Initial release: 500K single-frame decoder

.gitattributes

@@ -33,3 +33,18 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+examples/edit/product-material/transition-grid.png filter=lfs diff=lfs merge=lfs -text
+examples/edit/sketch-render/transition-grid.png filter=lfs diff=lfs merge=lfs -text
+examples/edit/sofa-recolor/transition-grid.png filter=lfs diff=lfs merge=lfs -text
+examples/reconstruction/native-h3-single-frame.png filter=lfs diff=lfs merge=lfs -text
+examples/reconstruction/reconstruction-diagram.png filter=lfs diff=lfs merge=lfs -text
+examples/reconstruction/reconstruction-stress.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/diagram-community.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/diagram-comparison.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/diagram-ours.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/natural-community.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/natural-comparison.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/natural-ours.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/product-community.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/product-comparison.png filter=lfs diff=lfs merge=lfs -text
+examples/t2i/product-ours.png filter=lfs diff=lfs merge=lfs -text

MODIFICATIONS.md

@@ -0,0 +1,13 @@
+# Modifications
+
+`minimax_h3_single_frame_decoder_500k.safetensors` is a modified decoder-side
+checkpoint derived from MiniMax H3 and Mamad8/MiniMax-H3-Image-VAE.
+
+The H3-compatible encoder and `quant_conv` remained frozen. The full decoder
+and `post_quant_conv` were further trained for 62,500 optimizer updates on
+500,000 unique still-image reconstruction examples using the curriculum and
+objective documented in `README.md`.
+
+This modification targets independent single-frame decoding. It does not claim
+full-sequence video compatibility.
+

