Instructions to use opsci/Claire-Modes-7b-0.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use opsci/Claire-Modes-7b-0.1 with Transformers:

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

pipe = pipeline("text-generation", model="opsci/Claire-Modes-7b-0.1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("opsci/Claire-Modes-7b-0.1")
model = AutoModelForCausalLM.from_pretrained("opsci/Claire-Modes-7b-0.1")
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 opsci/Claire-Modes-7b-0.1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "opsci/Claire-Modes-7b-0.1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "opsci/Claire-Modes-7b-0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/opsci/Claire-Modes-7b-0.1

SGLang

How to use opsci/Claire-Modes-7b-0.1 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 "opsci/Claire-Modes-7b-0.1" \
    --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": "opsci/Claire-Modes-7b-0.1",
		"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 "opsci/Claire-Modes-7b-0.1" \
        --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": "opsci/Claire-Modes-7b-0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use opsci/Claire-Modes-7b-0.1 with Docker Model Runner:
```
docker model run hf.co/opsci/Claire-Modes-7b-0.1
```

Claire-Modes-7b-0.1

Claire-Modes-7b-0.1 is a version of Mistral-Hermes finetuned by OpSci on the French conversational dataset curated by LINAGORA and OpenLLM-France. Claire is structured around 6 conversational "modes" that illustrate the potential of finetuning to create different cultural environement of communication.

The same dataset was used to train Claire-7B-0.1.

Claire-Modes-7b-0.1 uses a different approach that illustrate the potential of the French conversational dataset. It is a light finetune of one of the best available chat model. It is structure by default for chat conversations in chatML.

A temporary demo of the Theater mode is available on this HuggingFace space: Claire-Théâtre

Conversational modes

The main sources of conversational data have been integrated with special prompts of modes. Currently, Claire-Modes supports six modes:

Theater ("Tu es Claire-Théâtre le chatbot théâtral. Répond à la question posée avec un style littéraire et une emphase comique ou dramatique.")
Politics ("Tu es Claire-politique le chatbot de la vie politique française. Répond en utilisant le language élégant et les références d’une personnalité politique française.")
Meetings ("Tu es Claire-meeting le chatbot des rendez-vous d’affaire. Répond en utilisant un langage d’entreprise simple et spontané.")
Interviews ("Tu es Claire-Interview le chatbot des entretiens. Répond en utilisant un français oral, simple, accessible, mais soigné.")
Dialogues divers ("Tu es Claire-dialogue, le chatbot audio. Répond en utilisant un français oral spontané.")
Situation d’assistance ("Tu es Claire-Situation le chatbot des entreprise. Répond à la situation d’assistance soumise en étant aussi aidant que possible.")

The style of Claire-modes and its persona should change depending on the chosen mode.

License

Given that some of the corpora used for training are only available under CC-BY-NC-SA licenses, Claire-7B-0.1 is made available under the CC-BY-NC-SA 4.0 license.

Downloads last month: 10

Model tree for opsci/Claire-Modes-7b-0.1

Base model

mistralai/Mistral-7B-v0.1

Finetuned

teknium/OpenHermes-2-Mistral-7B

Finetuned

(5)

this model