Instructions to use FluffyKaeloky/Midnight-Miqu-103B-v1.5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use FluffyKaeloky/Midnight-Miqu-103B-v1.5 with Transformers:

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

pipe = pipeline("text-generation", model="FluffyKaeloky/Midnight-Miqu-103B-v1.5")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("FluffyKaeloky/Midnight-Miqu-103B-v1.5")
model = AutoModelForCausalLM.from_pretrained("FluffyKaeloky/Midnight-Miqu-103B-v1.5", device_map="auto")
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 Settings

vLLM

How to use FluffyKaeloky/Midnight-Miqu-103B-v1.5 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "FluffyKaeloky/Midnight-Miqu-103B-v1.5"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FluffyKaeloky/Midnight-Miqu-103B-v1.5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/FluffyKaeloky/Midnight-Miqu-103B-v1.5

SGLang

How to use FluffyKaeloky/Midnight-Miqu-103B-v1.5 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 "FluffyKaeloky/Midnight-Miqu-103B-v1.5" \
    --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": "FluffyKaeloky/Midnight-Miqu-103B-v1.5",
		"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 "FluffyKaeloky/Midnight-Miqu-103B-v1.5" \
        --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": "FluffyKaeloky/Midnight-Miqu-103B-v1.5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use FluffyKaeloky/Midnight-Miqu-103B-v1.5 with Docker Model Runner:
```
docker model run hf.co/FluffyKaeloky/Midnight-Miqu-103B-v1.5
```

exl2 quants

by Wsdfsdf - opened Mar 21, 2024

Discussion

Wsdfsdf

Mar 21, 2024

@LoneStriker 5.0BPW please

FluffyKaeloky

Owner Mar 21, 2024

I have uploaded the 5.0bpw quant, sorry for the delay.

Wsdfsdf

Mar 21, 2024

I have uploaded the 5.0bpw quant, sorry for the delay.

Thanks, but I compared Goliath-120B-rpcal months ago to the default calibration dataset and the rpcal will actually produce noticeably worse output even for RP and higher perplexity compared to the default dataset. Turboderp even uploaded Goliath-120B quants because of this to test. I'll just wait for LoneStriker I guess, as I don't want you to waste your time requantizing this large model while LoneStriker has probably already downloaded and is quantizing it as I type this. Just a warning in the future to stick with the default dataset, because I wouldn't be surprised if it's worse then 70B-v1.0 because of that.

FluffyKaeloky

Owner Mar 21, 2024

•

edited Mar 21, 2024

Alright, I didn't know that. I'll take it into account for future quants and try to requant if someone doesn't do it before me.
Edit : He just did ^^

Wsdfsdf changed discussion status to closed Mar 22, 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