Instructions to use nonetrix/Japanese-Migu-70b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nonetrix/Japanese-Migu-70b with Transformers:

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

pipe = pipeline("text-generation", model="nonetrix/Japanese-Migu-70b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nonetrix/Japanese-Migu-70b")
model = AutoModelForCausalLM.from_pretrained("nonetrix/Japanese-Migu-70b")
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 nonetrix/Japanese-Migu-70b with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/nonetrix/Japanese-Migu-70b

SGLang

How to use nonetrix/Japanese-Migu-70b 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 "nonetrix/Japanese-Migu-70b" \
    --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": "nonetrix/Japanese-Migu-70b",
		"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 "nonetrix/Japanese-Migu-70b" \
        --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": "nonetrix/Japanese-Migu-70b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nonetrix/Japanese-Migu-70b with Docker Model Runner:
```
docker model run hf.co/nonetrix/Japanese-Migu-70b
```

Successfully merging with Miqu

by Ont - opened Mar 8, 2024

Discussion

Ont

Mar 8, 2024

Miqu weights differ more than differences between many other 70B models. As demonstrated by this model, merges between models with greater differences may not make sense unless done carefully. For example, Sophosympatheia has shared a YAML configuration for a Miqu merge that resulted in a working model:
https://huggingface.co/sophosympatheia/Midnight-Miqu-70B-v1.0#configuration

A different Japanese model might possibly merge better (or worse). Here's another Japanese model:
https://huggingface.co/tokyotech-llm/Swallow-70b-NVE-instruct-hf

I hope this information helps if you're interested in trying again.

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