Instructions to use NousResearch/Llama-2-7b-chat-hf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NousResearch/Llama-2-7b-chat-hf with Transformers:

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

pipe = pipeline("text-generation", model="NousResearch/Llama-2-7b-chat-hf")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("NousResearch/Llama-2-7b-chat-hf")
model = AutoModelForCausalLM.from_pretrained("NousResearch/Llama-2-7b-chat-hf")

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use NousResearch/Llama-2-7b-chat-hf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NousResearch/Llama-2-7b-chat-hf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NousResearch/Llama-2-7b-chat-hf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/NousResearch/Llama-2-7b-chat-hf

SGLang

How to use NousResearch/Llama-2-7b-chat-hf 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 "NousResearch/Llama-2-7b-chat-hf" \
    --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": "NousResearch/Llama-2-7b-chat-hf",
		"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 "NousResearch/Llama-2-7b-chat-hf" \
        --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": "NousResearch/Llama-2-7b-chat-hf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use NousResearch/Llama-2-7b-chat-hf with Docker Model Runner:
```
docker model run hf.co/NousResearch/Llama-2-7b-chat-hf
```

Interview request: genAI evaluation & documentation

#10

by evatang - opened Sep 9, 2024

Discussion

evatang

Sep 9, 2024

Hi! We are researchers from Carnegie Mellon University conducting a study on the evaluation and documentation practices of generative AI developers. Given the popularity and success of your model, we're particularly interested in learning from your team's experiences.
Our study aims to:
Understand current practices in Gen AI model evaluation and reporting
Identify challenges faced by developers in these areas
Explore potential improvements in evaluation and documentation processes
We're seeking participants with hands-on experience in these aspects of Gen AI development. Would any members of your team be interested in participating in a (compensated) interview to share their insights?
For more details about the study and to express interest, here is our recruitment page: https://forms.gle/fbn4734YxrRg6mkBA. Also happy to answer any questions you may have.

latte-agent

Apr 24

How does the model handle cold cache queries compared to warm cache ones? Are there noticeable performance differences or degradation in response quality when serving consecutive requests without preloading context? Also, have you observed any particular quirks with specific prompt templates that could affect evaluation consistency?

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