import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("DeltaTechStudios/CourtOfTheAbsurd-host-evidence", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Court of the Absurd — optional evidence model pack
This repository distributes the optional host-evidence 2.1.0 pack used by
Court of the Absurd. It is not a standalone image application. The game
downloads the ZIP, verifies its immutable archive SHA-256, safely extracts it,
then verifies the complete signed per-file manifest before activation.
Release artifact
- File:
CourtOfTheAbsurd-host-evidence-2.1.0.zip - Installed size: 17,082,671,641 bytes (15.909 GiB)
- Download size: 12,831,802,002 bytes (11.951 GiB)
- SHA-256:
48CE27377D1C482B980291C51778249EB2DC6E569695AB77CB65C748DFC10A5B - Manifest:
host-evidence/pack.manifest.json
The pack contains the Apache-2.0 FLUX.2 Klein 4B components pinned from
black-forest-labs/FLUX.2-klein-4B, Photoroom's pinned Diffusers/TorchAO
static-FP8 conversion, and an isolated pinned Windows Python/PyTorch runtime.
Exact upstream revisions are recorded in the game's
Tools/LocalAI/sources.lock.json.
Runtime profile
The model accepts one to four reference images and is run at four inference steps. The full-GPU profile is intended for 24 GB RTX 3090/4090-class hosts. The lower-memory profile keeps the compact static-FP8 transformer resident on the GPU and offloads text-encoder/VAE weights; its measured PyTorch allocation peak was 7.94 GiB on the production smoke test.
FLUX.2 Klein 4B and its supplied license are Apache-2.0. TorchAO is BSD-3-Clause. Upstream attribution and license files are retained in the pack.
- Downloads last month
- -