Core ML Converted Model:
- This model was converted to Core ML for use on Apple Silicon devices. Conversion instructions can be found here.
- Provide the model to an app such as Mochi Diffusion Github - Discord to generate images.
split_einsumversion is compatible with all compute unit options including Neural Engine.originalversion is only compatible with CPU & GPU option.- Custom resolution versions are tagged accordingly.
vaetagged files have a vae embedded into the model.- Descriptions are posted as-is from original model source. Not all features and/or results may be available in CoreML format.
- This model was converted with
vae-encoderfor i2i.
Note: Some models do not have the unet split into chunks.
miniSD:
Source(s): Hugging Face
Training details
Fine tuned from the stable-diffusion 1.4 checkpoint as follows:
-
22,000 steps fine-tuning only the attention layers of the unet, learn rate=1e-5, batch size=256
-
66,000 steps training the full unet, learn rate=5e-5, batch size=552
-
GPUs provided by Lambda GPU Cloud
-
Trained on LAION Improved Aesthetics 6plus.
-
Trained using https://github.com/justinpinkney/stable-diffusion, original checkpoint available here
License
This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies:
- You can't use the model to deliberately produce nor share illegal or harmful outputs or content
- The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
- You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here