Instructions to use vhab10/llama-3-8b-merged-linear with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vhab10/llama-3-8b-merged-linear with Transformers:

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

pipe = pipeline("text-generation", model="vhab10/llama-3-8b-merged-linear")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("vhab10/llama-3-8b-merged-linear")
model = AutoModelForCausalLM.from_pretrained("vhab10/llama-3-8b-merged-linear", 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 vhab10/llama-3-8b-merged-linear with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vhab10/llama-3-8b-merged-linear"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vhab10/llama-3-8b-merged-linear",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/vhab10/llama-3-8b-merged-linear

SGLang

How to use vhab10/llama-3-8b-merged-linear 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 "vhab10/llama-3-8b-merged-linear" \
    --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": "vhab10/llama-3-8b-merged-linear",
		"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 "vhab10/llama-3-8b-merged-linear" \
        --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": "vhab10/llama-3-8b-merged-linear",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use vhab10/llama-3-8b-merged-linear with Docker Model Runner:
```
docker model run hf.co/vhab10/llama-3-8b-merged-linear
```

llama-3-8b-merged-linear

Overview

This model represents a linear merge of three distinct Llama 3-8b models using the Mergekit tool. The primary goal of this merge is to leverage the unique strengths of each base model, such as multilingual capabilities and specialized domain knowledge, into a more versatile and generalized language model.

By merging these models linearly, we combine their expertise into a unified model that performs well across various tasks, such as text generation, multilingual understanding, and domain-specific tasks.

Model Details

Model Description

Models Used:
- Danielbrdz/Barcenas-Llama3-8b-ORPO
- DeepMount00/Llama-3-8b-Ita
- lightblue/suzume-llama-3-8B-multilingual
Merging Tool: Mergekit
Merge Method: Linear merge with equal weighting (1.0) for all models
Tokenizer Source: Union
Data Type: float16 (FP16) precision
License: MIT License
Languages Supported: Multilingual, including English, Italian, and potentially others from the multilingual base models

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: Danielbrdz/Barcenas-Llama3-8b-ORPO
    parameters:
      weight: 1.0
  - model: DeepMount00/Llama-3-8b-Ita
    parameters:
      weight: 1.0
  - model: lightblue/suzume-llama-3-8B-multilingual
    parameters:
      weight: 1.0
merge_method: linear
tokenizer_source: union
dtype: float16

Downloads last month: 18

Safetensors

Model size

8B params

Tensor type

F16

F32

Model tree for vhab10/llama-3-8b-merged-linear

Base model

meta-llama/Meta-Llama-3-8B

Quantized

(280)

this model