Benign huggingface Download

ProCreations/Image-2.1-Calibrated-FP8 main

Python imports the hashlib libraryPython module-level function call

SHA-25651141918352c624189128e0dc6ebd6d7e36df750f7c5c23af8671a8d366411ff

URLhttps://huggingface.co/ProCreations/Image-2.1-Calibrated-FP8

Evidence

Python imports the hashlib library lines 1–14

1import json,time,hashlib,collections
2from pathlib import Path
3import torch
4from safetensors.torch import save_file
5from fp8_runtime import load_pipeline,eligible_modules,quantize_activation
6from prompts import CALIBRATION,EDIT_CALIBRATION
7ROOT=Path(__file__).parent
8OUT=ROOT/'calibration'; OUT.mkdir(exist_ok=True)
9BASE='/home/user/models/qwen-image-2.1-b3179ad'
10
⋯4 lines

English function-word token "for" lines 18–21

41… n_features),dtype=torch.bfloat16,device='cuda') for n,m in modules.items()}
  pos=collections.defaultdict(int)
  ctx={'step':0,'prompt':0}
   …

English function-word token "the" lines 25–27

5…         # Deterministic spatial/token sampling, distributed across the sequence.
          offset=(ctx['prompt']*97+ctx['step']*31)%max(1,x.shape[0]//4)
          …