README.md

@@ -0,0 +1,345 @@
+---
+base_model:
+  - MiniMaxAI/MiniMax-H3
+  - Mamad8/MiniMax-H3-Image-VAE
+library_name: diffusers
+tags:
+  - minimax-h3
+  - vae
+  - decoder
+  - image-reconstruction
+  - single-frame
+  - experimental
+---
+
+# MiniMax H3 Single-Frame VAE 500K
+
+This is an independent, experimental **decoder-only** checkpoint for decoding
+one image from a MiniMax H3 latent slice. It is not a full MiniMax H3 model, it
+does not include an encoder or transformer, and it is not used in production.
+
+The decoder started from
+[`Mamad8/MiniMax-H3-Image-VAE`](https://huggingface.co/Mamad8/MiniMax-H3-Image-VAE)
+and was further trained on 500,000 unique image-reconstruction examples using
+the frozen H3-compatible encoder. The full decoder and `post_quant_conv` were
+trainable. Text and captions were not used during decoder training.
+
+## Scope
+
+Use this checkpoint when the intended output is one independently decoded
+image. It is most reliable on structured content such as product contours,
+line art, diagrams, documents, and UI-like layouts.
+
+Do **not** treat it as a replacement decoder for MiniMax H3 video. Still-image
+training supervised one temporal boundary condition, while the full temporal
+decoder weights were updated. Later frames in a full-sequence decode can show
+grid or block artifacts, flicker, abrupt transitions, and texture drift.
+
+## Examples
+
+All decoder comparisons use the same input image or the same frozen generated
+latent for every displayed decoder. The new T2I prompts were fixed before the
+outputs were reviewed.
+
+### Reconstruction
+
+![Diagram reconstruction comparison](examples/reconstruction/reconstruction-diagram.png)
+
+The fixed benchmark also includes natural photographs and scene text. Their
+aggregate metrics are reported below, but the source images are not
+redistributed here. A procedurally generated
+[high-frequency stress case](examples/reconstruction/reconstruction-stress.png)
+and a [native H3 single-frame comparison](examples/reconstruction/native-h3-single-frame.png)
+are included.
+
+### H3 text-to-image through a single latent slice
+
+The H3 `t2va` workflow generated a full normalized latent sequence. One fixed
+temporal slice was then independently decoded with the community decoder and
+this checkpoint.
+
+![Product T2I comparison](examples/t2i/product-comparison.png)
+
+Prompt: “Studio product photograph of a compact futuristic electric espresso
+machine, brushed aluminum and matte black, precise industrial design, centered
+three-quarter view, soft gray seamless background, crisp edges, realistic
+materials, no people, no text.”
+
+Additional fixed examples:
+
+- [Industrial-design board](examples/t2i/diagram-comparison.png)
+- [Natural photograph](examples/t2i/natural-comparison.png)
+
+These examples test the decoder, not a native text-to-image pipeline. H3 still
+constructs and denoises its joint video/audio latent, so this is substantially
+more expensive than a purpose-built image generator.
+
+### First-frame-conditioned image editing
+
+The following examples use H3's `fl2va` workflow. The contact sheets show the
+source followed by independently decoded temporal slices. They intentionally
+show the transition instead of selecting only the most flattering frame.
+
+![Sofa recolor transition](examples/edit/sofa-recolor/transition-grid.png)
+
+Instruction: “Change only the sofa upholstery from rust red to deep navy blue
+velvet. Preserve the room layout, window, lamp, rug, floor, lighting, and
+camera.”
+
+Additional edits:
+
+- [Sketch-to-render transition](examples/edit/sketch-render/transition-grid.png)
+- [Product-material transition](examples/edit/product-material/transition-grid.png)
+
+This is image-conditioned video used as a static-edit experiment. Early slices
+usually reconstruct the source, later slices apply more of the requested edit,
+and composition, lighting, material, or geometry can drift during the
+transition. A useful result may depend on searching multiple slices. That is a
+fundamental limitation compared with a native image-edit model.
+
+## Fixed reconstruction benchmark
+
+The benchmark contains 288 held-out cases spanning photographs, text,
+documents, diagrams, line art, product contours, UI-like content, and synthetic
+high-frequency stress patterns.
+
+| Decoder | PSNR dB ↑ | SSIM ↑ | LPIPS-Alex ↓ | DISTS ↓ | Edge F1 ↑ |
+|---|---:|---:|---:|---:|---:|
+| Official MiniMax H3 VAE | 30.3544 | **0.9258** | **0.0323** | **0.0533** | **0.7387** |
+| Mamad8/MiniMax-H3-Image-VAE | 22.7022 | 0.8060 | 0.1671 | 0.2494 | 0.4647 |
+| **This checkpoint** | **31.1185** | 0.9208 | 0.0454 | 0.0883 | 0.7236 |
+
+This checkpoint has the highest PSNR in this three-way benchmark, but the
+official H3 decoder remains better on SSIM, LPIPS, DISTS, and Edge F1. The
+result should not be read as universal superiority over the official decoder.
+
+### What the aggregate metrics mean
+
+- PSNR emphasizes average pixel accuracy. This checkpoint is `+0.7641` dB
+  above official H3 and `+8.4163` dB above the community initialization.
