datasets:
- wikipedia
- Ichsan2895/OASST_Top1_Indonesian
- Ichsan2895/alpaca-gpt4-indonesian
language:
- id
- en
pipeline_tag: text-generation
license: cc-by-nc-sa-4.0
HAPPY TO ANNOUNCE THE RELEASE OF MERAK-7B-V4!
Merak-7B is the Large Language Model of Indonesian Language
This model is based on Mistral-7B-OpenOrca and fine tuned by some of Indonesia Wikipedia articles that I cleaned before.
Leveraging QLoRA (QLora: Efficient Finetuning of Quantized LLMs), Merak-7B is able to run with 16 GB VRAM
Licensed under Creative Commons-By Attribution-Share Alike-Non Commercial (CC-BY-SA-NC 4.0) Merak-7B empowers AI enthusiasts, researchers alike.
Big thanks to all my friends and communities that help to build our first model. Thanks for Axolotl for a great fine tuning tool which designed to streamline the fine-tuning of various AI models.
Feel free, to ask me about the model and please share the news on your social media.
HOW TO USE
Installation
Please make sure you have installed CUDA driver in your system, Python 3.10 and PyTorch 2. Then install this library in terminal
pip install protobuf==4.24.4
pip install bitsandbytes==0.41.1
pip install transformers==4.34.1
pip install peft==0.5.0
pip install accelerate==0.23.0
pip install einops==0.6.1 scipy sentencepiece datasets
Using BitsandBytes and it run with >= 10 GB VRAM GPU
import torch
from transformers import AutoTokenizer, AutoConfig, AutoModelForCausalLM, BitsAndBytesConfig, LlamaTokenizer
from peft import PeftModel, PeftConfig
model_id = "Ichsan2895/Merak-7B-v4"
config = AutoConfig.from_pretrained(model_id)
BNB_CONFIG = BitsAndBytesConfig(load_in_4bit=True,
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_use_double_quant=True,
bnb_4bit_quant_type="nf4",
)
model = AutoModelForCausalLM.from_pretrained(model_id,
quantization_config=BNB_CONFIG,
device_map="auto",
trust_remote_code=True)
tokenizer = LlamaTokenizer.from_pretrained(model_id)
def generate_response(question: str) -> str:
chat = [
{"role": "system", "content": "Anda adalah Merak, sebuah model kecerdasan buatan yang dilatih oleh Muhammad Ichsan. Mohon jawab pertanyaan berikut dengan benar, faktual, dan ramah."},
{"role": "user", "content": question},
]
prompt = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt", return_attention_mask=True)
with torch.no_grad():
outputs = model.generate(input_ids=inputs["input_ids"].to("cuda"),
attention_mask=inputs.attention_mask,
eos_token_id=tokenizer.eos_token_id,
pad_token_id=tokenizer.eos_token_id,
max_new_tokens=256)
response = tokenizer.batch_decode(outputs.detach().cpu().numpy(), skip_special_tokens=True)[0]
assistant_start = f'''{question} \n assistant\n '''
response_start = response.find(assistant_start)
return response[response_start + len(assistant_start) :].strip()
prompt = "Siapa penulis naskah proklamasi kemerdekaan Indonesia?"
print(generate_response(prompt))
From my experience, For better answer, please don’t use BitsandBytes 4-bit Quantization, but it using higher VRAM
import torch
from transformers import AutoTokenizer, AutoConfig, AutoModelForCausalLM, BitsAndBytesConfig, LlamaTokenizer
from peft import PeftModel, PeftConfig
model_id = "Ichsan2895/Merak-7B-v4"
config = AutoConfig.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id,
device_map="auto",
trust_remote_code=True)
tokenizer = LlamaTokenizer.from_pretrained(model_id)
def generate_response(question: str) -> str:
chat = [
{"role": "system", "content": "Anda adalah Merak, sebuah model kecerdasan buatan yang dilatih oleh Muhammad Ichsan. Mohon jawab pertanyaan berikut dengan benar, faktual, dan ramah."},
{"role": "user", "content": question},
]
prompt = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt", return_attention_mask=True)
with torch.no_grad():
outputs = model.generate(input_ids=inputs["input_ids"].to("cuda"),
attention_mask=inputs.attention_mask,
eos_token_id=tokenizer.eos_token_id,
pad_token_id=tokenizer.eos_token_id,
max_new_tokens=256)
response = tokenizer.batch_decode(outputs.detach().cpu().numpy(), skip_special_tokens=True)[0]
assistant_start = f'''{question} \n assistant\n '''
response_start = response.find(assistant_start)
return response[response_start + len(assistant_start) :].strip()
prompt = "Siapa penulis naskah proklamasi kemerdekaan Indonesia?"
