Nurture-intelligence/kEy-llama3.1-8b-v0.1

Overview

The purpose of this model is to create a base model for creating Nurture Intelligence. After merging a model that is good at ASEAN language, a model that is good at Japanese language, and a model with high basic intelligence, the instruct Models trained on the dataset.

We would like to thank all those who created the original model！

How to use

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "Nurture-intelligence/kEy-llama3.1-8b-v0.1"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")

messages = [
    {"role": "user", "content": "自己成長するLLMを作るにあたって大切なことを5つおしえてください。"},
]
input_ids = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)

outputs = model.generate(input_ids,
    do_sample=True,
    temperature=0.2, ## Recommend 0.6 or lower
    max_new_tokens=1024
)
response = outputs[0][input_ids.shape[-1]:]
print(tokenizer.decode(response, skip_special_tokens=True))

Used model

• meditsolutions/Llama-3.1-MedIT-SUN-8B
• elyza/Llama-3-ELYZA-JP-8B
• aisingapore/llama3-8b-cpt-sea-lionv2.1-instruct

Configuration

The following YAML configuration was used to produce this model: ```yaml

models: model: meditsolutions/Llama-3.1-MedIT-SUN-8B


- model: meditsolutions/Llama-3.1-MedIT-SUN-8B
parameters: weight: 1.0
weight: 1.0
- model: elyza/Llama-3-ELYZA-JP-8B
parameters: weight: 0.3
weight: 0.3
- model: aisingapore/llama3-8b-cpt-sea-lionv2.1-instruct
parameters: weight: 0.5
weight: 0.5


merge_method: breadcrumbs
base_model: meditsolutions/Llama-3.1-MedIT-SUN-8B
dtype: bfloat16