Add files using upload-large-folder tool

README.md

@@ -0,0 +1,85 @@
+---
+license: apache-2.0
+library_name: mlx
+pipeline_tag: image-to-image
+base_model: hustvl/Moebius
+tags:
+  - mlx
+  - image-inpainting
+  - inpainting
+  - diffusion
+  - moebius
+  - pretrained
+  - q8
+---
+
+# Moebius pretrained MLX q8
+
+This folder contains a converted MLX version of the **pretrained** Moebius checkpoint in **q8** form.
+
+Original upstream model: [hustvl/Moebius](https://huggingface.co/hustvl/Moebius)  
+Original source repository: [hustvl/Moebius](https://github.com/hustvl/Moebius)
+
+Base Moebius checkpoint before the dataset-specific inpainting fine-tunes. Use this when you want the upstream base model rather than a Places2 or face-specialized fine-tune.
+
+## Identity
+
+| Field | Value |
+|---|---|
+| Variant name | `pretrained-q8` |
+| Original Moebius checkpoint family | `pretrained` |
+| Original checkpoint type | base / pretrained checkpoint |
+| Source PyTorch checkpoint | `Moebius-Models/pretrained/diffusion_pytorch_model.bin` |
+| MLX precision / quantization label | `q8` |
+| Image size | 512 x 512 |
+| Latent size | 64 x 64 |
+| Latent channels | 4 |
+| Mask channels | 1 |
+| Conditioning IDs | 20 |
+| VAE scaling factor | 0.13025 |
+| Noise offset | 0.0357 |
+
+## Quantization
+
+8-bit MLX quantized export. The manifest selects `unet_quantized.safetensors` for the UNet; VAE encoder and decoder remain regular f16 safetensors unless a quantized VAE file is explicitly listed. Quantization uses MLX grouped quantization metadata and the runtime applies the matching quantized module layout.
+
+- UNet precision mode: `q8`.
+- MLX grouped quantization is used for supported linear layers; grouped quantized weights are loaded through the Moebius-MLX manifest/runtime.
+- Quantization config: 8 bits, group size 64, standard MLX quantized mode.
+- The VAE encoder and decoder remain regular f16 safetensors for this variant.
+
+## Manifest-selected deployment files
+
+These are the files selected by `manifest.json` when the Moebius-MLX runtime loads this variant.
+
+| Component | File | Size |
+|---|---|---:|
+| UNet | `unet_quantized.safetensors` | 432.89 MB |
+| VAE encoder | `vae_encoder.safetensors` | 68.34 MB |
+| VAE decoder | `vae_decoder.safetensors` | 99.00 MB |
+
+## Files in this folder
+
+- `unet.safetensors`
+- `unet_quantized.safetensors` (selected by manifest)
+- `vae_decoder.safetensors` (selected by manifest)
+- `vae_encoder.safetensors` (selected by manifest)
+- `manifest.json` (runtime metadata and file selection)
+
+A minimal runtime package needs `manifest.json` and the manifest-selected files above. Extra source or fallback files are optional and are not required for inference.
+
+## Runtime expectations
+
+This is not a Transformers or Diffusers-native checkpoint. It is intended for the Swift/MLX runtime in [Moebius-MLX](https://github.com/kylehowells/Moebius-MLX). The runtime reads `manifest.json`, loads the selected safetensors files, builds the Moebius UNet and VAE modules, and runs the DDIM inpainting pipeline.
+
+Pipeline constants must match the manifest:
+
+- DDIM scheduler: `scaled_linear`, beta start 0.00085, beta end 0.012, 1000 train timesteps, clip sample false
+- 512 x 512 image resolution and 64 x 64 latent resolution
+- 9-channel UNet input: noisy latent, mask, and masked-image latent
+- VAE scaling factor 0.13025
+
+
+## Attribution
+
+Moebius was released by the original authors as [hustvl/Moebius](https://huggingface.co/hustvl/Moebius). This folder is a format conversion and/or quantized MLX packaging of the original PyTorch weights, not a newly trained model.

manifest.json

@@ -0,0 +1,26 @@
+{
+  "conditioningIDs": 20,
+  "files": {
+    "unet": "unet.safetensors",
+    "unetQuantized": "unet_quantized.safetensors",
+    "vaeDecoder": "vae_decoder.safetensors",
+    "vaeEncoder": "vae_encoder.safetensors"
+  },
+  "formatVersion": 1,
+  "imageSize": 512,
+  "latentChannels": 4,
+  "latentSize": 64,
+  "maskChannels": 1,
+  "modelName": "pretrained",
+  "noiseOffset": 0.0357,
+  "precision": "q8",
+  "scalingFactor": 0.13025,
+  "scheduler": {
+    "betaEnd": 0.012,
+    "betaSchedule": "scaled_linear",
+    "betaStart": 0.00085,
+    "clipSample": false,
+    "trainTimesteps": 1000
+  },
+  "sourceCheckpoint": "Moebius-Models/pretrained/diffusion_pytorch_model.bin"
+}

unet_quantized.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:165adf970384279df3191c7bc8a5609291d1d7b837f0fb162b42c4e97e690e9e
+size 432889605

vae_decoder.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08778bbb4b48640ad16381904eeeb99635846eb24409f70ae7fdc8eb01a87e76
+size 98995758

vae_encoder.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f096b397453c9d77ad935eaee614fb573d2b07fd57bde80a8986ed203f08a5de
+size 68339216