+- SSIM measures structural similarity. Official H3 remains ahead by `0.0050`.
+- LPIPS and DISTS are perceptual-distance metrics. Official H3 remains clearly
+  better, especially on photographs and small text.
+- Edge F1 measures alignment of detected edges. This checkpoint is close to
+  official in aggregate, but the result varies substantially by domain.
+
+Compared with the community initialization, this checkpoint reduces LPIPS by
+about 72.8%, reduces DISTS by about 64.6%, and raises Edge F1 from `0.4647` to
+`0.7236`. Compared with official H3, it trades slightly better average pixel
+accuracy for worse perceptual texture and slightly worse edge alignment.
+
+### Per-domain study
+
+Each cell below is `this checkpoint / official H3`. Lower is better for LPIPS
+and DISTS; higher is better for the other metrics.
+
+| Domain | Cases | PSNR dB ↑ | SSIM ↑ | LPIPS ↓ | DISTS ↓ | Edge F1 ↑ |
+|---|---:|---:|---:|---:|---:|---:|
+| Broad photographs | 64 | 29.7037 / 30.9661 | 0.8504 / 0.8816 | 0.0873 / 0.0372 | 0.1439 / 0.0413 | 0.7679 / 0.8377 |
+| UI | 32 | 29.1883 / 28.3791 | 0.9571 / 0.9510 | 0.0316 / 0.0251 | 0.0746 / 0.0424 | 0.8530 / 0.8546 |
+| Documents | 32 | 27.3963 / 27.0216 | 0.9460 / 0.9333 | 0.0516 / 0.0366 | 0.0779 / 0.0465 | 0.7632 / 0.7326 |
+| Scene text | 32 | 32.5351 / 33.3653 | 0.9270 / 0.9312 | 0.0435 / 0.0204 | 0.0898 / 0.0304 | 0.8541 / 0.8937 |
+| Line art | 32 | 36.3202 / 34.1489 | 0.9975 / 0.9962 | 0.0058 / 0.0078 | 0.0463 / 0.0357 | 0.6666 / 0.7013 |
+| Diagrams | 32 | 34.0079 / 29.9501 | 0.9928 / 0.9815 | 0.0063 / 0.0256 | 0.0346 / 0.0570 | 0.8656 / 0.8225 |
+| Product contours | 32 | 37.2439 / 34.4608 | 0.9956 / 0.9905 | 0.0058 / 0.0164 | 0.0300 / 0.0421 | 0.6966 / 0.7082 |
+| Synthetic stress patterns | 32 | 23.9673 / 23.9315 | 0.7707 / 0.7850 | 0.0892 / 0.0846 | 0.1536 / 0.1433 | 0.2770 / 0.2598 |
+
+The strongest result is on diagrams, where this checkpoint beats official H3
+on all five metrics. It also performs well on product contours, line art,
+documents, and UI. These domains reward clean geometry, flat regions, and
+structured edges, which match the decoder's training signal well.
+
+The clearest weakness is natural photography. Pixel values remain close, but
+LPIPS and DISTS show that texture, tone, and fine local structure drift more
+than with official H3. Scene text is another weak area: small lettering and
+thin glyphs may soften, change shape, or become illegible. Dense grids,
+checkerboards, and repeated high-frequency patterns remain phase-sensitive and
+can show moiré or small line shifts.
+
+| Intended use | Assessment | Main caveat |
+|---|---|---|
+| Diagrams and structured graphics | Strong | Fine repeated patterns can still shift phase |
+| Product contours and industrial-design forms | Strong | Very thin edges can move by a pixel or soften |
+| UI and documents | Competitive | Small text is less reliable than large layout structure |
+| Line art | Strong pixel fidelity | Edge F1 remains below official H3 |
+| Natural photographs | Limited | Texture and perceptual detail trail official H3 |
+| Exact scene text | Limited | Glyph identity and thin strokes can change |
+| Full H3 video | Unsupported | Still-image training did not supervise temporal consistency |
+
+Scaling from the earlier 75K decoder to 500K improved SSIM, LPIPS, DISTS, and
+Edge F1, but it did not close the photograph or scene-text gap. This suggests
+that additional generic reconstruction data alone is unlikely to solve those
+weaknesses. Targeted natural-texture, tiny-text, and phase-aware training would
+be more directly relevant.
+
+An exact 74-frame native H3 latent panel also completed with no non-finite or
+geometry failures. Those generated latents have no ground-truth image, so
+decoder-to-decoder distances are disagreement measurements rather than quality
+scores. An AI visual review found the 500K outputs consistently sharper and
+more coherent than the community decoder, but this was not a human preference
+study and does not establish superiority over official H3 on generator
+latents.
+
+### Study boundaries
+
+- Reconstruction metrics use 288 fixed held-out cases with known targets.
+- T2I and editing examples are illustrative fixed-seed cases, not a benchmark.
+- Generator-latent comparisons have no ground-truth image.
+- The public study does not include human preference ratings or confidence
+  intervals.
+- The 1024 stage ran near the end of a cosine schedule with almost zero learning
+  rate, so it does not establish mature 1024-resolution specialization.
+
+## Training
+
+| Stage | Unique images | Resolution | Microbatch |
+|---|---:|---:|---:|
+| 0 | 250,000 | 256 | 8 |
+| 1 | 150,000 | 384 | 8 |
+| 2 | 75,000 | 512 | 8 |
+| 3 | 20,000 | 768 | 2 |
+| 4 | 5,000 | 1024 | 1 |
+
+- Optimizer updates: 62,500
+- Effective batch size: 8
+- Learning rate: `2e-6`
+- Warmup: 625 updates
+- Schedule: one continuous cosine decay
+- Trainable modules: full decoder and `post_quant_conv`
+- Frozen modules: encoder and `quant_conv`
+- Distillation coefficient: `0.0`
+
+The reconstruction objective was:
+
+```text
+1.0          * Charbonnier reconstruction
