Instructions to use gopi30/rlaif-safety-alligned with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use gopi30/rlaif-safety-alligned with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("ybelkada/falcon-7b-sharded-bf16")
model = PeftModel.from_pretrained(base_model, "gopi30/rlaif-safety-alligned")

Transformers

How to use gopi30/rlaif-safety-alligned with Transformers:

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

pipe = pipeline("text-generation", model="gopi30/rlaif-safety-alligned")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("gopi30/rlaif-safety-alligned", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use gopi30/rlaif-safety-alligned with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gopi30/rlaif-safety-alligned"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "gopi30/rlaif-safety-alligned",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/gopi30/rlaif-safety-alligned

SGLang

How to use gopi30/rlaif-safety-alligned 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 "gopi30/rlaif-safety-alligned" \
    --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": "gopi30/rlaif-safety-alligned",
		"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 "gopi30/rlaif-safety-alligned" \
        --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": "gopi30/rlaif-safety-alligned",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use gopi30/rlaif-safety-alligned with Docker Model Runner:
```
docker model run hf.co/gopi30/rlaif-safety-alligned
```

Model Card for LegalSafe-Falcon-7B

Model Details

Model Description

LegalSafe-Falcon-7B is a safety-aligned large language model fine-tuned for the Indian legal domain. The model focuses on reducing toxic, biased, and unsafe outputs while maintaining high accuracy in legal reasoning and responses. It integrates Constitutional AI principles and Reinforcement Learning from AI Feedback (RLAIF) to improve ethical alignment.

Developed by: Gopi M
Model type: Causal Language Model (LLM)
Language(s): English (Legal Domain - Indian Context)
License: Apache 2.0 (inherits from base model)
Finetuned from model: ybelkada/falcon-7b-sharded-bf16

Uses

Direct Use

Legal question answering
Legal document summarization
Safe text generation in sensitive domains
Educational tools for legal studies

Downstream Use

Integration into legal chatbots
AI-based legal assistants
Compliance and advisory systems
Safety-critical NLP applications

Out-of-Scope Use

Providing official legal advice
Use in high-risk legal decision-making without human oversight
Generating harmful, biased, or illegal content

Bias, Risks, and Limitations

May still exhibit residual bias from training data
Performance may vary outside Indian legal domain
Not a substitute for professional legal consultation
Risk of hallucinations in complex legal scenarios

Recommendations

Use with human oversight in legal applications
Validate outputs before deployment in critical systems
Avoid relying on model for final legal decisions

How to Get Started

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "gopi30/rlaif-safety-alligned"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

prompt = "Explain the concept of fundamental rights in Indian law."
inputs = tokenizer(prompt, return_tensors="pt")

outputs = model.generate(**inputs, max_new_tokens=200)
print(tokenizer.decode(outputs[0]))

Downloads last month: 1

Model tree for gopi30/rlaif-safety-alligned

Base model

ybelkada/falcon-7b-sharded-bf16

Adapter

(134)

this model