Instructions to use GSAI-ML/ReFusion with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use GSAI-ML/ReFusion with Transformers:

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

pipe = pipeline("text-generation", model="GSAI-ML/ReFusion", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("GSAI-ML/ReFusion", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("GSAI-ML/ReFusion", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Local Apps Settings

vLLM

How to use GSAI-ML/ReFusion with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/GSAI-ML/ReFusion

SGLang

How to use GSAI-ML/ReFusion 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 "GSAI-ML/ReFusion" \
    --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": "GSAI-ML/ReFusion",
		"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 "GSAI-ML/ReFusion" \
        --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": "GSAI-ML/ReFusion",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use GSAI-ML/ReFusion with Docker Model Runner:
```
docker model run hf.co/GSAI-ML/ReFusion
```

Add `library_name: transformers` to metadata

by nielsr HF Staff - opened Dec 21, 2025

base: refs/heads/main

←

from: refs/pr/2

Discussion Files changed

+12

-3

Files changed (1) hide show

README.md +12 -3

README.md CHANGED Viewed

@@ -1,11 +1,12 @@
 ---
 license: apache-2.0
 tags:
 - dllm
 - diffusion
 - llm
 - text_generation
-pipeline_tag: text-generation
 ---
 # ReFusion
@@ -91,7 +92,7 @@ def generate_refusion(model, tokenizer, prompt, gen_length=128, temperature=0.,
         cur_gen_pos_ids = gen_pos_ids[:, serial_num_block*block_length:(serial_num_block+1)*block_length] # (batch_size, block_length)
         cur_gen_blocks_x = cur_gen_x.reshape(batch_size, -1, cur_slot_size)
-        cur_gen_blocks_pos_ids = cur_gen_pos_ids.reshape(batch_size, -1, cur_slot_size)
         # slot level generation
         while cur_gen_blocks_x.numel() > 0:
@@ -416,7 +417,15 @@ def main():
     model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, torch_dtype=torch.bfloat16).to(device).eval()
     tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
-    prompt = "You are an expert Python programmer. Your task is to write a single Python function to solve the problem described below, and here is your task: Write a function to sum all amicable numbers from 1 to a specified number.\n\nDirectly after the '[BEGIN]' marker, you must write only the Python code for the function. Do not provide any explanations, comments, or introductory text. The function must include the 'def' line, its arguments, the function body, and a 'return' statement. Your code should pass these tests:\n\nassert amicable_numbers_sum(999)==504\nassert amicable_numbers_sum(9999)==31626\nassert amicable_numbers_sum(99)==0\n[BEGIN]\n"
     m = [{"role": "user", "content": prompt}, ]
     prompt = tokenizer.apply_chat_template(m, add_generation_prompt=True, tokenize=False, enable_thinking=True)

 ---
 license: apache-2.0
+pipeline_tag: text-generation
 tags:
 - dllm
 - diffusion
 - llm
 - text_generation
+library_name: transformers
 ---
 # ReFusion
         cur_gen_pos_ids = gen_pos_ids[:, serial_num_block*block_length:(serial_num_block+1)*block_length] # (batch_size, block_length)
         cur_gen_blocks_x = cur_gen_x.reshape(batch_size, -1, cur_slot_size)
+        cur_gen_blocks_pos_ids = cur_gen_blocks_pos_ids.reshape(batch_size, -1, cur_slot_size)
         # slot level generation
         while cur_gen_blocks_x.numel() > 0:
     model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, torch_dtype=torch.bfloat16).to(device).eval()
     tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+    prompt = "You are an expert Python programmer. Your task is to write a single Python function to solve the problem described below, and here is your task: Write a function to sum all amicable numbers from 1 to a specified number.
+Directly after the '[BEGIN]' marker, you must write only the Python code for the function. Do not provide any explanations, comments, or introductory text. The function must include the 'def' line, its arguments, the function body, and a 'return' statement. Your code should pass these tests:
+assert amicable_numbers_sum(999)==504
+assert amicable_numbers_sum(9999)==31626
+assert amicable_numbers_sum(99)==0
+[BEGIN]
+"
     m = [{"role": "user", "content": prompt}, ]
     prompt = tokenizer.apply_chat_template(m, add_generation_prompt=True, tokenize=False, enable_thinking=True)