+0.2          * first-order edge reconstruction
+0.1          * (1 - SSIM)
+0.05         * log-frequency-magnitude reconstruction
+0.0113518046 * LPIPS-Alex
+```
+
+The final 5,000-image 1024 stage occurred near the end of the cosine schedule,
+when the learning rate was almost zero. It was neutral to slightly positive,
+but it does not establish meaningful 1024-resolution specialization.
+
+## Training data
+
+Of the 500,000 training examples, 475,140 came from disjoint shards of
+[`Spawning/pd12m-full`](https://huggingface.co/datasets/Spawning/pd12m-full).
+The other 24,860 were native-resolution hard cases: 19,860 at 768 and 5,000
+at 1024, balanced across WebSight, DocLayNet, Quick, Draw!, deterministic
+procedural diagrams, and deterministic procedural product contours. No
+training images are included in this repository. Users should review each
+source dataset's current terms and attribution requirements independently.
+
+## Loading
+
+Clone the repository and install the pinned dependencies:
+
+```shell
+git clone https://huggingface.co/iamkaikai/MiniMax-H3-Single-Frame-VAE-500K
+cd MiniMax-H3-Single-Frame-VAE-500K
+pip install -r requirements.txt
+```
+
+The following CUDA example loads one normalized H3 generator-latent sequence,
+selects a temporal slice, decodes it independently, and writes a PNG. The
+input safetensors file must contain a tensor named `latents` with shape
+`[batch, channels, time, height, width]` in H3's normalized generator-latent
+space.
+
+```python
+import torch
+from diffusers import AutoencoderKLMiniMaxH3
+from PIL import Image
+from safetensors.torch import load_file
+
+from load_decoder import load_single_frame_decoder
+
+device = "cuda"
+dtype = torch.float16
+
+vae = AutoencoderKLMiniMaxH3.from_pretrained(
+    "MiniMaxAI/MiniMax-H3",
+    subfolder="vae",
+    revision="42ed227ee7df40d41602854ae760620d6eb651fe",
+    torch_dtype=dtype,
+).to(device).eval()
+
+load_single_frame_decoder(vae, "minimax_h3_single_frame_decoder_500k.safetensors")
+
+# This is the normalized latent emitted by H3 generation, before applying
+# vae.config.latents_std and vae.config.latents_mean.
+normalized = load_file("normalized-latents.safetensors")["latents"]
+latent_index = 0
+z = normalized[:, :, latent_index : latent_index + 1].to(device, dtype=dtype)
+
+latent_mean = torch.tensor(
+    vae.config.latents_mean, device=device, dtype=dtype
+).view(1, -1, 1, 1, 1)
+latent_std = torch.tensor(
+    vae.config.latents_std, device=device, dtype=dtype
+).view(1, -1, 1, 1, 1)
+z = z * latent_std + latent_mean
+
+# Direct decoder use is intentional for a temporal extent of one.
+with torch.inference_mode(), torch.autocast("cuda", dtype=dtype):
+    decoded = vae.decoder(vae.post_quant_conv(z))
+
+pixel_mean = torch.tensor(
+    [0.485, 0.456, 0.406], device=device
+).view(1, 3, 1, 1)
+pixel_std = torch.tensor(
+    [0.229, 0.224, 0.225], device=device
+).view(1, 3, 1, 1)
+frame = (decoded[:, :, -1].float() * pixel_std + pixel_mean).clamp(0, 1)
+
+rgb = frame[0].mul(255).round().to(torch.uint8).permute(1, 2, 0).cpu().numpy()
+Image.fromarray(rgb, mode="RGB").save("decoded-frame.png")
+```
+
+The file contains 585 tensors with `decoder.` and `post_quant_conv.` prefixes.
+It intentionally contains no encoder weights. `load_single_frame_decoder`
+strictly verifies the 9 GB file's SHA-256 by default, so the first load can take
+several seconds before GPU inference begins.
+
+## Known limitations
+
+- Single-frame use only. Full video decoding is unsupported.
+- Natural textures can be smoothed or shifted.
+- Small lettering and thin glyphs can change.
+- Dense lines, checkerboards, and fine grids can show moiré or phase errors.
+- H3 T2I and FL2VA examples still pay the cost of full video/audio latent
+  generation.
+- FL2VA edits can require searching temporal slices and can drift in crop,
+  lighting, material, geometry, or unrelated content.
+- The public comparisons are a bounded benchmark, not a human preference
+  study and not evidence of universal superiority.
+
+## Reproducibility
+
+- Training exposures: `500000`
+- Optimizer step: `62500`
+- Training config fingerprint:
+  `225279cbb4cd1e2db94b5b2ae15e123a921fefada0782e5c68d35302e4b4d0a9`
+- Original final decoder SHA-256:
+  `4a5e4d0041c6fdd15af5292db3e10574e28500b216ad2d1f78b5951f2b8abfe2`
+- Audited release export SHA-256:
+  `6c5ff2caa8fade6769f4dd53ee244f77a06652c8cb66b2fedc93f75046d9f001`
+- Community initialization revision:
+  `c7b9252c73707dba494cf4d99ca45d3f33f561b3`
+- Official MiniMax H3 revision:
+  `42ed227ee7df40d41602854ae760620d6eb651fe`
+
+The release export contains the same 585 decoder tensors as the exact final
+resume state. Its file hash differs from the training export because the
+release file includes provenance metadata.
+
+## License
+
+This checkpoint is a modified derivative of MiniMax H3 and remains subject to
+the [MiniMax H3 Community License Agreement](LICENSE). The upstream license
+contains territorial and use restrictions. Review it before downloading,
+using, or redistributing this checkpoint. See [NOTICE](NOTICE) and
+[MODIFICATIONS.md](MODIFICATIONS.md).

