ ```
138
+ entropack.registry.DispatchError: No backend can handle 'lattice_rans': cuda: cupy is not installed (No module named 'cupy');
139
+ install a matching build, e.g. `pip install cupy-cuda13x` for CUDA 13 or `pip install cupy-cuda12x` for CUDA 12
140
+ ```
141
+
142
+ Install the CuPy build that matches the CUDA major version of your PyTorch (`pip install cupy-cuda13x` for CUDA 13, `pip install cupy-cuda12x` for CUDA 12), or reinstall EntroPack with the matching extra (`pip install "entropack[cuda13]"` / `pip install "entropack[cuda12]"`).
143
+
144
+ ### ComfyUI
145
+
146
+ The packages are also available as ComfyUI nodes through [DiffSynth-ComfyUI](https://github.com/modelscope/DiffSynth-ComfyUI). A ready-to-run workflow is shipped as [`comfyui/featured_minimax_h3_image_entropack.json`](comfyui/featured_minimax_h3_image_entropack.json): it loads the 7-bit INT8 DiT, the 7bpp text encoder and the INT8 VAE, and generates a 4096×4096 image in six distilled steps. Load the file in ComfyUI and queue it; the models are downloaded and registered from their json sidecars automatically.
147
+
148
+ ### Example scripts
149
+
150
+ * Inference: [standard](https://github.com/modelscope/DiffSynth-Studio/blob/main/examples/minimax_h3_image/model_inference/MiniMax-H3-Image-Pruned-EntroPack.py) / [low VRAM](https://github.com/modelscope/DiffSynth-Studio/blob/main/examples/minimax_h3_image/model_inference_low_vram/MiniMax-H3-Image-Pruned-EntroPack.py)
151
+ * LoRA training: [train](https://github.com/modelscope/DiffSynth-Studio/blob/main/examples/minimax_h3_image/model_training/lora/MiniMax-H3-Image-EntroPack.sh) / [validation](https://github.com/modelscope/DiffSynth-Studio/blob/main/examples/minimax_h3_image/model_training/validate_lora/MiniMax-H3-Image-EntroPack.py)
152
+
153
+ `dit_package` and `text_encoder_package` are chosen independently: any combination works (e.g. `dit_int8_8bit` with `text_encoder_4bpp`). The 7-bit DiT and the 4-bit text encoder are pre-registered in DiffSynth-Studio, so they load with a plain `ModelConfig`; the helper below registers any package's json sidecar before loading:
154
+
155
+ ```python
156
+ import json
157
+
158
+ import torch
159
+ from PIL import Image
160
+
161
+ from diffsynth.configs import MODEL_CONFIGS
162
+ from diffsynth.pipelines.minimax_h3_image import MiniMaxH3ImagePipeline, ModelConfig
163
+
164
+ dit_model_id = "DiffSynth-Studio/MiniMax-H3-Image-EntroPack"
165
+ text_encoder_model_id = "DiffSynth-Studio/EntroPackPreQuants"
166
+ dit_package = "models/dit_7bit.safetensors"
167
+ text_encoder_package = "models/MiniMax-H3/text_encoder_3bpp.safetensors"
168
+
169
+
170
+ def register_prequantized(model_id, package_pattern):
171
+ package = ModelConfig(model_id=model_id, origin_file_pattern=package_pattern)
172
+ package.download_if_necessary()
173
+ sidecar = ModelConfig(model_id=model_id, origin_file_pattern=package_pattern.replace(".safetensors", ".json"))
174
+ sidecar.download_if_necessary()
175
+ with open(sidecar.path) as handle:
176
+ entry = json.load(handle)
177
+ if not any(config.get("model_hash") == entry["model_hash"] for config in MODEL_CONFIGS):
178
+ MODEL_CONFIGS.append(entry)
179
+ return package
180
+
181
+
182
+ pipe = MiniMaxH3ImagePipeline.from_pretrained(
183
+ torch_dtype=torch.bfloat16,
184
+ device="cuda",
185
+ model_configs=[
186
+ register_prequantized(dit_model_id, dit_package),
187
+ register_prequantized(text_encoder_model_id, text_encoder_package),
188
+ ModelConfig(model_id="MiniMax/MiniMax-H3-Image", origin_file_pattern="pdd_fused_heads.safetensors"),
189
+ ModelConfig(model_id="DiffSynth-Studio/MiniMax-H3-Image-EntroPack", origin_file_pattern="models/vae_int8.safetensors"),
190
+ ],
191
+ processor_config=ModelConfig(model_id="MiniMax/MiniMax-H3-Image", origin_file_pattern="processor/"),
192
+ )
193
+
194
+ prompt = (
195
+ "Full-color cinematic outdoor night photograph of a stunning young Asian woman in her early twenties with a"
196
+ "natural warm ivory complexion and a healthy glow, wearing a soft blush pink off-shoulder dress with a fitted"
197
+ "bodice, framed from the waist up against softly illuminated historic architecture. She stands near the"
198
+ "centre of the frame, facing the camera with a gentle closed-lip smile, her hands resting loosely clasped in"
199
+ "front of her waist with relaxed elbows; she has a slim oval face with softly defined cheekbones and a"
200
+ "delicate narrow jaw, dark expressive eyes, natural-shaped brows, and subtle natural makeup with soft red"
201
+ "lips. Her long black hair falls in loose natural waves over her shoulders, with fine strands framing her"
202
+ "face. The fabric of the dress falls in soft folds across her shoulders and chest. In the frame-left"
203
+ "background, the word \"EntroPack\" glows as a hand-drawn bubbly yellow neon sign mounted on a softly floodlit"
204
+ "dark stone building above the street; rows of warm-lit windows, stone cornices and shopfronts remain clearly"
205
+ "readable around the sign in gentle amber light, while the thick rounded irregular letters ride a gently"
206
+ "bouncing baseline with a bright core and a soft outer glow, kept sharp and unblurred against the night. The"
207
+ "background recedes into deep blue tones: the Eiffel Tower rises above distant buildings on the frame-right"
208
+ "as a softly silhouetted tapering lattice outline in the shallow depth of field, while small amber, muted"
209
+ "turquoise and rose-coloured lights dissolve into rounded bokeh across the upper background, leaving a darker"
210
+ "area behind her head. Soft neutral frontal illumination with gentle falloff reveals her face with natural"
211
+ "warm skin tone and subtle texture, a faint cooler rim separates her hair from the night, and the skirt of"
212
+ "her dress fills the lower edge of the frame."
213
+ )
214
+ image = pipe(prompt=prompt, height=4096, width=4096, seed=42)
215
+ image.save("image.jpg")
216
+ ```
assets/bf16.jpg ADDED

