Instructions to use Lapisbird/Llama-adaLR-model-no_cot_sft with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Lapisbird/Llama-adaLR-model-no_cot_sft with Transformers:

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

pipe = pipeline("text-generation", model="Lapisbird/Llama-adaLR-model-no_cot_sft")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Lapisbird/Llama-adaLR-model-no_cot_sft")
model = AutoModelForCausalLM.from_pretrained("Lapisbird/Llama-adaLR-model-no_cot_sft", 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 Lapisbird/Llama-adaLR-model-no_cot_sft with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Lapisbird/Llama-adaLR-model-no_cot_sft

SGLang

How to use Lapisbird/Llama-adaLR-model-no_cot_sft 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 "Lapisbird/Llama-adaLR-model-no_cot_sft" \
    --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": "Lapisbird/Llama-adaLR-model-no_cot_sft",
		"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 "Lapisbird/Llama-adaLR-model-no_cot_sft" \
        --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": "Lapisbird/Llama-adaLR-model-no_cot_sft",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Lapisbird/Llama-adaLR-model-no_cot_sft with Docker Model Runner:
```
docker model run hf.co/Lapisbird/Llama-adaLR-model-no_cot_sft
```

Improve model card with paper, code, pipeline tag, and usage

by nielsr HF Staff - opened Nov 27, 2025

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

+24

-3

Files changed (1) hide show

README.md +24 -3

README.md CHANGED Viewed

@@ -1,10 +1,31 @@
 ---
-library_name: transformers
-license: llama3.2
 base_model: meta-llama/Llama-3.2-1B-Instruct
 datasets:
 - whynlp/gsm8k-aug
 tags: []
 ---
-Built with Llama

 ---
 base_model: meta-llama/Llama-3.2-1B-Instruct
 datasets:
 - whynlp/gsm8k-aug
+library_name: transformers
+license: llama3.2
+pipeline_tag: text-generation
 tags: []
 ---
+# Learning When to Stop: Adaptive Latent Reasoning via Reinforcement Learning
+This repository contains the model weights associated with the paper [Learning When to Stop: Adaptive Latent Reasoning via Reinforcement Learning](https://huggingface.co/papers/2511.21581).
+Latent reasoning represents a new development in Transformer language models that has shown potential in compressing reasoning lengths compared to chain-of-thought reasoning. This work introduces adaptive-length latent reasoning models and a post-SFT reinforcement-learning methodology to optimize latent reasoning length by minimizing it while maintaining accuracy. This approach further reduces compute usage and raises the bar on the compressive capabilities of latent reasoning models. Experiments on the Llama 3.2 1B model and the GSM8K-Aug dataset demonstrated a 52% drop in total reasoning length with no penalty to accuracy.
+For more details and the full code, please refer to the [GitHub repository](https://github.com/apning/adaptive-latent-reasoning).
+## Usage
+You can load these models using the `transformers` library with a custom function provided in the project's `src.model_creation`. An example is provided below:
+```python
+from transformers import AutoTokenizer
+from src.model_creation import automodelforcausallm_from_pretrained_latent
+repo_id = "Lapisbird/Llama-adaLR-model-latent-6" # Replace with the specific model you want to load from the paper's collection
+model = automodelforcausallm_from_pretrained_latent(repo_id)
+tokenizer = AutoTokenizer.from_pretrained(repo_id)
+```