metadata
tags:
- text-to-image
- lora
- diffusers
- template:diffusion-lora
widget:
- output:
url: images/kwis_android.png
text: >-
Photo of an android kwis man with mechanical joints and limbs straight
from a Phillip K. Dick novel but with a human-looking face and eyes. The
kwis man is looking at the camera with a sad expression, (((wires and
cables connecting to head:1.2))), ((mechanical limbs:0.8), ((mechanical
cervical attaching to the neck)),
parameters:
negative_prompt: >-
blurry, cartoon, poorly drawn, fused fingers, deformed hands, text,
watermark, jpeg artifacts,
base_model: lodestones/Chroma1-Base
instance_prompt: kwis
license: mit
chroma-kwis-v1
- Prompt
- Photo of an android kwis man with mechanical joints and limbs straight from a Phillip K. Dick novel but with a human-looking face and eyes. The kwis man is looking at the camera with a sad expression, (((wires and cables connecting to head:1.2))), ((mechanical limbs:0.8), ((mechanical cervical attaching to the neck)),
- Negative Prompt
- blurry, cartoon, poorly drawn, fused fingers, deformed hands, text, watermark, jpeg artifacts,
Model description
Rank 32 LoRA using https://github.com/tdrussell/diffusion-pipe optim.AdamW at an LR of 2e-4, batch size of six on a single Nvidia L40S for 40 epochs.
https://api.wandb.ai/links/zakack/utmlycsw
Trigger words
You should use kwis to trigger the image generation.
Weights & Biases Metrics
https://api.wandb.ai/links/zakack/utmlycsw
Download model
Download them in the Files & versions tab.