Files changed (1) hide show
  1. README.md +26 -12
README.md CHANGED
@@ -17,20 +17,32 @@ By organizing Core ML models in one place, it will be easier to find them and fo
17
 
18
  <hr>
19
 
20
- ## Conversion Flags
21
 
22
- The models were converted using the following flags:\
23
- `--convert-vae-decoder --convert-vae-encoder --convert-unet --convert-text-encoder --bundle-resources-for-swift-cli --attention-implementation {SPLIT_EINSUM or ORIGINAL}`
 
 
 
 
 
 
 
24
 
25
- ## Model version: `split_einsum` VS `original`
26
 
27
- Depending on what compute unit you select, you will need to use the correct model version:
28
- - `split_einsum` is compatible with all compute unit
29
- - `original` is only compatible with CPU & GPU
30
 
31
  ## Usage
32
 
33
- Once the chosen model has been downloaded, simply unzip it to use it.
 
 
 
 
 
34
 
35
  <hr>
36
 
@@ -51,15 +63,16 @@ Models have the following naming scheme:
51
  1. Original model name
52
  1. Model version (`split-einsum` or `original`)
53
  1. Model size (only if different from `512x512`)
54
- 1. VAE name (only if different from the original VAE)
 
55
 
56
  Each label is separated by an underscore `_`, and all capitalization from the original name is preserved.\
57
- For example: `stable-diffusion-2-1_original_512x768_ema-vae`.
58
 
59
  ## Repo Name
60
 
61
- Repos are named with the original diffusers Hugging Face / Civitai repo name prefixed by `coreml-`.\
62
- For example: `coreml-stable-diffusion-2-1`.
63
 
64
  ## Repo README Contents
65
 
@@ -72,6 +85,7 @@ tags:
72
  - coreml
73
  - stable-diffusion
74
  - text-to-image
 
75
  ---
76
 
77
  # Core ML converted model
 
17
 
18
  <hr>
19
 
20
+ ## Base Model Types
21
 
22
+ Model files with `split_einsum` in the file name are compatible with all compute units.
23
+
24
+ Model files with `original` in the file name are only compatible with CPU & GPU.
25
+
26
+ Model files with a `no-i2i` suffix in the file name only work for Text2Image.
27
+
28
+ Models with a `cn` suffix in the file name (or in their repo name) will work for Text2Image, Image2Image and ControlNet.
29
+
30
+ Model files with neither a `no-i2i` nor a `cn` suffix in the file name will work for Text2Image and Image2Image.
31
 
32
+ **Other Model Types**
33
 
34
+ If you are using Mochi Diffusion v3.2, v4.0, or later versions, some model files with neither a `no-i2i` nor a `_cn` suffix in the file name might need a simple modification to enable Image2Image to work correctly. Please go [HERE](https://huggingface.co/coreml/VAEEncoders) for more information.
35
+
36
+ The various ControlNet types supported by Mochi Diffusion v4.0 and later are enabled by individual ControlNet models. Please go [HERE](https://huggingface.co/coreml/ControlNet-Models-For-Core-ML) for a selection of ControlNet models that you can download, and information on their use.
37
 
38
  ## Usage
39
 
40
+ Once the chosen base model has been downloaded, simply unzip it and put it in your model folder to use it.
41
+
42
+ ## Conversion Flags
43
+
44
+ The models were converted using the following flags:\
45
+ `--convert-vae-decoder --convert-vae-encoder --convert-unet --unet-support-controlnet --convert-text-encoder --bundle-resources-for-swift-cli --attention-implementation {SPLIT_EINSUM or ORIGINAL}`
46
 
47
  <hr>
48
 
 
63
  1. Original model name
64
  1. Model version (`split-einsum` or `original`)
65
  1. Model size (only if different from `512x512`)
66
+ 1. VAE name (only if different from the original `VAE`)
67
+ 1. cn (for models compatible with `ControlNet`)
68
 
69
  Each label is separated by an underscore `_`, and all capitalization from the original name is preserved.\
70
+ For example: `stable-diffusion-1-5_original_512x768_ema-vae_cn`.
71
 
72
  ## Repo Name
73
 
74
+ Repos are named with the original diffusers Hugging Face / Civitai repo name prefixed by `coreml-` and have a `_cn` suffix if they are ControlNet compatibile.\
75
+ For example: `coreml-stable-diffusion-1-5_cn`.
76
 
77
  ## Repo README Contents
78
 
 
85
  - coreml
86
  - stable-diffusion
87
  - text-to-image
88
+ - not-for-all-eyes (models where viewer discretion is advised)
89
  ---
90
 
91
  # Core ML converted model