| license: gpl-3.0 | |
| This a an fp16 variant of Proteus V2.0 | |
| https://huggingface.co/dataautogpt3/ProteusV0.2 | |
| currently under the gpl-v3 licence. | |
| simply created by | |
| ```py | |
| import torch | |
| from diffusers import DiffusionPipeline | |
| pipeline = DiffusionPipeline.from_pretrained("dataautogpt3/ProteusV0.2", torch_dtype=torch.float16) | |
| pipeline.save_pretrained('fp16_ProteusV0.2', safe_serialization=True, variant='fp16') | |
| ``` | |
| See the original model for details. | |
| The fp32 version of the model, even when converted to fp16 when loading, uses up to much RAM | |
| hence my need for this version. | |
| Dave | |