Git LFS Details

  • SHA256: 779c91d2923d8ee4176633c0820d861525f0430e8e08237b4023896e80b46b0a
  • Pointer size: 132 Bytes
  • Size of remote file: 2.92 MB
assets/bf16_pruned.jpg ADDED

Git LFS Details

  • SHA256: 779c91d2923d8ee4176633c0820d861525f0430e8e08237b4023896e80b46b0a
  • Pointer size: 132 Bytes
  • Size of remote file: 2.92 MB
assets/dit_4p5bit.jpg ADDED

Git LFS Details

  • SHA256: e0fc6a605d92b21782f9e58108c526f56114dfac1a03902a2aa69ae8950d91a2
  • Pointer size: 132 Bytes
  • Size of remote file: 2.92 MB
assets/dit_4p5bit_int8.jpg ADDED

Git LFS Details

  • SHA256: 010217af3f75a58b151bc9dba89336957b49a2da1035b374513b40225394702b
  • Pointer size: 132 Bytes
  • Size of remote file: 2.95 MB
assets/dit_6bit.jpg ADDED

Git LFS Details

  • SHA256: 0f5d6d801a8c7ab7b9ee1bf3f2e149e3249e58ec2d25136c8a94cc1722be7a2f
  • Pointer size: 132 Bytes
  • Size of remote file: 2.95 MB
assets/dit_6bit_int8.jpg ADDED

Git LFS Details

  • SHA256: d72e2e10f5677b0e9cd151ca7854e7ac9eada9a4bb8adb8c6faa3c654dca0728
  • Pointer size: 132 Bytes
  • Size of remote file: 3.05 MB
assets/dit_7bit.jpg ADDED

Git LFS Details

  • SHA256: 551942b98374f5cc0ccda41d1a208e9edc0ccf9e7009204b7f41689cc20a2fb2
  • Pointer size: 132 Bytes
  • Size of remote file: 2.86 MB
assets/dit_7bit_int8.jpg ADDED

Git LFS Details

  • SHA256: 0aee2d3c91301bd477e8a24c0973cec4da5e00374a5b719214c8d2b8ed6db7d8
  • Pointer size: 132 Bytes
  • Size of remote file: 2.95 MB
assets/dit_8bit.jpg ADDED

Git LFS Details

  • SHA256: dcb65a129a6c7f5a0ef44cd534777d8cf7c276fd0d321a3cceb94377078f973b
  • Pointer size: 132 Bytes
  • Size of remote file: 2.92 MB
assets/dit_8bit_int8.jpg ADDED

Git LFS Details

  • SHA256: cf3f0f1cf64b6fe798a0a340b3ba22d1d8655112819470ac4ec6eb94295c889f
  • Pointer size: 132 Bytes
  • Size of remote file: 2.95 MB
assets/dit_extreme_3bpp.jpg ADDED

