Instructions to use vicgalle/Roleplay-Llama-3-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vicgalle/Roleplay-Llama-3-8B with Transformers:

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

pipe = pipeline("text-generation", model="vicgalle/Roleplay-Llama-3-8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("vicgalle/Roleplay-Llama-3-8B")
model = AutoModelForMultimodalLM.from_pretrained("vicgalle/Roleplay-Llama-3-8B")
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 vicgalle/Roleplay-Llama-3-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vicgalle/Roleplay-Llama-3-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vicgalle/Roleplay-Llama-3-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/vicgalle/Roleplay-Llama-3-8B

SGLang

How to use vicgalle/Roleplay-Llama-3-8B 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 "vicgalle/Roleplay-Llama-3-8B" \
    --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": "vicgalle/Roleplay-Llama-3-8B",
		"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 "vicgalle/Roleplay-Llama-3-8B" \
        --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": "vicgalle/Roleplay-Llama-3-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use vicgalle/Roleplay-Llama-3-8B with Docker Model Runner:
```
docker model run hf.co/vicgalle/Roleplay-Llama-3-8B
```

<|eot_id|><|start_header_id|>assistant<|end_header_id|> in model outputs

by willsims - opened May 22, 2024

Discussion

willsims

May 22, 2024

This model has a ton of potential and thanks for making it available to everyone. I'm hosting your model via inference endpoints on hugging face and we have an issue where <|eot_id|><|start_header_id|>assistant<|end_header_id|> is included in model outputs. Is this expected or do you know how we can fix it? Additionally, when we limit the token length, responses get cut off. Is the best way to target a certain message length through prompt engineering?

vicgalle

Owner May 28, 2024

Hi @willsims !

Is this expected or do you know how we can fix it?

Yes, it is expected, because that's the chat template both Meta and me used. If you use huggingface you can use this https://huggingface.co/docs/transformers/main/en/chat_templating to control it, but you can just filter out all the <| ... |>

Is the best way to target a certain message length through prompt engineering?
I think so. If you are using a low max_tokens in the response, you can try specifying in the system prompt that it answers with short and concise responses, or something like it. Hope it helps!

willsims

May 28, 2024

This is very helpful, thank you @vicgalle !

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