Hugging Face's logo

moonbox-dev
/
FLUX.1-Fill-dev-MLX

Image-to-Image
MLX
Safetensors
moonbox
flux
inpainting
outpainting
flux-fill
Model card Files
xet
 Community
jennyxue commited on
Commit
1714f9e
·
verified ·
1 Parent(s): dde7b02

Model card and the original license

Browse files
Files changed (1) hide show
  1. README.md +20 -0
README.md CHANGED
@@ -66,6 +66,23 @@ for what was measured).
66
  | 6bit | [`6bit/transformer/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/6bit/transformer/model.safetensors.index.json) | 0.1 MB | `302702931b26` | below threshold (experimental): vs 8bit PSNR 37.8 dB mean, LPIPS 0.056 mean / 0.163 worst |
67
  | 6bit | [`6bit/vae/0.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/6bit/vae/0.safetensors) | 165.2 MB | `5fb147b7776d` | below threshold (experimental): vs 8bit PSNR 37.8 dB mean, LPIPS 0.056 mean / 0.163 worst |
68
  | 6bit | [`6bit/vae/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/6bit/vae/model.safetensors.index.json) | 17.4 kB | `20fb1f73bf5e` | below threshold (experimental): vs 8bit PSNR 37.8 dB mean, LPIPS 0.056 mean / 0.163 worst |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
  ## How to run
71
 
@@ -95,6 +112,9 @@ seed: `6bit/` and `4bit/` against `8bit/` (nothing bigger fits in 24 GB). Pass:
95
  |---|---|---|---|---|
96
  | `8bit/` | 17.0 GB | reference | - | not compared with bf16 |
97
  | `6bit/` | 13.8 GB | 37.8 dB | 0.056 / 0.163 | **just below threshold, experimental** |
 
 
 
98
 
99
  ## How these were made
100
 
 
66
  | 6bit | [`6bit/transformer/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/6bit/transformer/model.safetensors.index.json) | 0.1 MB | `302702931b26` | below threshold (experimental): vs 8bit PSNR 37.8 dB mean, LPIPS 0.056 mean / 0.163 worst |
67
  | 6bit | [`6bit/vae/0.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/6bit/vae/0.safetensors) | 165.2 MB | `5fb147b7776d` | below threshold (experimental): vs 8bit PSNR 37.8 dB mean, LPIPS 0.056 mean / 0.163 worst |
68
  | 6bit | [`6bit/vae/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/6bit/vae/model.safetensors.index.json) | 17.4 kB | `20fb1f73bf5e` | below threshold (experimental): vs 8bit PSNR 37.8 dB mean, LPIPS 0.056 mean / 0.163 worst |
69
+ | 4bit | [`4bit/mflux_model.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/mflux_model.json) | 0.4 kB | `9a500615e1c2` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
70
+ | 4bit | [`4bit/text_encoder/0.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/text_encoder/0.safetensors) | 69.4 MB | `80401ef4d0cb` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
71
+ | 4bit | [`4bit/text_encoder/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/text_encoder/model.safetensors.index.json) | 25.0 kB | `87918a51d110` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
72
+ | 4bit | [`4bit/text_encoder_2/0.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/text_encoder_2/0.safetensors) | 2.15 GB | `daf9b170b819` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
73
+ | 4bit | [`4bit/text_encoder_2/1.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/text_encoder_2/1.safetensors) | 533.3 MB | `3db2a6e9309a` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
74
+ | 4bit | [`4bit/text_encoder_2/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/text_encoder_2/model.safetensors.index.json) | 44.1 kB | `03f5a8751ab8` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
75
+ | 4bit | [`4bit/tokenizer/tokenizer.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/tokenizer/tokenizer.json) | 3.6 MB | `6d9109cc8389` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
76
+ | 4bit | [`4bit/tokenizer/tokenizer_config.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/tokenizer/tokenizer_config.json) | 0.4 kB | `d0eda7377bb9` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
77
+ | 4bit | [`4bit/tokenizer_2/tokenizer.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/tokenizer_2/tokenizer.json) | 2.4 MB | `2432e8aa8341` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
78
+ | 4bit | [`4bit/tokenizer_2/tokenizer_config.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/tokenizer_2/tokenizer_config.json) | 2.5 kB | `e31ed19bc545` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
79
+ | 4bit | [`4bit/transformer/0.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/transformer/0.safetensors) | 2.13 GB | `dbf7be651fcf` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
80
+ | 4bit | [`4bit/transformer/1.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/transformer/1.safetensors) | 2.14 GB | `ae7fa499db8e` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
81
+ | 4bit | [`4bit/transformer/2.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/transformer/2.safetensors) | 2.13 GB | `0bdccd27d8fb` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
82
+ | 4bit | [`4bit/transformer/3.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/transformer/3.safetensors) | 297.6 MB | `6437f5d4cba1` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
83
+ | 4bit | [`4bit/transformer/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/transformer/model.safetensors.index.json) | 0.1 MB | `ef05979b2043` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
84
+ | 4bit | [`4bit/vae/0.safetensors`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/vae/0.safetensors) | 164.7 MB | `53aa93d75594` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
85
+ | 4bit | [`4bit/vae/model.safetensors.index.json`](https://huggingface.co/moonbox-ai/FLUX.1-Fill-dev-MLX/blob/main/4bit/vae/model.safetensors.index.json) | 17.4 kB | `56348f08a8c9` | vs 8bit PSNR 32.0 dB mean, LPIPS 0.051 mean / 0.107 worst |
86
 
87
  ## How to run
88
 
 
112
  |---|---|---|---|---|
113
  | `8bit/` | 17.0 GB | reference | - | not compared with bf16 |
114
  | `6bit/` | 13.8 GB | 37.8 dB | 0.056 / 0.163 | **just below threshold, experimental** |
115
+ | `4bit/` | 9.6 GB | 32.0 dB | 0.051 / 0.107 | passes |
116
+
117
+ About 5 minutes per 512x512 picture (20 steps) for `8bit/` on an Apple M5 with 24 GB in low-RAM mode.
118
 
119
  ## How these were made
120