Git LFS Details

  • SHA256: 7159b75463df43202949b8369aca62545f5ddbde7d65fa92432fa62a591e8171
  • Pointer size: 132 Bytes
  • Size of remote file: 2.78 MB
comfyui/featured_minimax_h3_image_entropack.json ADDED
@@ -0,0 +1,296 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "last_node_id": 11,
3
+ "last_link_id": 13,
4
+ "nodes": [
5
+ {
6
+ "id": 1,
7
+ "type": "DiffSynthVRAMConfig",
8
+ "pos": [20, 20],
9
+ "size": {"0": 280, "1": 180},
10
+ "flags": {},
11
+ "order": 0,
12
+ "mode": 0,
13
+ "inputs": [],
14
+ "outputs": [
15
+ {
16
+ "name": "vram_config",
17
+ "type": "DIFFSYNTH_VRAM_CONFIG",
18
+ "links": [1, 3, 5, 7],
19
+ "slot_index": 0
20
+ }
21
+ ],
22
+ "properties": {"Node name for S&R": "DiffSynthVRAMConfig"},
23
+ "widgets_values": ["cpu", "bfloat16", "cpu", "bfloat16", "cuda", "bfloat16", "cuda", "bfloat16"]
24
+ },
25
+ {
26
+ "id": 2,
27
+ "type": "DiffSynthModelConfig",
28
+ "pos": [420, 20],
29
+ "size": {"0": 320, "1": 220},
30
+ "flags": {},
31
+ "order": 1,
32
+ "mode": 0,
33
+ "inputs": [
34
+ {"name": "vram_config", "type": "DIFFSYNTH_VRAM_CONFIG", "link": 1},
35
+ {"name": "quant_config", "type": "DIFFSYNTH_QUANT_CONFIG", "link": null}
36
+ ],
37
+ "outputs": [
38
+ {
39
+ "name": "model_config",
40
+ "type": "DIFFSYNTH_MODEL_CONFIG",
41
+ "links": [2],
42
+ "slot_index": 0
43
+ }
44
+ ],
45
+ "properties": {"Node name for S&R": "DiffSynthModelConfig"},
46
+ "widgets_values": ["DiffSynth-Studio/EntroPackPreQuants", "models/MiniMax-H3/text_encoder_7bpp.safetensors", "", "modelscope", true],
47
+ "title": "Text Encoder (EntroPack 7bpp)"
48
+ },
49
+ {
50
+ "id": 3,
51
+ "type": "DiffSynthModelConfig",
52
+ "pos": [420, 300],
53
+ "size": {"0": 320, "1": 220},
54
+ "flags": {},
55
+ "order": 2,
56
+ "mode": 0,
57
+ "inputs": [
58
+ {"name": "vram_config", "type": "DIFFSYNTH_VRAM_CONFIG", "link": 3},
59
+ {"name": "quant_config", "type": "DIFFSYNTH_QUANT_CONFIG", "link": null}
60
+ ],
61
+ "outputs": [
62
+ {
63
+ "name": "model_config",
64
+ "type": "DIFFSYNTH_MODEL_CONFIG",
65
+ "links": [4],
66
+ "slot_index": 0
67
+ }
68
+ ],
69
+ "properties": {"Node name for S&R": "DiffSynthModelConfig"},
70
+ "widgets_values": ["DiffSynth-Studio/MiniMax-H3-Image-EntroPack", "models/dit_7bit_int8.safetensors", "", "modelscope", true],
71
+ "title": "DiT (EntroPack 7bit INT8)"
72
+ },
73
+ {
74
+ "id": 4,
75
+ "type": "DiffSynthModelConfig",
76
+ "pos": [420, 580],
77
+ "size": {"0": 320, "1": 150},
78
+ "flags": {},
79
+ "order": 3,
80
+ "mode": 0,
81
+ "inputs": [
82
+ {"name": "vram_config", "type": "DIFFSYNTH_VRAM_CONFIG", "link": 5},
83
+ {"name": "quant_config", "type": "DIFFSYNTH_QUANT_CONFIG", "link": null}
84
+ ],
85
+ "outputs": [
86
+ {
87
+ "name": "model_config",
88
+ "type": "DIFFSYNTH_MODEL_CONFIG",
89
+ "links": [6],
90
+ "slot_index": 0
91
+ }
92
+ ],
93
+ "properties": {"Node name for S&R": "DiffSynthModelConfig"},
94
+ "widgets_values": ["MiniMax/MiniMax-H3-Image", "pdd_fused_heads.safetensors"],
95
+ "title": "PDD Heads"
96
+ },
97
+ {
98
+ "id": 5,
99
+ "type": "DiffSynthMergeModelConfigs",
100
+ "pos": [800, 20],
101
+ "size": {"0": 280, "1": 150},
102
+ "flags": {},
103
+ "order": 7,
104
+ "mode": 0,
105
+ "inputs": [
106
+ {"name": "model_config_1", "type": "DIFFSYNTH_MODEL_CONFIG", "link": 2},
107
+ {"name": "model_config_2", "type": "DIFFSYNTH_MODEL_CONFIG", "link": 4},
108
+ {"name": "model_config_3", "type": "DIFFSYNTH_MODEL_CONFIG", "link": 6},
109
+ {"name": "model_config_4", "type": "DIFFSYNTH_MODEL_CONFIG", "link": 8},
110
+ {"name": "model_config_5", "type": "DIFFSYNTH_MODEL_CONFIG", "link": null},
111
+ {"name": "model_config_6", "type": "DIFFSYNTH_MODEL_CONFIG", "link": null},
112
+ {"name": "model_config_7", "type": "DIFFSYNTH_MODEL_CONFIG", "link": null},
113
+ {"name": "model_config_8", "type": "DIFFSYNTH_MODEL_CONFIG", "link": null}
114
+ ],
115
+ "outputs": [
116
+ {
117
+ "name": "model_configs",
118
+ "type": "DIFFSYNTH_MODEL_CONFIG_LIST",
119
+ "links": [10],
120
+ "slot_index": 0
121
+ }
122
+ ],
123
+ "properties": {"Node name for S&R": "DiffSynthMergeModelConfigs"},
124
+ "widgets_values": []
125
+ },
126
+ {
127
+ "id": 6,
128
+ "type": "DiffSynthModelConfig",
129
+ "pos": [1140, 20],
130
+ "size": {"0": 320, "1": 150},
