Instructions to use Rakuten/RakutenAI-7B-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Rakuten/RakutenAI-7B-instruct with Transformers:

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

pipe = pipeline("text-generation", model="Rakuten/RakutenAI-7B-instruct")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Rakuten/RakutenAI-7B-instruct")
model = AutoModelForCausalLM.from_pretrained("Rakuten/RakutenAI-7B-instruct")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Rakuten/RakutenAI-7B-instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Rakuten/RakutenAI-7B-instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Rakuten/RakutenAI-7B-instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Rakuten/RakutenAI-7B-instruct

SGLang

How to use Rakuten/RakutenAI-7B-instruct 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 "Rakuten/RakutenAI-7B-instruct" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Rakuten/RakutenAI-7B-instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "Rakuten/RakutenAI-7B-instruct" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Rakuten/RakutenAI-7B-instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Rakuten/RakutenAI-7B-instruct with Docker Model Runner:
```
docker model run hf.co/Rakuten/RakutenAI-7B-instruct
```

Very limited model response... am I doing something wrong?

by echa00 - opened Aug 3, 2024

Discussion

echa00

Aug 3, 2024

Tried the following prompt and the response was worse than 'Mistral-7B-Instruct-v0.3' without the Japanese fine-tuning. Am I doing something wrong?

import requests

API_URL = "XXX"
headers = {
    "Accept" : "application/json",
    "Authorization": "Bearer YYY",
    "Content-Type": "application/json" 
}

def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

system_prompt = "You are a Japanese language teacher. Your task is to draft a beginner-level Japanese roleplay for making a restaurant reservation, focusing on specifying any dietary restrictions or special requests. Output the dialogue in JSON format with only the following keys: 'Japanese' and 'speaker'. "
prompt = "Draft a beginner-level Japanese roleplay to help me practice making a restaurant reservation, focusing on specifying any dietary restrictions or special requests. Use real-life examples."

msg = "<s> [INST] "+system_prompt+" \n "+prompt+" [/INST]"

output = query({
    "inputs": msg,
    "parameters": {
        "top_k": 50,
        "top_p": 0.95,
        "temperature": 0.7,
        "max_new_tokens": 1024
    }
})

print(output)

HenriqueSch

Nov 1, 2024

The issue might be your prompt formatting. From what I've worked with rakuten, your prompt should be "General Instructions \n USER:\n user provided instructions\n ASSISTANT:\n". The model likes to follow this format. In my experience, passing any general instructions didn't really help, I've contained my prompt only to the user section and that works best for me. In general it is a bit finnicky with instructions and sometimes it will just keep babbling random stuff if it doesn't like your prompt.

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