Instructions to use AlexCuadron/dpo_task_vio with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AlexCuadron/dpo_task_vio with Transformers:

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

pipe = pipeline("text-generation", model="AlexCuadron/dpo_task_vio")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AlexCuadron/dpo_task_vio")
model = AutoModelForCausalLM.from_pretrained("AlexCuadron/dpo_task_vio")
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

vLLM

How to use AlexCuadron/dpo_task_vio with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/AlexCuadron/dpo_task_vio

SGLang

How to use AlexCuadron/dpo_task_vio 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 "AlexCuadron/dpo_task_vio" \
    --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": "AlexCuadron/dpo_task_vio",
		"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 "AlexCuadron/dpo_task_vio" \
        --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": "AlexCuadron/dpo_task_vio",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AlexCuadron/dpo_task_vio with Docker Model Runner:
```
docker model run hf.co/AlexCuadron/dpo_task_vio
```

output

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the Task Arithmetic merge method using Nitral-AI/Captain-Eris_Violet-V0.420-12B as a base.

Models Merged

The following models were included in the merge:

AlexCuadron/chai-ddpo

Configuration

The following YAML configuration was used to produce this model:

# MergeKit Configuration for merging alignment-DPO with roleplay capabilities
merge_method: task_arithmetic

# Define the base model (your alignment model)
base_model:
  model: Nitral-AI/Captain-Eris_Violet-V0.420-12B # Nitral-AI/Community_Request-02-12B  # Path to your DPO-trained alignment model
  # If your model is in Hugging Face format, you could use:
  model_type: huggingface
  # If it's a local model, just use the path:
  # model: /path/to/alex_dpo

# Models to merge with the base model
models:
  - model: AlexCuadron/chai-ddpo
    # Modify if needed:
    model_type: huggingface
    parameters:
      weight: 0.2  # Adjust between 0.2-0.4 based on how much roleplay vs alignment you want

# Output settings
dtype: bfloat16  # Or bfloat16 if your GPU supports it
output_dir: ./aligned_roleplay_model
# Optional: include parameters you want to skip or include specifically
# parameters_settings:
#   include: [".*"] # Include all parameters, or specify patterns to include
#   exclude: [] # No exclusions, or specify patterns to exclude

# Add this if you want to save disk space
output_optimizations:
  safetensors: true  # Save in safetensors format (more secure and efficient)
  # set to false if you want to keep intermediate models
  skip_saving_intermediate_models: true

Downloads last month: 4

Safetensors

Model size

12B params

Tensor type

BF16

Model tree for AlexCuadron/dpo_task_vio

AlexCuadron/chai-ddpo

Nitral-AI/Captain-Eris_Violet-V0.420-12B

Merge model

this model

Paper for AlexCuadron/dpo_task_vio

Editing Models with Task Arithmetic

Paper • 2212.04089 • Published Dec 8, 2022 • 8