Instructions to use UT-IFML/Nemotron-Labs-Diffusion-3B-PIG with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use UT-IFML/Nemotron-Labs-Diffusion-3B-PIG with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="UT-IFML/Nemotron-Labs-Diffusion-3B-PIG", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("UT-IFML/Nemotron-Labs-Diffusion-3B-PIG", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use UT-IFML/Nemotron-Labs-Diffusion-3B-PIG with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "UT-IFML/Nemotron-Labs-Diffusion-3B-PIG"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UT-IFML/Nemotron-Labs-Diffusion-3B-PIG",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/UT-IFML/Nemotron-Labs-Diffusion-3B-PIG

SGLang

How to use UT-IFML/Nemotron-Labs-Diffusion-3B-PIG with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "UT-IFML/Nemotron-Labs-Diffusion-3B-PIG" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UT-IFML/Nemotron-Labs-Diffusion-3B-PIG",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "UT-IFML/Nemotron-Labs-Diffusion-3B-PIG" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UT-IFML/Nemotron-Labs-Diffusion-3B-PIG",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use UT-IFML/Nemotron-Labs-Diffusion-3B-PIG with Docker Model Runner:
```
docker model run hf.co/UT-IFML/Nemotron-Labs-Diffusion-3B-PIG
```

Nemotron-Labs-Diffusion-3B (instruct) with PIG

Nemotron-Labs-Diffusion-3B, NVIDIA's instruct model, with Parallel Inconsistency Guidance (PIG), which lets the model commit more tokens per forward pass at the same accuracy. This repository is the whole model: the backbone, a 1.9M-parameter PIG adapter and its calibration tables.

Paper: PIG: Parallel Inconsistency Guidance for Accelerating Diffusion Language Models, Fruytier, Xue, Shakkottai, Sanghavi (UT Austin), 2026; arXiv:2610.17375. Code: https://github.com/Quentin-Fruytier/PIG-DLM-release Companion: UT-IFML/LLaDA-8B-Instruct-PIG

Use

Needs transformers 5.x and a GPU with 8 GB.

from transformers import AutoModel

model = AutoModel.from_pretrained("UT-IFML/Nemotron-Labs-Diffusion-3B-PIG", trust_remote_code=True)
out = model.generate("Write a Python function that returns the n-th Fibonacci number.", mode="pig.ext", theta=0.25)
print(out.text, out.tpf)          # tpf: generated tokens per forward pass

argument	what it does	default
`mode`	`"pig.int"`, `"pig.ext"`, or `"confidence"` (the Fast-dLLM baseline)	`"pig.int"`
`phi`	PIG.INT budget; larger is faster, smaller more accurate (paper range 0.015 to 3)	`0.1`
`theta`	PIG.EXT budget; same direction (paper range 0.125 to 3)	`0.5`
`floor`	PIG.EXT: never commit below this confidence	none
`nu`	baseline: commit everything above this confidence	`0.9`
`geometry`, `block`	which positions are candidates: `"block"`, `"sliding"` or `"nested"`	`"block"`, `32`

Implementation details, every argument, and how the adapter and tables were fitted: AGENTS.md.

License

Licensed by NVIDIA Corporation under the NVIDIA Nemotron Model License: the backbone's weights, tokenizer and code are NVIDIA's (LICENSE, NOTICE.md). The PIG adapter, tables and wrapper code are released under the MIT License. NVIDIA's names and marks are used only to identify the model.

Citation

@misc{fruytier2026pig,
  title  = {{PIG}: Parallel Inconsistency Guidance for Accelerating Diffusion Language Models},
  author = {Fruytier, Quentin and Xue, Anton and Shakkottai, Sanjay and Sanghavi, Sujay},
  year   = {2026},
  eprint = {2610.17375},
  archivePrefix = {arXiv},
  primaryClass = {cs.CL},
  url    = {https://arxiv.org/abs/2610.17375}
}

Downloads last month: 82

Safetensors

Model size

4B params

Tensor type

BF16

I64

F64

F32

Model tree for UT-IFML/Nemotron-Labs-Diffusion-3B-PIG

Base model

nvidia/Nemotron-Labs-Diffusion-3B

Adapter

(1)

this model

Paper for UT-IFML/Nemotron-Labs-Diffusion-3B-PIG

PIG: Parallel Inconsistency Guidance for Accelerating Diffusion Language Models

Paper • 2610.17375 • Published 4 days ago