example-prompts.json

@@ -0,0 +1,50 @@
+{
+  "schema_version": 1,
+  "selection_policy": "Prompts fixed before generation; no best-of-seed search.",
+  "t2i": [
+    {
+      "id": "product",
+      "prompt": "Studio product photograph of a compact futuristic electric espresso machine, brushed aluminum and matte black, precise industrial design, centered three-quarter view, soft gray seamless background, crisp edges, realistic materials, no people, no text.",
+      "seed": 2026081901,
+      "inference_steps": 30,
+      "latent_index": 0
+    },
+    {
+      "id": "diagram",
+      "prompt": "Clean industrial design concept board showing a modular electric bicycle from side view, one large hero rendering with three small exploded-view callouts, white background, graphite and cobalt accents, precise linework, premium presentation.",
+      "seed": 2026081902,
+      "inference_steps": 30,
+      "latent_index": 0
+    },
+    {
+      "id": "natural",
+      "prompt": "Cinematic close-up photograph of a red fox standing in wet green moss after rain, detailed fur and droplets, soft morning light, shallow depth of field, natural colors, 35mm lens.",
+      "seed": 2026081903,
+      "inference_steps": 30,
+      "latent_index": 0
+    }
+  ],
+  "image_editing": [
+    {
+      "id": "sketch-render",
+      "prompt": "Transform the hand-drawn compact desktop coffee maker into a photorealistic premium industrial design render in brushed aluminum and matte black. Preserve the exact silhouette, camera, and layout.",
+      "seed": 2026081911,
+      "inference_steps": 25,
+      "decoded_latent_indices": [0, 9, 18, 27, 36]
+    },
+    {
+      "id": "sofa-recolor",
+      "prompt": "Change only the sofa upholstery from rust red to deep navy blue velvet. Preserve the room layout, window, lamp, rug, floor, lighting, and camera.",
+      "seed": 2026081912,
+      "inference_steps": 25,
+      "decoded_latent_indices": [0, 9, 18, 27, 36]
+    },
+    {
+      "id": "product-material",
+      "prompt": "Change only the coffee machine body from warm silver to glossy cobalt blue anodized metal. Preserve the geometry, black front panel, controls, shadow, background, and camera.",
+      "seed": 2026081913,
+      "inference_steps": 25,
+      "decoded_latent_indices": [0, 9, 18, 27, 36]
+    }
+  ]
+}

load_decoder.py

@@ -0,0 +1,69 @@
+"""Strictly load the 500K single-frame weights into a MiniMax H3 VAE."""
+
+from __future__ import annotations
+
+import hashlib
+from pathlib import Path
+from typing import Any
+
+from safetensors.torch import load_file
+
+
+EXPECTED_SHA256 = "6c5ff2caa8fade6769f4dd53ee244f77a06652c8cb66b2fedc93f75046d9f001"
+EXPECTED_TENSOR_COUNT = 585
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for chunk in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def _split_state(state: dict[str, Any]) -> tuple[dict[str, Any], dict[str, Any]]:
+    decoder = {
+        key.removeprefix("decoder."): value
+        for key, value in state.items()
+        if key.startswith("decoder.")
+    }
+    post_quant_conv = {
+        key.removeprefix("post_quant_conv."): value
+        for key, value in state.items()
+        if key.startswith("post_quant_conv.")
+    }
+    unexpected = sorted(
+        key
+        for key in state
+        if not key.startswith("decoder.")
+        and not key.startswith("post_quant_conv.")
+    )
+    if unexpected or not decoder or not post_quant_conv:
+        raise RuntimeError(
+            "single-frame decoder key contract differs: "
+            f"unexpected={unexpected[:5]}, decoder={len(decoder)}, "
+            f"post_quant_conv={len(post_quant_conv)}"
+        )
+    return decoder, post_quant_conv
+
+
+def load_single_frame_decoder(
+    vae: Any,
+    checkpoint_path: str | Path,
+    *,
+    verify_sha256: bool = True,
+) -> Any:
+    """Replace only ``decoder`` and ``post_quant_conv`` on an H3 VAE."""
+
+    checkpoint = Path(checkpoint_path)
+    if verify_sha256 and _sha256(checkpoint) != EXPECTED_SHA256:
+        raise RuntimeError("single-frame decoder SHA-256 differs")
+    state = load_file(str(checkpoint), device="cpu")
+    if len(state) != EXPECTED_TENSOR_COUNT:
+        raise RuntimeError(
+            f"expected {EXPECTED_TENSOR_COUNT} tensors, found {len(state)}"
+        )
+    decoder, post_quant_conv = _split_state(state)
+    vae.decoder.load_state_dict(decoder, strict=True)
+    vae.post_quant_conv.load_state_dict(post_quant_conv, strict=True)
+    return vae

minimax_h3_single_frame_decoder_500k.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c5ff2caa8fade6769f4dd53ee244f77a06652c8cb66b2fedc93f75046d9f001
+size 9694193992