131
+ "flags": {},
132
+ "order": 4,
133
+ "mode": 0,
134
+ "inputs": [
135
+ {"name": "vram_config", "type": "DIFFSYNTH_VRAM_CONFIG", "link": null},
136
+ {"name": "quant_config", "type": "DIFFSYNTH_QUANT_CONFIG", "link": null}
137
+ ],
138
+ "outputs": [
139
+ {
140
+ "name": "model_config",
141
+ "type": "DIFFSYNTH_MODEL_CONFIG",
142
+ "links": [9],
143
+ "slot_index": 0
144
+ }
145
+ ],
146
+ "properties": {"Node name for S&R": "DiffSynthModelConfig"},
147
+ "widgets_values": ["MiniMax/MiniMax-H3-Image", "processor/"],
148
+ "title": "Processor / Tokenizer"
149
+ },
150
+ {
151
+ "id": 7,
152
+ "type": "DiffSynthModelConfig",
153
+ "pos": [420, 790],
154
+ "size": {"0": 320, "1": 150},
155
+ "flags": {},
156
+ "order": 5,
157
+ "mode": 0,
158
+ "inputs": [
159
+ {"name": "vram_config", "type": "DIFFSYNTH_VRAM_CONFIG", "link": 7},
160
+ {"name": "quant_config", "type": "DIFFSYNTH_QUANT_CONFIG", "link": null}
161
+ ],
162
+ "outputs": [
163
+ {
164
+ "name": "model_config",
165
+ "type": "DIFFSYNTH_MODEL_CONFIG",
166
+ "links": [8],
167
+ "slot_index": 0
168
+ }
169
+ ],
170
+ "properties": {"Node name for S&R": "DiffSynthModelConfig"},
171
+ "widgets_values": ["DiffSynth-Studio/MiniMax-H3-Image-EntroPack", "models/vae_int8.safetensors", "", "modelscope", true],
172
+ "title": "VAE"
173
+ },
174
+ {
175
+ "id": 8,
176
+ "type": "DiffSynthVRAMLimit",
177
+ "pos": [20, 300],
178
+ "size": {"0": 240, "1": 100},
179
+ "flags": {},
180
+ "order": 6,
181
+ "mode": 0,
182
+ "inputs": [],
183
+ "outputs": [
184
+ {
185
+ "name": "vram_limit",
186
+ "type": "DIFFSYNTH_VRAM_LIMIT",
187
+ "links": [11],
188
+ "slot_index": 0
189
+ }
190
+ ],
191
+ "properties": {"Node name for S&R": "DiffSynthVRAMLimit"},
192
+ "widgets_values": ["cuda", 2.0]
193
+ },
194
+ {
195
+ "id": 9,
196
+ "type": "DiffSynthMiniMaxH3ImageLoader",
197
+ "pos": [1560, 20],
198
+ "size": {"0": 340, "1": 200},
199
+ "flags": {},
200
+ "order": 8,
201
+ "mode": 0,
202
+ "inputs": [
203
+ {"name": "model_configs", "type": "DIFFSYNTH_MODEL_CONFIG_LIST", "link": 10},
204
+ {"name": "processor_config", "type": "DIFFSYNTH_MODEL_CONFIG", "link": 9},
205
+ {"name": "vram_limit", "type": "DIFFSYNTH_VRAM_LIMIT", "link": 11}
206
+ ],
207
+ "outputs": [
208
+ {
209
+ "name": "pipe",
210
+ "type": "DIFFSYNTH_PIPE",
211
+ "links": [12],
212
+ "slot_index": 0
213
+ }
214
+ ],
215
+ "properties": {"Node name for S&R": "DiffSynthMiniMaxH3ImageLoader"},
216
+ "widgets_values": ["bfloat16", "cuda"]
217
+ },
218
+ {
219
+ "id": 10,
220
+ "type": "DiffSynthMiniMaxH3ImageInference",
221
+ "pos": [2000, 20],
222
+ "size": {"0": 380, "1": 600},
223
+ "flags": {},
224
+ "order": 9,
225
+ "mode": 0,
226
+ "inputs": [
227
+ {"name": "pipe", "type": "DIFFSYNTH_PIPE", "link": 12},
228
+ {"name": "edit_image", "type": "IMAGE", "link": null},
229
+ {"name": "latents", "type": "*", "link": null},
230
+ {"name": "reference_noise", "type": "STRING", "link": null}
231
+ ],
232
+ "outputs": [
233
+ {
234
+ "name": "image",
235
+ "type": "IMAGE",
236
+ "links": [13],
237
+ "slot_index": 0
238
+ }
239
+ ],
240
+ "properties": {"Node name for S&R": "DiffSynthMiniMaxH3ImageInference"},
241
+ "widgets_values": [
242
+ "An overhead photograph of a rustic farm-to-table spread on a weathered wooden table: a round loaf of sourdough on burlap, wedges of cheese on a dark board, a copper bowl of vegetables, vine tomatoes, olives and walnuts, warm afternoon light raking across the grain of the wood.",
243
+ 42,
244
+ "fixed",
245
+ 6,
246
+ 4096,
247
+ 4096,
248
+ "",
249
+ 1.0,
250
+ 4096,
251
+ 4194304,
252
+ true,
253
+ true
254
+ ]
255
+ },
256
+ {
257
+ "id": 11,
258
+ "type": "SaveImage",
259
+ "pos": [2440, 20],
260
+ "size": {"0": 300, "1": 140},
261
+ "flags": {},
262
+ "order": 10,
263
+ "mode": 0,
264
+ "inputs": [
265
+ {"name": "images", "type": "IMAGE", "link": 13}
266
+ ],
267
+ "outputs": [],
268
+ "properties": {"Node name for S&R": "SaveImage"},
269
+ "widgets_values": ["DiffSynth/featured_minimax_h3_image_entropack"]
270
+ }
271
+ ],
272
+ "links": [
273
+ [1, 1, 0, 2, 0, "DIFFSYNTH_VRAM_CONFIG"],
274
+ [2, 2, 0, 5, 0, "DIFFSYNTH_MODEL_CONFIG"],
275
+ [3, 1, 0, 3, 0, "DIFFSYNTH_VRAM_CONFIG"],
276
+ [4, 3, 0, 5, 1, "DIFFSYNTH_MODEL_CONFIG"],
277
