Instructions to use yingfanbot/gsm-lotus-llama3b-nl with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yingfanbot/gsm-lotus-llama3b-nl with Transformers:

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

pipe = pipeline("text-generation", model="yingfanbot/gsm-lotus-llama3b-nl")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("yingfanbot/gsm-lotus-llama3b-nl")
model = AutoModelForCausalLM.from_pretrained("yingfanbot/gsm-lotus-llama3b-nl", device_map="auto")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use yingfanbot/gsm-lotus-llama3b-nl with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yingfanbot/gsm-lotus-llama3b-nl"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yingfanbot/gsm-lotus-llama3b-nl",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/yingfanbot/gsm-lotus-llama3b-nl

SGLang

How to use yingfanbot/gsm-lotus-llama3b-nl 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 "yingfanbot/gsm-lotus-llama3b-nl" \
    --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": "yingfanbot/gsm-lotus-llama3b-nl",
		"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 "yingfanbot/gsm-lotus-llama3b-nl" \
        --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": "yingfanbot/gsm-lotus-llama3b-nl",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use yingfanbot/gsm-lotus-llama3b-nl with Docker Model Runner:
```
docker model run hf.co/yingfanbot/gsm-lotus-llama3b-nl
```

gsm-lotus-llama3b-nl / README.md

yingfanbot

Card: add verified GSM8K accuracy (68.54%)

9a4fbb8 verified about 1 month ago

preview code

Raw

History Blame Contribute Delete

1.66 kB

	---
	license: mit
	base_model: meta-llama/Llama-3.2-3B-Instruct
	library_name: transformers
	pipeline_tag: text-generation
	tags: [latent-reasoning, looped-transformer, gsm8k, lotus]
	---

	# gsm-lotus-llama3b-nl

	LOTUS trained on natural-language chain-of-thought, fine-tuned from `meta-llama/Llama-3.2-3B-Instruct`,
	from the paper
	[Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers](https://arxiv.org/abs/2606.31779).
	Direct step-aligned supervision through the base LM head (no aux decoder, no CODI); the latent blocks are
	supervised against natural-language reasoning steps rather than formatted math expressions.

	- GSM8K (GSM8k-Aug) test accuracy: 68.54% (904/1319)
	- Latent config: K = 6 blocks, c_thought = 50 tokens/block (NL steps are ~2.7x longer than math-expr,
	so c=50 covers ~98.8% of steps), n_looped_iters = 6
	- Base: meta-llama/Llama-3.2-3B-Instruct (vocab 128256 -> 128259 for 3 latent tokens)

	## Loading
	`from_pretrained` loads the weights only — the looped padded architecture needs the LOTUS code
	([code repo](https://github.com/yingfan-bot/lotus)).

	## Reproduce the number above
	Run in the pinned env (torch 2.7 / transformers 4.46.2) — note `--c_thought 50` for this NL model:
	```bash
	python scripts/eval.py \
	--model_id yingfanbot/gsm-lotus-llama3b-nl \
	--datasets gsm8k --n_looped_iters 6 --c_thought 50 --bf16
	```

	## Citation
	```bibtex
	@article{fan2026bridging,
	title={Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers},
	author={Fan, Ying and Svete, Anej and Lee, Kangwook},
	journal={arXiv preprint arXiv:2606.31779},
	year={2026}
	}
	```