Instructions to use Nabbers1999/Melpomene-70B-0307-Uncensored with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Nabbers1999/Melpomene-70B-0307-Uncensored with Transformers:

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

pipe = pipeline("text-generation", model="Nabbers1999/Melpomene-70B-0307-Uncensored")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Nabbers1999/Melpomene-70B-0307-Uncensored")
model = AutoModelForCausalLM.from_pretrained("Nabbers1999/Melpomene-70B-0307-Uncensored", 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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Nabbers1999/Melpomene-70B-0307-Uncensored with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Nabbers1999/Melpomene-70B-0307-Uncensored"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Nabbers1999/Melpomene-70B-0307-Uncensored",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Nabbers1999/Melpomene-70B-0307-Uncensored

SGLang

How to use Nabbers1999/Melpomene-70B-0307-Uncensored 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 "Nabbers1999/Melpomene-70B-0307-Uncensored" \
    --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": "Nabbers1999/Melpomene-70B-0307-Uncensored",
		"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 "Nabbers1999/Melpomene-70B-0307-Uncensored" \
        --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": "Nabbers1999/Melpomene-70B-0307-Uncensored",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Nabbers1999/Melpomene-70B-0307-Uncensored with Docker Model Runner:
```
docker model run hf.co/Nabbers1999/Melpomene-70B-0307-Uncensored
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Melpomene 70B - Uncensored

Thalia is a chat model I have merged to provide a distillation model for future projects. Combining a Lumimaid model with Strawberry Lemonade over a heavy base of Deepseek R1 Distill Llama 70B has produced a thinking model and healed its safety alignments. This model contains the writing creativity of its two chat model parents, while adding the deep reasoning of Deepseek.

Melpomene is the same model as Thalia, abliterated via orthogonalization and reinforced via direction-only DoRA training. Use at your own risk.

In order for this model to function properly, you should prefill the opening `<think>` tag. This model's ancestry results in a hybrid thinker that sometimes chooses to think without `<think>` tags.

Merge Details

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

Merge Method

This model was merged using the DARE TIES merge method using unsloth/Llama-3.3-70B-Instruct as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: deepseek-ai/DeepSeek-R1-Distill-Llama-70B
    parameters:
      density: 0.8
      weight: 1.0
  - model: sophosympatheia/Strawberrylemonade-L3-70B-v1.1
    parameters:
      density: 0.5
      weight: 0.4
  - model: NeverSleep/Lumimaid-v0.2-70B
    parameters:
      density: 0.5
      weight: 0.4
merge_method: dare_ties
base_model: unsloth/Llama-3.3-70B-Instruct
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16