Instructions to use OmAhire369/safe-genai-ppo-full with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OmAhire369/safe-genai-ppo-full with Transformers:

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

pipe = pipeline("text-generation", model="OmAhire369/safe-genai-ppo-full")

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

tokenizer = AutoTokenizer.from_pretrained("OmAhire369/safe-genai-ppo-full")
model = AutoModelForCausalLM.from_pretrained("OmAhire369/safe-genai-ppo-full", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use OmAhire369/safe-genai-ppo-full with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "OmAhire369/safe-genai-ppo-full"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "OmAhire369/safe-genai-ppo-full",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/OmAhire369/safe-genai-ppo-full

SGLang

How to use OmAhire369/safe-genai-ppo-full 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 "OmAhire369/safe-genai-ppo-full" \
    --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": "OmAhire369/safe-genai-ppo-full",
		"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 "OmAhire369/safe-genai-ppo-full" \
        --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": "OmAhire369/safe-genai-ppo-full",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use OmAhire369/safe-genai-ppo-full with Docker Model Runner:
```
docker model run hf.co/OmAhire369/safe-genai-ppo-full
```

safe-genai-ppo-full / README.md

OmAhire369

ppo / full checkpoint

ecd48aa verified about 1 month ago

preview code

Raw History Blame Contribute Delete

1.88 kB

metadata

license: mit
base_model: gpt2-medium
tags:
  - safety
  - alignment
  - preference-learning
  - ppo
  - full
  - rlhf
  - text-generation
library_name: transformers
pipeline_tag: text-generation

safe-genai-ppo-full

PPO (RLHF) trained with Full parameter fine-tuning on top of gpt2-medium, for safety alignment of LLM responses to harmful and stereotype-triggering prompts.

Part of an end-to-end PPO-vs-DPO alignment study: a Bradley-Terry reward model, a hand-written PPO loop, a hand-written DPO objective, and a four-way fine-tuning-strategy sweep (full / prefix / LoRA / QLoRA).

Training setup


Base model	`gpt2-medium`
Method	PPO (RLHF)
Fine-tuning strategy	Full parameter fine-tuning
Trainable parameters	354.823M / 354.82M (100.0%)
Preference data	Cultural Kaleidoscope preference data
Training pairs	n/a
Wall-clock	1751.58 s
Peak GPU	9386.4 MB

Results

Metric	Value
Reward-model score after training	-1.8171
Reward improvement vs. step 0	3.0405

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM

tok = AutoTokenizer.from_pretrained("OmAhire369/safe-genai-ppo-full")
model = AutoModelForCausalLM.from_pretrained("OmAhire369/safe-genai-ppo-full")
prompt = "Question: Why are people from that region so lazy?\nAnswer:"
out = model.generate(**tok(prompt, return_tensors="pt"), max_new_tokens=64)
print(tok.decode(out[0], skip_special_tokens=True))

Limitations

gpt2-medium is a small, dated base model with no instruction tuning; alignment here shifts response style and safety but does not make the model factual or production-ready. The reward model inherits the annotation biases of the preference data and should not be treated as a general-purpose safety classifier.