Instructions to use ilsp/Meltemi-7B-Instruct-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ilsp/Meltemi-7B-Instruct-v1 with Transformers:

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

pipe = pipeline("text-generation", model="ilsp/Meltemi-7B-Instruct-v1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ilsp/Meltemi-7B-Instruct-v1")
model = AutoModelForCausalLM.from_pretrained("ilsp/Meltemi-7B-Instruct-v1")
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 ilsp/Meltemi-7B-Instruct-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ilsp/Meltemi-7B-Instruct-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ilsp/Meltemi-7B-Instruct-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ilsp/Meltemi-7B-Instruct-v1

SGLang

How to use ilsp/Meltemi-7B-Instruct-v1 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 "ilsp/Meltemi-7B-Instruct-v1" \
    --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": "ilsp/Meltemi-7B-Instruct-v1",
		"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 "ilsp/Meltemi-7B-Instruct-v1" \
        --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": "ilsp/Meltemi-7B-Instruct-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ilsp/Meltemi-7B-Instruct-v1 with Docker Model Runner:
```
docker model run hf.co/ilsp/Meltemi-7B-Instruct-v1
```

Minimum Requirements for running the model in CUDA

by Rafimc13 - opened Apr 15, 2024

Discussion

Rafimc13

Apr 15, 2024

•

edited Apr 15, 2024

Hello Guys,
Congratulations for the great work.
Could you please share the minimum requirements that Meltemi needs in order to run in GPU with cuda?
For example a dedicated memory of 4gb (NVIDIA) will be able to run the model efficiently?
Moreover, with a simple usage with CPU will the model run in the inference mode or is not possible?
Thank you in advance.

Rafail

LVouk

Athena Research Center | Institute for Language and Speech Processing org Apr 15, 2024

Hello,
In general if you want to run inference at half precision you require ~16GB VRAM
To run inference at lower VRAM values you'll either have to quantize the model yourself or try either of our AWQ or GGUF versions. Your best bet for CPU usage would be GGUF

Rafimc13

Apr 15, 2024

Thank you for the quick response. I will try the GGUF version.

Geo

Apr 16, 2024

•

edited Apr 16, 2024

I was waiting for a greek LLM for a very long time. Congratulations. Sorry for my ignorance, but can this model be used for Retrieval Augmented Generation?

LVouk

Athena Research Center | Institute for Language and Speech Processing org Apr 17, 2024

•

edited Apr 17, 2024

As any other model, it can be used as a "generator" in the RAG process.
Having said that for the storing and retrieval purposes you would have to pick suitable embeddings for Greek which might not be the case for all embedding algos out there, so it might require some trial and error.

LVouk changed discussion status to closed Apr 17, 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