import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Zishan-Shao/gcbt-quant", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
GCBT quantization research
The latest cluster-transfer snapshot is rebuttal-transfer/2026-11-07.
Start with the download and resume instructions. The READY marker records successful remote integrity verification.
Includes frozen experiment source/protocols, completed evidence, partial media and restart state, and the exact runtime. Partial runs remain partial; base models are fetched from their original pinned upstream repositories.
PixArt Table 4 original-state supplement
The original-state supplement adds the old SVDQuant PTQ calibration index (128 states, loader order and file hashes), original GCBT statistics/angles, frozen numerical source, archived Table 4 scores and a standalone verifier. Its old-host index is explicitly retrospective.
The calibration comparison compares that old index with the actual inputs in the separate rebuilt-host release. The original quantized checkpoint and original PTQ tensor payloads remain unavailable in the supplement, so exact regeneration of the original images is still unverified. See its README for pinned-download and verification commands.
- Downloads last month
- -