+ [5, 1, 0, 4, 0, "DIFFSYNTH_VRAM_CONFIG"],
278
+ [6, 4, 0, 5, 2, "DIFFSYNTH_MODEL_CONFIG"],
279
+ [7, 1, 0, 7, 0, "DIFFSYNTH_VRAM_CONFIG"],
280
+ [8, 7, 0, 5, 3, "DIFFSYNTH_MODEL_CONFIG"],
281
+ [9, 6, 0, 9, 1, "DIFFSYNTH_MODEL_CONFIG"],
282
+ [10, 5, 0, 9, 0, "DIFFSYNTH_MODEL_CONFIG_LIST"],
283
+ [11, 8, 0, 9, 2, "DIFFSYNTH_VRAM_LIMIT"],
284
+ [12, 9, 0, 10, 0, "DIFFSYNTH_PIPE"],
285
+ [13, 10, 0, 11, 0, "IMAGE"]
286
+ ],
287
+ "groups": [
288
+ {"title": "VRAM/Quantization Config", "bounding": [0, -70, 320, 490], "color": "#88aACC", "font_size": 24, "flags": {}},
289
+ {"title": "Model Config", "bounding": [400, -70, 1080, 1030], "color": "#CC88AA", "font_size": 24, "flags": {}},
290
+ {"title": "Loader", "bounding": [1540, -70, 380, 310], "color": "#aACC88", "font_size": 24, "flags": {}},
291
+ {"title": "Inference", "bounding": [1980, -70, 780, 710], "color": "#CCAA88", "font_size": 24, "flags": {}}
292
+ ],
293
+ "config": {},
294
+ "extra": {},
295
+ "version": 0.4
296
+ }
config.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "MiniMaxH3DiTModel",
3
+ "_diffusers_version": "0.32.2",
4
+ "hidden_size": 5376,
5
+ "num_layers": 50,
6
+ "token_refiner_num_layers": 2,
7
+ "num_attention_heads": 56,
8
+ "attention_head_dim": 128,
9
+ "ffn_hidden_size": 14336,
10
+ "latents_dim": 24,
11
+ "audio_latents_dim": 32,
12
+ "patch_size": [
13
+ 1,
14
+ 2,
15
+ 2
16
+ ],
17
+ "text_dim": 5120,
18
+ "timestep_input_dim": 256,
19
+ "time_embed_hidden_size": 5376,
20
+ "time_embed_dim": 2688,
21
+ "adaln_out_features": 96768,
22
+ "final_adaln_out_features": 10752,
23
+ "rope_inv_freq_len": 16,
24
+ "norm_eps": 1e-05,
25
+ "qk_norm_eps": 1e-05,
26
+ "final_norm_eps": 1e-05,
27
+ "audio_enabled": false
28
+ }
configuration.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"task":"text-to-image-synthesis"}
models/bf16_table_pruned.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ffa988f6c5596f28490be6b57c8f346bc4e31af136e65f2aa4a6ab1dcc00218
3
+ size 40202825664
models/dit_4p5bit.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "4c0400645cb029731d24e93c4c84705d",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 4.3341785714285725,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20
14
+ },
15
+ "exclude_modules": [
16
+ "video_patch_proj",
17
+ "condition_proj",
18
+ "token_refiner.blocks.0.attn.qkv_proj",
19
+ "token_refiner.blocks.0.attn.out_proj",
20
+ "token_refiner.blocks.0.mlp.fc1",
21
+ "token_refiner.blocks.0.mlp.fc2",
22
+ "token_refiner.blocks.1.attn.qkv_proj",
23
+ "token_refiner.blocks.1.attn.out_proj",
24
+ "token_refiner.blocks.1.mlp.fc1",
25
+ "token_refiner.blocks.1.mlp.fc2"
26
+ ]
27
+ },
28
+ {
29
+ "method": "entropack_lossy_quant",
30
+ "backend_config_kwargs": {
31
+ "target_bpp": 8.5,
32
+ "row_rdo_iterations": 2,
33
+ "row_rdo_candidates": 20
34
+ },
35
+ "target_modules": [
36
+ "video_patch_proj",
37
+ "condition_proj",
38
+ "token_refiner.blocks.0.attn.qkv_proj",
39
+ "token_refiner.blocks.0.attn.out_proj",
40
+ "token_refiner.blocks.0.mlp.fc1",
41
+ "token_refiner.blocks.0.mlp.fc2",
42
+ "token_refiner.blocks.1.attn.qkv_proj",
43
+ "token_refiner.blocks.1.attn.out_proj",
44
+ "token_refiner.blocks.1.mlp.fc1",
45
+ "token_refiner.blocks.1.mlp.fc2"
46
+ ]
47
+ }
48
+ ]
49
+ }
50
+ }
models/dit_4p5bit.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f53920785ed7fc54b39f018eafca38c3891fcc2877a97a3f0be4ea5fa724060c
3
+ size 11506687152
models/dit_4p5bit_int8.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "f6b2aaaadaa64d954c739f14f70fa16e",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant_int8",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 4.3341785714285725,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20,
14
+ "use_convrot": true,
15
+ "convrot_groupsize": 256
16
+ },
17
+ "exclude_modules": [
18
+ "video_patch_proj",
19
+ "condition_proj",
20
+ "token_refiner.blocks.0.attn.qkv_proj",
21
+ "token_refiner.blocks.0.attn.out_proj",
22
+ "token_refiner.blocks.0.mlp.fc1",
23
+ "token_refiner.blocks.0.mlp.fc2",
24
+ "token_refiner.blocks.1.attn.qkv_proj",
25
+ "token_refiner.blocks.1.attn.out_proj",
26
+ "token_refiner.blocks.1.mlp.fc1",
27
+ "token_refiner.blocks.1.mlp.fc2"
28
+ ]
29
+ },
30
+ {
31
+ "method": "entropack_lossy_quant",