print(generate_response(prompt))
CHANGELOG
v4 = We use Mistral-7B-OpenOrca instead of Llama-2-Chat-HF. We did it throught uncounted trial-and-error. We pick the best one to do this model.
What we have done so far:
1st). We fine tuned it with Wikipedia articles that we cleaned it before. It use QLora and speed up by Deepspeed Zero 2 for 1 epoch. Axolotl was used for easier fine tuning configuration.
2nd). We got extra funds. Thanks all.. We did it again like first step but it was Full Parameter fine tuning (FFT) instead of QLora.
3rd). We fine tuned it with Ichsan2895/OASST_Top1_Indonesian & Ichsan2895/alpaca-gpt4-indonesian with minor modification, so it was suitable with ChatML format. It was FFT for 4 epochs.
v3 = Fine tuned by Ichsan2895/OASST_Top1_Indonesian & Ichsan2895/alpaca-gpt4-indonesian
v2 = Finetuned version of first Merak-7B model. We finetuned again with the same ID Wikipedia articles except it changes prompt-style in the questions. It has 600k ID wikipedia articles.
v1 = The first Merak-7B model. We selected and cleaned about 200k ID wikipedia articles.
CITATION
@software{lian2023mistralorca1
title = {MistralOrca: Mistral-7B Model Instruct-tuned on Filtered OpenOrcaV1 GPT-4 Dataset},
author = {Wing Lian and Bleys Goodson and Guan Wang and Eugene Pentland and Austin Cook and Chanvichet Vong and "Teknium"},
year = {2023},
publisher = {HuggingFace},
journal = {HuggingFace repository},
howpublished = {\url{https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca},
}
@misc{mukherjee2023orca,
title={Orca: Progressive Learning from Complex Explanation Traces of GPT-4},
author={Subhabrata Mukherjee and Arindam Mitra and Ganesh Jawahar and Sahaj Agarwal and Hamid Palangi and Ahmed Awadallah},
year={2023},
eprint={2306.02707},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
@ONLINE{wikidump,
author = "Wikimedia Foundation",
title = "Wikimedia Downloads",
url = "https://dumps.wikimedia.org"
}
@inproceedings{wolf-etal-2020-transformers,
title = "Transformers: State-of-the-Art Natural Language Processing",
author = "Thomas Wolf and Lysandre Debut and Victor Sanh and Julien Chaumond and Clement Delangue and Anthony Moi and Pierric Cistac and Tim Rault and Rémi Louf and Morgan Funtowicz and Joe Davison and Sam Shleifer and Patrick von Platen and Clara Ma and Yacine Jernite and Julien Plu and Canwen Xu and Teven Le Scao and Sylvain Gugger and Mariama Drame and Quentin Lhoest and Alexander M. Rush",
booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
month = oct,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/2020.emnlp-demos.6",
pages = "38--45"
}
@article{dettmers2023qlora,
title = {QLoRA: Efficient Finetuning of Quantized LLMs},
author = {Dettmers, Tim and Pagnoni, Artidoro and Holtzman, Ari and Zettlemoyer, Luke},
journal = {arXiv preprint arXiv:2305.14314},
year = {2023}
}
HOW TO CITE THIS PROJECT
If you use the Merak-7B model in your research or project, please cite it as:
@article{Merak,
title={Merak-7B: The LLM for Bahasa Indonesia},
author={Muhammad Ichsan},
publisher={Hugging Face}
journal={Hugging Face Repository},
year={2023}
}