Kalista is a state-of-the-art stylized realism model that is extremely consistent in generating high quality outputs!
This model was meant to be a quick experiment on 'simulating' additional training steps by finding the mean values of multiple different checkpoints that were trained on separate datasets.
While prompting did prove quite difficult to maneuver as compared to other models, I was blown away by the consistency of the raw quality of the images generated.
since my GPU is not powerful enough to generate high resolution images I'm completely open to include your art on Kalista's repository over at https://huggingface.co/skylersterling/kalista
The below images were generated in succession to demonstrate Kalista's consistency:
描述:
训练词语:
名称: kalista_type1.ckpt
大小 (KB): 3761850
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: **Detected Pickle imports (3)**
```
collections.OrderedDict
torch.HalfStorage
torch._utils._rebuild_tensor_v2
```
病毒扫描结果: Success
