Instructions to use CohereLabs/c4ai-command-r-08-2024 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CohereLabs/c4ai-command-r-08-2024 with Transformers:

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

pipe = pipeline("text-generation", model="CohereLabs/c4ai-command-r-08-2024")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("CohereLabs/c4ai-command-r-08-2024")
model = AutoModelForCausalLM.from_pretrained("CohereLabs/c4ai-command-r-08-2024")
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
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use CohereLabs/c4ai-command-r-08-2024 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CohereLabs/c4ai-command-r-08-2024"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CohereLabs/c4ai-command-r-08-2024",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/CohereLabs/c4ai-command-r-08-2024

SGLang

How to use CohereLabs/c4ai-command-r-08-2024 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 "CohereLabs/c4ai-command-r-08-2024" \
    --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": "CohereLabs/c4ai-command-r-08-2024",
		"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 "CohereLabs/c4ai-command-r-08-2024" \
        --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": "CohereLabs/c4ai-command-r-08-2024",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use CohereLabs/c4ai-command-r-08-2024 with Docker Model Runner:
```
docker model run hf.co/CohereLabs/c4ai-command-r-08-2024
```

The `safety_mode` parameter?

by jukofyork - opened Aug 30, 2024

Discussion

jukofyork

Aug 30, 2024

•

edited Aug 30, 2024

https://docs.cohere.com/docs/safety-modes

Is this just something sent to the official Cohere API, or is it something we can control using the Transformers version in this repo via the preemble in the system message?

jukofyork

Aug 30, 2024

Also the "Chat Annotation Style Guide: Model Output" link is gated on that page.

algorithm

Aug 30, 2024

•

edited Aug 30, 2024

Good question, wondering the same, I'm trying to disable it.

I found some clues in chat_template within tokenizer_config.json, it starts with:

{{ '# Safety Preamble' }}{{ '\nThe instructions in this section override those in the task description and style guide sections............

That lead me to believe that it might be part of the system prompt, though I haven't looked into it deep enough.

Curious to know what the default is considering it's only defined at "name": "tool_use", and "name": "rag", but not at "name": "default",.

alexrs

Cohere Labs org Sep 9, 2024

Hi @jukofyork , the safety mode is only available through the API for now. Hope that helps!

alexrs changed discussion status to closed Sep 9, 2024

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment