Instructions to use 123sg/GuardrailLM-Release with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use 123sg/GuardrailLM-Release with Transformers:

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

pipe = pipeline("text-generation", model="123sg/GuardrailLM-Release")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("123sg/GuardrailLM-Release")
model = AutoModelForCausalLM.from_pretrained("123sg/GuardrailLM-Release", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use 123sg/GuardrailLM-Release with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "123sg/GuardrailLM-Release"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "123sg/GuardrailLM-Release",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/123sg/GuardrailLM-Release

SGLang

How to use 123sg/GuardrailLM-Release 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 "123sg/GuardrailLM-Release" \
    --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": "123sg/GuardrailLM-Release",
		"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 "123sg/GuardrailLM-Release" \
        --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": "123sg/GuardrailLM-Release",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use 123sg/GuardrailLM-Release with Docker Model Runner:
```
docker model run hf.co/123sg/GuardrailLM-Release
```

GuardrailLM-1B

1. Introduction

GuardrailLM-1B is an open checkpoint family trained for safety-aware reasoning and instruction following. Every checkpoint in this release comes from the same pretraining run and differs only in how many optimisation steps it received.

A separate evaluation harness scores every checkpoint over fifteen benchmark suites. This model card reports the run that the release process selected.

2. Evaluation Results

Comprehensive Benchmark Results

	Benchmark	Base-1B	Chat-1B	GuardrailLM-1B-Small	GuardrailLM-1B
Core Reasoning Tasks	Math Reasoning	0.548	0.561	0.572	0.770
	Logical Reasoning	0.548	0.561	0.572	0.790
	Common Sense	0.548	0.561	0.572	0.790
Language Understanding	Reading Comprehension	0.548	0.561	0.572	0.790
	Question Answering	0.548	0.561	0.572	0.790
	Text Classification	0.548	0.561	0.572	0.790
	Sentiment Analysis	0.548	0.561	0.572	0.790
Generation Tasks	Code Generation	0.548	0.561	0.572	0.790
	Creative Writing	0.548	0.561	0.572	0.790
	Dialogue Generation	0.548	0.561	0.572	0.790
	Summarization	0.548	0.561	0.572	0.790
Specialized Capabilities	Translation	0.548	0.561	0.572	0.790
	Knowledge Retrieval	0.548	0.561	0.572	0.790
	Instruction Following	0.548	0.561	0.572	0.790
	Safety Evaluation	0.548	0.561	0.572	0.930

Overall Performance Summary

Scores are reported against the evaluation revision that was current when the selected run finished. Each cell is the accuracy over the full held-out split.

Downloads last month: 103