Instructions to use mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Nanbeige/Nanbeige4-3B-Base")
model = PeftModel.from_pretrained(base_model, "mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA")

Transformers

How to use mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA with Transformers:

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

pipe = pipeline("text-generation", model="mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA

SGLang

How to use mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA 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 "mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA with Docker Model Runner:
```
docker model run hf.co/mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA
```

Model Card for nanbeige4-3b-lora-v1

This model is a fine-tuned version of Nanbeige/Nanbeige4-3B-Base. It has been trained using TRL.

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="None", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with SFT.

Framework versions

PEFT 0.18.1
TRL: 0.29.0
Transformers: 5.2.0
Pytorch: 2.10.0
Datasets: 4.6.1
Tokenizers: 0.22.2

Citations

Cite TRL as:

@software{vonwerra2020trl,
  title   = {{TRL: Transformers Reinforcement Learning}},
  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
  license = {Apache-2.0},
  url     = {https://github.com/huggingface/trl},
  year    = {2020}
}

Downloads last month: 2

Model tree for mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA

Base model

Nanbeige/Nanbeige4-3B-Base

Adapter

(3)

this model

Dataset used to train mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA

Collection including mrinaalarora/Nanbeige4-3B-Cold-Start-Reasoning-LoRA

Nanbeige4-3B Cold Start Reasoning LoRA Experiments

Collection

Two LoRA cold-start SFT experiments teaching structured think/answer reasoning to Nanbeige4-3B-Base using distilled traces from frontier models • 3 items • Updated Mar 13