32
+ "backend_config_kwargs": {
33
+ "target_bpp": 8.5,
34
+ "row_rdo_iterations": 2,
35
+ "row_rdo_candidates": 20
36
+ },
37
+ "target_modules": [
38
+ "video_patch_proj",
39
+ "condition_proj",
40
+ "token_refiner.blocks.0.attn.qkv_proj",
41
+ "token_refiner.blocks.0.attn.out_proj",
42
+ "token_refiner.blocks.0.mlp.fc1",
43
+ "token_refiner.blocks.0.mlp.fc2",
44
+ "token_refiner.blocks.1.attn.qkv_proj",
45
+ "token_refiner.blocks.1.attn.out_proj",
46
+ "token_refiner.blocks.1.mlp.fc1",
47
+ "token_refiner.blocks.1.mlp.fc2"
48
+ ]
49
+ }
50
+ ]
51
+ }
52
+ }
models/dit_4p5bit_int8.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e57709eb492542bfbd0f53bad7a8e961ee76ef1c32ef0529280a5ca64a00a89
3
+ size 11459114520
models/dit_6bit.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "c82d7ef090bce613aeaceb3008eb6fd2",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 5.896361607142858,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20
14
+ },
15
+ "exclude_modules": [
16
+ "video_patch_proj",
17
+ "condition_proj",
18
+ "token_refiner.blocks.0.attn.qkv_proj",
19
+ "token_refiner.blocks.0.attn.out_proj",
20
+ "token_refiner.blocks.0.mlp.fc1",
21
+ "token_refiner.blocks.0.mlp.fc2",
22
+ "token_refiner.blocks.1.attn.qkv_proj",
23
+ "token_refiner.blocks.1.attn.out_proj",
24
+ "token_refiner.blocks.1.mlp.fc1",
25
+ "token_refiner.blocks.1.mlp.fc2"
26
+ ]
27
+ },
28
+ {
29
+ "method": "entropack_lossy_quant",
30
+ "backend_config_kwargs": {
31
+ "target_bpp": 8.5,
32
+ "row_rdo_iterations": 2,
33
+ "row_rdo_candidates": 20
34
+ },
35
+ "target_modules": [
36
+ "video_patch_proj",
37
+ "condition_proj",
38
+ "token_refiner.blocks.0.attn.qkv_proj",
39
+ "token_refiner.blocks.0.attn.out_proj",
40
+ "token_refiner.blocks.0.mlp.fc1",
41
+ "token_refiner.blocks.0.mlp.fc2",
42
+ "token_refiner.blocks.1.attn.qkv_proj",
43
+ "token_refiner.blocks.1.attn.out_proj",
44
+ "token_refiner.blocks.1.mlp.fc1",
45
+ "token_refiner.blocks.1.mlp.fc2"
46
+ ]
47
+ }
48
+ ]
49
+ }
50
+ }
models/dit_6bit.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d78fae76ab7e861c8c349b17e369652bced1bfcc077d49942145830fa1496dcc
3
+ size 15347793552
models/dit_6bit_int8.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "0f91c352e019af773165e04a2ab4b2e6",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant_int8",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 5.896361607142858,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20,
14
+ "use_convrot": true,
15
+ "convrot_groupsize": 256
16
+ },
17
+ "exclude_modules": [
18
+ "video_patch_proj",
19
+ "condition_proj",
20
+ "token_refiner.blocks.0.attn.qkv_proj",
21
+ "token_refiner.blocks.0.attn.out_proj",
22
+ "token_refiner.blocks.0.mlp.fc1",
23
+ "token_refiner.blocks.0.mlp.fc2",
24
+ "token_refiner.blocks.1.attn.qkv_proj",
25
+ "token_refiner.blocks.1.attn.out_proj",
26
+ "token_refiner.blocks.1.mlp.fc1",
27
+ "token_refiner.blocks.1.mlp.fc2"
28
+ ]
29
+ },
30
+ {
31
+ "method": "entropack_lossy_quant",
32
+ "backend_config_kwargs": {
33
+ "target_bpp": 8.5,
34
+ "row_rdo_iterations": 2,
35
+ "row_rdo_candidates": 20
36
+ },
37
+ "target_modules": [
38
+ "video_patch_proj",
39
+ "condition_proj",
40
+ "token_refiner.blocks.0.attn.qkv_proj",
41
+ "token_refiner.blocks.0.attn.out_proj",
42
+ "token_refiner.blocks.0.mlp.fc1",
43
+ "token_refiner.blocks.0.mlp.fc2",
44
+ "token_refiner.blocks.1.attn.qkv_proj",
45
+ "token_refiner.blocks.1.attn.out_proj",
46
+ "token_refiner.blocks.1.mlp.fc1",
47
+ "token_refiner.blocks.1.mlp.fc2"
48
+ ]
49
+ }
50
+ ]
51
+ }
52
+ }
models/dit_6bit_int8.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e65f62429760e1e7542c857b464d2dd2714d1fde4c49611163b38847d1431426
3
+ size 15231185416
models/dit_7bit.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "9a07491167182716eccb303ded353b22",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 6.937816964285715,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20
14
+ },
15
+ "exclude_modules": [
16
+ "video_patch_proj",
17
+ "condition_proj",
18
+ "token_refiner.blocks.0.attn.qkv_proj",
19
+ "token_refiner.blocks.0.attn.out_proj",
20
+ "token_refiner.blocks.0.mlp.fc1",
21
+ "token_refiner.blocks.0.mlp.fc2",
22
+ "token_refiner.blocks.1.attn.qkv_proj",