release-manifest.json

@@ -0,0 +1,250 @@
+{
+  "files": [
+    {
+      "bytes": 1244,
+      "path": "ASSET_PROVENANCE.md",
+      "sha256": "7705779ec1e701831a17354aec25ac650552e6cac724a950a9db55eedbb2aa6e"
+    },
+    {
+      "bytes": 17604,
+      "path": "LICENSE",
+      "sha256": "59b99642b95ea21630e311198ddbfffbfe05aadba0c2f5d884cbdf4efcc90f44"
+    },
+    {
+      "bytes": 545,
+      "path": "MODIFICATIONS.md",
+      "sha256": "aa29c19da66df5f30c571118d3e1b8efce2bf2e97c2f2a2f0fc7452d42cd0c65"
+    },
+    {
+      "bytes": 122,
+      "path": "NOTICE",
+      "sha256": "0ee234b090f3f3f19484751f051668c145ca9d0c180417f7abef46e8378a923d"
+    },
+    {
+      "bytes": 15253,
+      "path": "README.md",
+      "sha256": "3bc8f2f5f91069b2610eb17d7388da791411ea984d40842848541cf9985bc4bb"
+    },
+    {
+      "bytes": 2245,
+      "path": "example-prompts.json",
+      "sha256": "025ab567aa1ccbfd292bb85ac2276fccbac9a06dc25f855432f860ec1cd99e02"
+    },
+    {
+      "bytes": 93081,
+      "path": "examples/edit/product-material/contact-sheet.jpg",
+      "sha256": "fe84c13209dc305c50592e92fdd08cc320b90f97a058cd7ddf9c1b7c6d65b12a"
+    },
+    {
+      "bytes": 316819,
+      "path": "examples/edit/product-material/transition-grid.png",
+      "sha256": "8f38664ff83fed8e22d58f7cc46ebea8130647b6be884340ebba8bb688763fd0"
+    },
+    {
+      "bytes": 62004,
+      "path": "examples/edit/sketch-render/contact-sheet.jpg",
+      "sha256": "6581b9d497752fb9c86e9715f02f34f1f8a7d46c9d22953db049b980f0134029"
+    },
+    {
+      "bytes": 202583,
+      "path": "examples/edit/sketch-render/transition-grid.png",
+      "sha256": "58f23d459bb7596a94788c27d40297a2bf5b22cfb7a6020bba4bc27c589d3131"
+    },
+    {
+      "bytes": 96059,
+      "path": "examples/edit/sofa-recolor/contact-sheet.jpg",
+      "sha256": "8e9e7a70ca9776468914df02cd126025b941a5f97b0f27900042631db016dc2d"
+    },
+    {
+      "bytes": 391559,
+      "path": "examples/edit/sofa-recolor/transition-grid.png",
+      "sha256": "8be0f1fd04a3602f89849f2ad48e53a4b59098dc43299b785129d1f6628ccf35"
+    },
+    {
+      "bytes": 3111329,
+      "path": "examples/reconstruction/native-h3-single-frame.png",
+      "sha256": "f196b6e2d86ccdd4937190d087f7582bd0d1226b4225611ea36d728768c58cbe"
+    },
+    {
+      "bytes": 1014965,
+      "path": "examples/reconstruction/reconstruction-diagram.png",
+      "sha256": "2e16eb24d6688aa09908d7458a599091422295fefb90197977ab42990b243115"
+    },
+    {
+      "bytes": 3999517,
+      "path": "examples/reconstruction/reconstruction-stress.png",
+      "sha256": "ddcb073ab2c95dc66a7055b356c0fa7ad4358b3b721a0418824fae59bdd23a42"
+    },
+    {
+      "bytes": 414247,
+      "path": "examples/t2i/diagram-community.png",
+      "sha256": "1fc1cb201cc64e397d963d668c3ad09bac335ebe7f35e6f74384c14131735232"
+    },
+    {
+      "bytes": 648473,
+      "path": "examples/t2i/diagram-comparison.png",
+      "sha256": "01170a97d39e25dbf93f7c157feb2e3857e0e15ea5cfdca234c3fb1fb6a14927"
+    },
+    {
+      "bytes": 243607,
+      "path": "examples/t2i/diagram-ours.png",
+      "sha256": "54f5fc862dce7ae440aaa802bbcae44c04e27e1cd306a6dc1ce06906ce6410f5"
+    },
+    {
+      "bytes": 650559,
+      "path": "examples/t2i/natural-community.png",
+      "sha256": "c2416e00300b5a47a0c17597b3f6e75d60ba843e89e9e8677cde5ec1cb39dde4"
+    },
+    {
+      "bytes": 1250127,
+      "path": "examples/t2i/natural-comparison.png",
+      "sha256": "9f6aacba304de88ff05055598fa7bd92f30c1081147fcd5162f3cec18325248b"
+    },
+    {
+      "bytes": 620719,
+      "path": "examples/t2i/natural-ours.png",
+      "sha256": "7c837927ea26213a9a40f756c2ff5722294b768534a20f925898463081a85559"
+    },
+    {
+      "bytes": 410060,
+      "path": "examples/t2i/product-community.png",
+      "sha256": "ffd3a45aafb66446a191ea4b5800180f9cfc0bc3f75c7b7bbf68e50eae6acb1f"
+    },
+    {
+      "bytes": 673826,
+      "path": "examples/t2i/product-comparison.png",
+      "sha256": "4885f6b2b035b26b41b5fc6f3d641cd73633801c1f83ec7088064233f7425761"
+    },
+    {
+      "bytes": 297480,
+      "path": "examples/t2i/product-ours.png",
+      "sha256": "8dcc562b2c1753c4e52f952666ea20693a05e6b85f2c8a644a3360ca093e3c67"
+    },
+    {
