Instructions to use prhegde/multi-merged-llama2-7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use prhegde/multi-merged-llama2-7b with Transformers:

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

pipe = pipeline("text-generation", model="prhegde/multi-merged-llama2-7b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("prhegde/multi-merged-llama2-7b")
model = AutoModelForCausalLM.from_pretrained("prhegde/multi-merged-llama2-7b")
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 prhegde/multi-merged-llama2-7b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "prhegde/multi-merged-llama2-7b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "prhegde/multi-merged-llama2-7b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/prhegde/multi-merged-llama2-7b

SGLang

How to use prhegde/multi-merged-llama2-7b 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 "prhegde/multi-merged-llama2-7b" \
    --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": "prhegde/multi-merged-llama2-7b",
		"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 "prhegde/multi-merged-llama2-7b" \
        --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": "prhegde/multi-merged-llama2-7b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use prhegde/multi-merged-llama2-7b with Docker Model Runner:
```
docker model run hf.co/prhegde/multi-merged-llama2-7b
```

Summary

This model is a merge of following 4 models

HWERI/Llama2-7b-sharegpt4
circulus/Llama-2-7b-orca-v1
starmpcc/Asclepius-Llama2-7B
meta-llama/Llama-2-7b-chat-hf

Merge is done through averaging of the parameters.

merged_wt.copy_( (base_wt + asclepius_wt + sharegpt_wt + orca_wt) / 4.0 )

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

MODEL_ID = "prhegde/multi-merged-llama2-7b"
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(MODEL_ID)

Prompt = "You are given a passage PASSAGE and a boolean question QUESTION. Return answer as True or False. \n PASSAGE: Calcium carbide is a chemical compound with the chemical formula of CaC. Its main use industrially is in the production of acetylene and calcium cyanamide. \n QUESTION: is calcium carbide cac2 the raw material for the production of acetylene? \n ANSWER: "

input_ids = tokenizer(Prompt, return_tensors="pt").input_ids
output = model.generate(input_ids)
output_seq = tokenizer.decode(output[0], skip_special_tokens=True)
print(output_seq)

Downloads last month: 3

Safetensors

Model size

7B params

Tensor type

F32