23
+ "token_refiner.blocks.1.attn.out_proj",
24
+ "token_refiner.blocks.1.mlp.fc1",
25
+ "token_refiner.blocks.1.mlp.fc2"
26
+ ]
27
+ },
28
+ {
29
+ "method": "entropack_lossy_quant",
30
+ "backend_config_kwargs": {
31
+ "target_bpp": 8.5,
32
+ "row_rdo_iterations": 2,
33
+ "row_rdo_candidates": 20
34
+ },
35
+ "target_modules": [
36
+ "video_patch_proj",
37
+ "condition_proj",
38
+ "token_refiner.blocks.0.attn.qkv_proj",
39
+ "token_refiner.blocks.0.attn.out_proj",
40
+ "token_refiner.blocks.0.mlp.fc1",
41
+ "token_refiner.blocks.0.mlp.fc2",
42
+ "token_refiner.blocks.1.attn.qkv_proj",
43
+ "token_refiner.blocks.1.attn.out_proj",
44
+ "token_refiner.blocks.1.mlp.fc1",
45
+ "token_refiner.blocks.1.mlp.fc2"
46
+ ]
47
+ }
48
+ ]
49
+ }
50
+ }
models/dit_7bit.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0853c9cbb8444759290e7e98f0c7887f7395683e315a99b0a596ec462931da59
3
+ size 17893766864
models/dit_7bit_int8.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "3251d5dbddd0359b3060750b6b323b4b",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant_int8",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 6.937816964285715,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20,
14
+ "use_convrot": true,
15
+ "convrot_groupsize": 256
16
+ },
17
+ "exclude_modules": [
18
+ "video_patch_proj",
19
+ "condition_proj",
20
+ "token_refiner.blocks.0.attn.qkv_proj",
21
+ "token_refiner.blocks.0.attn.out_proj",
22
+ "token_refiner.blocks.0.mlp.fc1",
23
+ "token_refiner.blocks.0.mlp.fc2",
24
+ "token_refiner.blocks.1.attn.qkv_proj",
25
+ "token_refiner.blocks.1.attn.out_proj",
26
+ "token_refiner.blocks.1.mlp.fc1",
27
+ "token_refiner.blocks.1.mlp.fc2"
28
+ ]
29
+ },
30
+ {
31
+ "method": "entropack_lossy_quant",
32
+ "backend_config_kwargs": {
33
+ "target_bpp": 8.5,
34
+ "row_rdo_iterations": 2,
35
+ "row_rdo_candidates": 20
36
+ },
37
+ "target_modules": [
38
+ "video_patch_proj",
39
+ "condition_proj",
40
+ "token_refiner.blocks.0.attn.qkv_proj",
41
+ "token_refiner.blocks.0.attn.out_proj",
42
+ "token_refiner.blocks.0.mlp.fc1",
43
+ "token_refiner.blocks.0.mlp.fc2",
44
+ "token_refiner.blocks.1.attn.qkv_proj",
45
+ "token_refiner.blocks.1.attn.out_proj",
46
+ "token_refiner.blocks.1.mlp.fc1",
47
+ "token_refiner.blocks.1.mlp.fc2"
48
+ ]
49
+ }
50
+ ]
51
+ }
52
+ }
models/dit_7bit_int8.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45bf78fa49f6d9ce634174a6307aa92388f99110e11314b98101784b6d71c6b2
3
+ size 17749759528
models/dit_8bit.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "d51af0845c9571866c191b765c7e9822",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 7.979272321428573,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20
14
+ },
15
+ "exclude_modules": [
16
+ "video_patch_proj",
17
+ "condition_proj",
18
+ "token_refiner.blocks.0.attn.qkv_proj",
19
+ "token_refiner.blocks.0.attn.out_proj",
20
+ "token_refiner.blocks.0.mlp.fc1",
21
+ "token_refiner.blocks.0.mlp.fc2",
22
+ "token_refiner.blocks.1.attn.qkv_proj",
23
+ "token_refiner.blocks.1.attn.out_proj",
24
+ "token_refiner.blocks.1.mlp.fc1",
25
+ "token_refiner.blocks.1.mlp.fc2"
26
+ ]
27
+ },
28
+ {
29
+ "method": "entropack_lossy_quant",
30
+ "backend_config_kwargs": {
31
+ "target_bpp": 8.5,
32
+ "row_rdo_iterations": 2,
33
+ "row_rdo_candidates": 20
34
+ },
35
+ "target_modules": [
36
+ "video_patch_proj",
37
+ "condition_proj",
38
+ "token_refiner.blocks.0.attn.qkv_proj",
39
+ "token_refiner.blocks.0.attn.out_proj",
40
+ "token_refiner.blocks.0.mlp.fc1",
41
+ "token_refiner.blocks.0.mlp.fc2",
42
+ "token_refiner.blocks.1.attn.qkv_proj",
43
+ "token_refiner.blocks.1.attn.out_proj",
44
+ "token_refiner.blocks.1.mlp.fc1",
45
+ "token_refiner.blocks.1.mlp.fc2"
46
+ ]
47
+ }
48
+ ]
49
+ }
50
+ }
models/dit_8bit.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bba3d748b2e90ae599ec8536e43e12e50b1e0ecd657d7b80e83c893e7fdb8b90
3
+ size 20335801888
models/dit_8bit_int8.json ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "38d987aeb947f79336734b0c5ae68804",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant_int8_raw",
10
+ "backend_config_kwargs": {
11
+ "use_convrot": true,
12
+ "convrot_groupsize": 256
13
+ },
14
+ "exclude_modules": [
15
+ "video_patch_proj",
16
+ "condition_proj",