+      "bytes": 2202,
+      "path": "load_decoder.py",
+      "sha256": "4638ee384e9abb0254f5e46bfd1212cd8b5f9d923b1560fc0b1da08dd9c861dd"
+    },
+    {
+      "bytes": 9694193992,
+      "path": "minimax_h3_single_frame_decoder_500k.safetensors",
+      "sha256": "6c5ff2caa8fade6769f4dd53ee244f77a06652c8cb66b2fedc93f75046d9f001"
+    },
+    {
+      "bytes": 155,
+      "path": "requirements.txt",
+      "sha256": "44a88473c0eda6e40e2489ab9323dd2f8dc5098bf8f8556d6d66ec49d839fb48"
+    },
+    {
+      "bytes": 6319,
+      "path": "source-fixtures/interior.png",
+      "sha256": "49e7096a1df03068051f99a6522df320d7cb75599eaa42408b69625cc57941a0"
+    },
+    {
+      "bytes": 5470,
+      "path": "source-fixtures/product-photo.png",
+      "sha256": "6e4aa4be38360d7c4bde2589535966e9f76c6a41f96b02736d33efdd28035808"
+    },
+    {
+      "bytes": 5760,
+      "path": "source-fixtures/product-sketch.png",
+      "sha256": "475ea1cf6abda5cffd57cedb4c75e95ef6fe250419be01be3799c6e0060b67b4"
+    },
+    {
+      "bytes": 2011,
+      "path": "vae/config.json",
+      "sha256": "78f67deec3d63aae807f2bfe7154bc1e26f6372cb20b63265fcbae1b62bb5745"
+    }
+  ],
+  "issues": [],
+  "publication_status": "blocked_pending_upstream_distribution_clearance",
+  "repository_target": "iamkaikai/MiniMax-H3-Single-Frame-VAE-500K",
+  "required_examples": [
+    {
+      "bytes": 1014965,
+      "height": 2812,
+      "mode": "RGB",
+      "path": "examples/reconstruction/reconstruction-diagram.png",
+      "sha256": "2e16eb24d6688aa09908d7458a599091422295fefb90197977ab42990b243115",
+      "width": 1280
+    },
+    {
+      "bytes": 3999517,
+      "height": 2812,
+      "mode": "RGB",
+      "path": "examples/reconstruction/reconstruction-stress.png",
+      "sha256": "ddcb073ab2c95dc66a7055b356c0fa7ad4358b3b721a0418824fae59bdd23a42",
+      "width": 1280
+    },
+    {
+      "bytes": 3111329,
+      "height": 1996,
+      "mode": "RGB",
+      "path": "examples/reconstruction/native-h3-single-frame.png",
+      "sha256": "f196b6e2d86ccdd4937190d087f7582bd0d1226b4225611ea36d728768c58cbe",
+      "width": 1152
+    },
+    {
+      "bytes": 673826,
+      "height": 596,
+      "mode": "RGB",
+      "path": "examples/t2i/product-comparison.png",
+      "sha256": "4885f6b2b035b26b41b5fc6f3d641cd73633801c1f83ec7088064233f7425761",
+      "width": 1920
+    },
+    {
+      "bytes": 648473,
+      "height": 596,
+      "mode": "RGB",
+      "path": "examples/t2i/diagram-comparison.png",
+      "sha256": "01170a97d39e25dbf93f7c157feb2e3857e0e15ea5cfdca234c3fb1fb6a14927",
+      "width": 1920
+    },
+    {
+      "bytes": 1250127,
+      "height": 596,
+      "mode": "RGB",
+      "path": "examples/t2i/natural-comparison.png",
+      "sha256": "9f6aacba304de88ff05055598fa7bd92f30c1081147fcd5162f3cec18325248b",
+      "width": 1920
+    },
+    {
+      "bytes": 202583,
+      "height": 632,
+      "mode": "RGB",
+      "path": "examples/edit/sketch-render/transition-grid.png",
+      "sha256": "58f23d459bb7596a94788c27d40297a2bf5b22cfb7a6020bba4bc27c589d3131",
+      "width": 1440
+    },
+    {
+      "bytes": 391559,
+      "height": 632,
+      "mode": "RGB",
+      "path": "examples/edit/sofa-recolor/transition-grid.png",
+      "sha256": "8be0f1fd04a3602f89849f2ad48e53a4b59098dc43299b785129d1f6628ccf35",
+      "width": 1440
+    },
+    {
+      "bytes": 316819,
+      "height": 632,
+      "mode": "RGB",
+      "path": "examples/edit/product-material/transition-grid.png",
+      "sha256": "8f38664ff83fed8e22d58f7cc46ebea8130647b6be884340ebba8bb688763fd0",
+      "width": 1440
+    }
+  ],
+  "schema_version": 1,
+  "status": "passed",
+  "weight": {
+    "bytes": 9694193992,
+    "metadata": {
+      "source_config_fingerprint": "225279cbb4cd1e2db94b5b2ae15e123a921fefada0782e5c68d35302e4b4d0a9",
+      "source_resume_sha256": "40e94c1bef5a8467b5d41ec7c039212c6246b67b37f61b9ae802f4f48dfb23b4",
+      "source_run_name": "minimax_h3_t1_hardwarm_scale500k_curriculum256_384_512_768_1024_20260818_v2",
+      "source_step": "62500"
+    },
+    "path": "minimax_h3_single_frame_decoder_500k.safetensors",
+    "sha256": "6c5ff2caa8fade6769f4dd53ee244f77a06652c8cb66b2fedc93f75046d9f001",
+    "tensor_count": 585
+  }
+}

