SiT-diffusers

Diffusers-ready checkpoints for Scalable Interpolant Transformers (SiT), converted for local/offline use.

This root folder is a model collection that contains:

• SiT-S-2-256
• SiT-B-2-256
• SiT-L-2-256
• SiT-XL-2-256
• SiT-XL-2-512

Each subfolder is a self-contained Diffusers model repo with:

• pipeline.py
• transformer/transformer_sit.py
• scheduler/scheduler_config.json (FlowMatchEulerDiscreteScheduler)
• transformer/diffusion_pytorch_model.safetensors
• vae/diffusion_pytorch_model.safetensors

Each variant embeds English id2label directly in model_index.json (DiT-style), so class labels can be passed as ImageNet ids or English synonym strings.

Demo

Class-conditional sample (ImageNet class 207, golden retriever), SiT-XL/2 at 512×512, 250 steps, CFG 4.0, seed 0.

Model Paths

Use paths relative to this root README:

Model	Resolution	Local path
SiT-S/2	256x256	./SiT-S-2-256
SiT-B/2	256x256	./SiT-B-2-256
SiT-L/2	256x256	./SiT-L-2-256
SiT-XL/2	256x256	./SiT-XL-2-256
SiT-XL/2	512x512	./SiT-XL-2-512

Inference Demo (Diffusers)

1) Load a local subfolder checkpoint

import torch
from diffusers import DiffusionPipeline

model_path = "./SiT-XL-2-512"  # change to any path in the table above
device = "cuda" if torch.cuda.is_available() else "cpu"

pipe = DiffusionPipeline.from_pretrained(
    model_path,
    trust_remote_code=True,
).to(device)

generator = torch.Generator(device=device).manual_seed(0)

# ImageNet class example: 207 = golden retriever
print(pipe.id2label[207])
print(pipe.get_label_ids("golden retriever"))  # [207]
result = pipe(
    class_labels="golden retriever",
    height=512,
    width=512,
    num_inference_steps=250,  # official SiT comparisons commonly use 250 steps
    guidance_scale=4.0,
    generator=generator,
)

image = result.images[0]
image.save("sit_xl_512_demo.png")

2) Quick variant switch (256 models)

model_path = "./SiT-S-2-256"
# model_path = "./SiT-B-2-256"
# model_path = "./SiT-L-2-256"
# model_path = "./SiT-XL-2-256"

pipe = DiffusionPipeline.from_pretrained(model_path, trust_remote_code=True).to(device)
image = pipe(
    class_labels=207,
    height=256,
    width=256,
    num_inference_steps=250,
    guidance_scale=4.0,
    generator=generator,
).images[0]
image.save("sit_256_demo.png")