Instructions to use yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated with Transformers:

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

pipe = pipeline("text-generation", model="yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated")
model = AutoModelForCausalLM.from_pretrained("yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated

SGLang

How to use yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated 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 "yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated with Docker Model Runner:
```
docker model run hf.co/yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Merged Model

This model is a combination of:

Base Model: shisa-ai/shisa-v2-mistral-nemo-12b
LoRA Adapter: nbeerbower/Mistral-Nemo-12B-abliterated-LORA

The model is saved in bfloat16 format and is ready for deployment or fine-tuning.

Code for LoRA merging (Generated by Qwen3)

import argparse
import os
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

def main():
    parser = argparse.ArgumentParser(description="Merge a LoRA into a large language model.")
    parser.add_argument("--model_name", type=str, required=True, help="Name or path of the base model.")
    parser.add_argument("--lora_name", type=str, required=True, help="Name or path of the LoRA adapter.")
    parser.add_argument("--output_model_name", type=str, required=True, help="Output directory for the merged model.")
    args = parser.parse_args()

    # Load the base model in bfloat16
    print("Loading base model...")
    base_model = AutoModelForCausalLM.from_pretrained(
        args.model_name,
        torch_dtype="bfloat16"
    )

    # Load the LoRA adapter
    print("Loading LoRA adapter...")
    peft_model = PeftModel.from_pretrained(base_model, args.lora_name)

    # Merge and unload the LoRA weights into the base model
    print("Merging LoRA into base model...")
    merged_model = peft_model.merge_and_unload()

    # Save the merged model
    print(f"Saving merged model to {args.output_model_name}...")
    merged_model.save_pretrained(args.output_model_name)

    # Save the tokenizer from the base model
    tokenizer = AutoTokenizer.from_pretrained(args.model_name)
    tokenizer.save_pretrained(args.output_model_name)

    # Create README.md file
    print("Creating README.md...")
    readme_content = f"""# Merged Model

This model is a combination of:

- **Base Model**: `{args.model_name}`
- **LoRA Adapter**: `{args.lora_name}`

The model is saved in `bfloat16` format and is ready for deployment or fine-tuning.
"""

    readme_path = os.path.join(args.output_model_name, "README.md")
    with open(readme_path, "w") as f:
        f.write(readme_content)

    print("✅ Merging complete. Model saved to:", args.output_model_name)

if __name__ == "__main__":
    main()

Downloads last month: 137

Safetensors

Model size

12B params

Tensor type

BF16

Model tree for yamatazen/Shisa-v2-Mistral-Nemo-12B-Lorablated

nbeerbower/Mistral-Nemo-12B-abliterated-LORA

shisa-ai/shisa-v2-mistral-nemo-12b

Merge model

this model

Quantizations

2 models