Instructions to use NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel with Transformers:

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

pipe = pipeline("text-generation", model="NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel")

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel")
model = AutoModelForMultimodalLM.from_pretrained("NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel

SGLang

How to use NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel 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 "NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel" \
    --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": "NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel",
		"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 "NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel" \
        --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": "NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel with Docker Model Runner:
```
docker model run hf.co/NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

GRPO RPG System 3.2 1B "Scalpel"

Overview

GRPO RPG System 3.2 1B Scalpel is a lightweight experimental merge based on Ultimate-RPG.System-3.2-1B and a GRPO-optimized conversational model.

The objective of this merge is to preserve the narrative strengths, roleplay capabilities and worldbuilding consistency of the original RPG System while incorporating part of the conversational flexibility and optimization learned through GRPO training.

Scalpel is the "lite" variant of the series: conservative, stable and focused on maintaining coherence rather than maximizing aggressiveness.

Architecture

Base architecture: Llama 3.2 1B
Parameters: 1B
Merge method: SLERP
Precision: FP16
Base model influence: ~70%
GRPO model influence: ~30%

Intended Behavior

GRPO RPG System 3.2 1B Scalpel is designed for:

Roleplay and interactive storytelling.
Fiction writing and worldbuilding.
Character simulation.
Fantasy, science fiction and alternate history scenarios.
General conversations with a narrative orientation.

The model prioritizes:

Coherent long-form responses.
Strong character consistency.
Rich descriptions.
Flexible dialogue generation.
Preservation of narrative context.

Strengths

Lightweight and efficient.
Strong roleplay capabilities.
Good balance between creativity and coherence.
Enhanced conversational fluidity inherited from GRPO optimization.
Compatible with most Llama 3.2 1B inference frameworks.

Limitations

Experimental merge.
Performance may vary depending on prompting style.
Long-context coherence remains limited by the underlying 1B architecture.
May exhibit inconsistent behavior outside its intended narrative and conversational domains.

Recommended Settings

Temperature: 0.8 – 1.2
Top-p: 0.9 – 0.98
Min-p: 0.05
Repetition penalty: 1.05 – 1.15
Context: 8K recommended

Version

GRPO RPG System 3.2 1B Scalpel

Lite Variant — Focused on preserving RPG capabilities while integrating a moderate amount of GRPO-derived behavior.

Experimental release.

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:



# Author: Dr. Novaciano
# Objective: GRPO RPG Unethic 3.2 1B AI Model
# =========================================================
# PROJECT: GRPO RPG System 3.2 1B - "Bisturí"
# =========================================================

models:
  - model: NovaCorp/Ultimate-RPG.System-3.2-1B
  - model: jtatman/llama3.2_1b_uncensored_pentest_grpo-merged

merge_method: slerp

base_model: NovaCorp/Ultimate-RPG.System-3.2-1B

parameters:
  t: 0.25

dtype: float16

Downloads last month: 36

Safetensors

Model size

1B params

Tensor type

F16

Model tree for NovaCorp/GRPO-RPG.System-3.2-1B-Scalpel

NovaCorp/Ultimate-RPG.System-3.2-1B

jtatman/llama3.2_1b_uncensored_pentest_grpo-merged

Merge model

this model

Quantizations

2 models