17
+ "token_refiner.blocks.0.attn.qkv_proj",
18
+ "token_refiner.blocks.0.attn.out_proj",
19
+ "token_refiner.blocks.0.mlp.fc1",
20
+ "token_refiner.blocks.0.mlp.fc2",
21
+ "token_refiner.blocks.1.attn.qkv_proj",
22
+ "token_refiner.blocks.1.attn.out_proj",
23
+ "token_refiner.blocks.1.mlp.fc1",
24
+ "token_refiner.blocks.1.mlp.fc2"
25
+ ]
26
+ },
27
+ {
28
+ "method": "entropack_lossy_quant",
29
+ "backend_config_kwargs": {
30
+ "target_bpp": 8.5,
31
+ "row_rdo_iterations": 2,
32
+ "row_rdo_candidates": 20
33
+ },
34
+ "target_modules": [
35
+ "video_patch_proj",
36
+ "condition_proj",
37
+ "token_refiner.blocks.0.attn.qkv_proj",
38
+ "token_refiner.blocks.0.attn.out_proj",
39
+ "token_refiner.blocks.0.mlp.fc1",
40
+ "token_refiner.blocks.0.mlp.fc2",
41
+ "token_refiner.blocks.1.attn.qkv_proj",
42
+ "token_refiner.blocks.1.attn.out_proj",
43
+ "token_refiner.blocks.1.mlp.fc1",
44
+ "token_refiner.blocks.1.mlp.fc2"
45
+ ]
46
+ }
47
+ ]
48
+ }
49
+ }
models/dit_8bit_int8.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6da1325049c1d51a6f085e7223e978dac11b881b41a9b8f007565e90d148380
3
+ size 20202699112
models/dit_extreme_3bpp.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "adab44d835193ce1b2bd0f85f0586f0e",
3
+ "model_name": "minimax_h3_image_dit",
4
+ "model_class": "diffsynth.models.minimax_h3_image_dit.MiniMaxH3ImageDiTPruned",
5
+ "quant_config": {
6
+ "load_prequantized": true,
7
+ "configs": [
8
+ {
9
+ "method": "entropack_lossy_quant",
10
+ "backend_config_kwargs": {
11
+ "target_bpp": 2.911995535714286,
12
+ "row_rdo_iterations": 2,
13
+ "row_rdo_candidates": 20
14
+ },
15
+ "exclude_modules": [
16
+ "video_patch_proj",
17
+ "condition_proj",
18
+ "token_refiner.blocks.0.attn.qkv_proj",
19
+ "token_refiner.blocks.0.attn.out_proj",
20
+ "token_refiner.blocks.0.mlp.fc1",
21
+ "token_refiner.blocks.0.mlp.fc2",
22
+ "token_refiner.blocks.1.attn.qkv_proj",
23
+ "token_refiner.blocks.1.attn.out_proj",
24
+ "token_refiner.blocks.1.mlp.fc1",
25
+ "token_refiner.blocks.1.mlp.fc2"
26
+ ]
27
+ },
28
+ {
29
+ "method": "entropack_lossy_quant",
30
+ "backend_config_kwargs": {
31
+ "target_bpp": 8.5,
32
+ "row_rdo_iterations": 2,
33
+ "row_rdo_candidates": 20
34
+ },
35
+ "target_modules": [
36
+ "video_patch_proj",
37
+ "condition_proj"
38
+ ]
39
+ },
40
+ {
41
+ "method": "entropack_lossy_quant",
42
+ "backend_config_kwargs": {
43
+ "target_bpp": 5.0,
44
+ "row_rdo_iterations": 2,
45
+ "row_rdo_candidates": 20
46
+ },
47
+ "target_modules": [
48
+ "token_refiner.blocks.0.attn.qkv_proj",
49
+ "token_refiner.blocks.0.attn.out_proj",
50
+ "token_refiner.blocks.0.mlp.fc1",
51
+ "token_refiner.blocks.0.mlp.fc2",
52
+ "token_refiner.blocks.1.attn.qkv_proj",
53
+ "token_refiner.blocks.1.attn.out_proj",
54
+ "token_refiner.blocks.1.mlp.fc1",
55
+ "token_refiner.blocks.1.mlp.fc2"
56
+ ]
57
+ }
58
+ ]
59
+ }
60
+ }
models/dit_extreme_3bpp.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7316ce8afb83170ca5c2595710682323931c97026c34939632d8b96b7690d5cc
3
+ size 7668229904
models/text_encoder_8bit.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "5e498d14ba9ab4f230685ca419f43949",
3
+ "model_name": "minimax_h3_text_encoder",
4
+ "model_class": "diffsynth.models.minimax_h3_text_encoder.MiniMaxH3TextEncoder",
5
+ "quant_config": {
6
+ "method": "entropack_lossy_quant",
7
+ "load_prequantized": true,
8
+ "backend_config_kwargs": {
9
+ "target_bpp": 8.0,
10
+ "row_rdo_iterations": 2,
11
+ "row_rdo_candidates": 20
12
+ }
13
+ }
14
+ }
models/text_encoder_8bit.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ee2925c90ac5bee29a7b11e954e534055d785b480c8b7da51ab05f0d51b528d
3
+ size 26831739344
models/vae_int8.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_hash": "14b7df9252600e262f14869446c77dfb",
3
+ "model_name": "minimax_h3_image_vae",
4
+ "model_class": "diffsynth.models.minimax_h3_image_vae.MiniMaxH3ImageVAE",
5
+ "quant_config": {
6
+ "method": "entropack_lossy_quant_int8_raw",
7
+ "load_prequantized": true,
8
+ "backend_config_kwargs": {
9
+ "use_convrot": true,
10
+ "convrot_groupsize": 256
11
+ }
12
+ }
13
+ }
models/vae_int8.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95d6c5195d73bb1afc19154bcb9facd69c935d5a8231aed42c418d11682e5287
3
+ size 766356456