import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("mann-e/mann-e_4-2-merged", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Mann-E 4.2 Merged
Technical Information about the model
- Base Model : runwayml/stable-diffusion-v1-5
- Merge : mann-e/mann-e_4_rev-1-3
- Merge amount : %70 fine-tuned SD 1.5 (or Mann-E version 4.2 base) and %30 of Mann-E 4.1.3 in order to get the old styles such as Model Shoot, Elden Ring, Arcane, Analog Style and GTA V Style. Also this merge can be helpful for Midjourney version 4 style artwork as well.
Training process
The code for pre-processing data and fine-tuning the model is available in this repository and you can run it on your own as well.
- Text encoder iterations : 1440 (number of pics times two in order to understand
mstylewhich can give the user a Midjourney version 5 vibe). - Stable Diffusion iterations : 16000 iterations for one epoch
- Time: around 4 hours on a single T4 GPU.
- Downloads last month
- 16