BiliSakura/ADM-diffusers

Self-contained OpenAI ADM-G checkpoints for Hugging Face diffusers. No external code repo is required — each subfolder ships its own pipeline.py, component modules, and weights.

This repo is derived from the development bundle in Visual-Generative-Foundation-Model-Collection, but inference only needs:

• This model repo (BiliSakura/ADM-diffusers)
• PyPI diffusers, torch, huggingface_hub

This Hugging Face repo hosts multiple self-contained checkpoints as subfolders. Each subfolder includes its own pipeline.py, model_index.json, weights, and component code (unet/, classifier/, scheduler/).

Available checkpoints

Subfolder	Resolution	Classifier scale	OpenAI sources
ADM-G-256/	256×256	1.0	256x256_diffusion.pt + 256x256_classifier.pt
ADM-G-512/	512×512	4.0	512x512_diffusion.pt + 512x512_classifier.pt

Both resolutions use the class-conditional diffusion checkpoint plus the noisy classifier (not the 256 uncond variant).

ImageNet class labels

Each variant keeps an id2label map directly in its own model_index.json (same style as DiT on the Hub). Runtime label resolution is English-only:

• pipe.id2label — inspect id → English label correspondence
• pipe.labels — reverse map (English synonym → id), sorted for browsing
• pipe.get_label_ids("golden retriever")
• pipe(class_labels="golden retriever", ...)

Chinese labels are still preserved in the main source repo under src/labels/id2label_cn.json for reference.

Demo

Load from Hugging Face

import sys
from pathlib import Path
import torch
from huggingface_hub import snapshot_download

repo_dir = Path(snapshot_download("BiliSakura/ADM-diffusers"))
variant = "ADM-G-512"  # or "ADM-G-256"

sys.path.insert(0, str(repo_dir / variant))
from pipeline import ADMPipeline

pipe = ADMPipeline.from_pretrained(".")
pipe.to("cuda")
pipe.unet.float()
pipe.classifier.float()
pipe.classifier.model.dtype = torch.float32

images = pipe(
    class_labels=207,
    num_inference_steps=250,
    classifier_guidance_scale=4.0 if variant == "ADM-G-512" else 1.0,
).images

# Human-readable ImageNet labels (English)
print(pipe.id2label[207])          # "golden retriever"
pipe.get_label_ids("golden retriever")  # [207]
images = pipe(class_labels="golden retriever", classifier_guidance_scale=1.0).images

Load from a local clone

import sys
from pathlib import Path

repo = Path("BiliSakura/ADM-diffusers").resolve()
variant = "ADM-G-256"

sys.path.insert(0, str(repo / variant))
from pipeline import ADMPipeline

pipe = ADMPipeline.from_pretrained(".")
pipe.to("cuda")

Repo layout

BiliSakura/ADM-diffusers/
├── README.md
├── ADM-G-256/
│   ├── pipeline.py
│   ├── model_index.json
│   ├── unet/
│   ├── classifier/
│   └── scheduler/
└── ADM-G-512/
    ├── pipeline.py
    ├── model_index.json
    ├── demo.png
    ├── unet/
    ├── classifier/
    └── scheduler/