requirements.txt

@@ -0,0 +1,7 @@
+git+https://github.com/huggingface/diffusers.git@9284607295a09f759aadd65ed08f48b35feea6d9
+accelerate
+huggingface_hub
+Pillow
+safetensors
+torch
+transformers

vae/config.json

@@ -0,0 +1,97 @@
+{
+  "_class_name": "AutoencoderKLMiniMaxH3",
+  "_diffusers_version": "0.36.0.dev0",
+  "in_channels": 3,
+  "out_channels": 3,
+  "latent_channels": 24,
+  "block_out_channels": [
+    128,
+    256,
+    256,
+    512,
+    512,
+    1024
+  ],
+  "layers_per_block": 2,
+  "spatial_downsample_factors": [
+    2,
+    2,
+    2,
+    2,
+    1,
+    1
+  ],
+  "temporal_downsample_factors": [
+    1,
+    2,
+    2,
+    1,
+    1,
+    1
+  ],
+  "norm_num_groups": 32,
+  "norm_eps": 1e-06,
+  "spatial_padding_mode": "reflect",
+  "decoder_num_layers": 36,
+  "decoder_num_attention_heads": 32,
+  "decoder_attention_head_dim": 64,
+  "decoder_num_register_tokens": 4,
+  "decoder_ffn_mult": 4,
+  "decoder_rope_theta": 100.0,
+  "decoder_rope_dim_ratio": 0.75,
+  "decoder_norm_eps": 1e-05,
+  "clip_length": 17,
+  "token_drop": 3,
+  "latents_mean": [
+    0.858090341091156,
+    -0.9606591463088989,
+    1.0661640167236328,
+    -0.5090325474739075,
+    -0.2727581858634949,
+    -1.3675414323806763,
+    -0.2553254961967468,
+    -0.26907554268836975,
+    -0.5376840829849243,
+    -0.0464097298681736,
+    0.6657370328903198,
+    0.19690127670764923,
+    -0.5460608005523682,
+    -0.4035342037677765,
+    -0.23683024942874908,
+    0.25928452610969543,
+    -0.30133944749832153,
+    0.211341992020607,
+    -1.1206848621368408,
+    0.3581933379173279,
+    -0.04225143790245056,
+    0.2604829967021942,
+    0.22864092886447906,
+    0.7056031823158264
+  ],
+  "latents_std": [
+    1.2223774194717407,
+    1.2767263650894165,
+    1.6831774711608887,
+    1.7549455165863037,
+    1.5636216402053833,
+    2.194143533706665,
+    0.9653137922286987,
+    1.0569885969161987,
+    0.841948926448822,
+    0.7729952931404114,
+    1.8955937623977661,
+    0.946841835975647,
+    0.7996809482574463,
+    0.44988900423049927,
+    0.7197399735450745,
+    0.6936293244361877,
+    2.961095094680786,
+    2.7694199085235596,
+    3.0496184825897217,
+    2.1088054180145264,
+    3.276226282119751,
+    3.1627357006073,
+    2.2816812992095947,
+    2.6127843856811523
+  ]
+}