Auto Data
1.0.0
Data Otomatis adalah pustaka yang dirancang untuk menghasilkan kumpulan data realistis secara otomatis, yang penting untuk menyempurnakan Model Bahasa Besar (LLM). Pustaka yang sangat efisien dan ringan ini memungkinkan pembuatan kumpulan data komprehensif dengan cepat dan mudah di berbagai topik, berapa pun ukurannya.
Salah satu tantangan utama yang dihadapi selama penyempurnaan model pengembangan agen kustom adalah kelangkaan dan ketidakseimbangan data. Kekurangan seperti ini dapat mengganggu pemahaman model terhadap satu fitur tertentu atau, dalam kasus yang lebih parah, dapat menyebabkan model menyimpang sepenuhnya dari parameter yang dipelajari karena jumlah contoh pelatihan yang tidak mencukupi. Untuk mengatasi masalah kritis ini, Data Otomatis dikembangkan.
Sebelum melanjutkan, siapkan OPENAI_API_KEY Anda sebagai variabel lingkungan. Jika Anda tidak mengetahui cara melakukannya, lihat panduan ini - https://help.openai.com/en/articles/5112595-best-practices-for-api-key-safety
git clone https://github.com/Itachi-Uchiha581/Auto-Data.git cd Auto-Datapip install -r requirements.txtpython main.py --helpOutput dari perintah di atas:
usage: Auto Data [-h] [--model MODEL] [--topic TOPIC] [--format {json,parquet}] [--engine {native}] [--threads THREADS] [--length LENGTH] [--system_prompt SYSTEM_PROMPT]
Auto Data is a tool which automatically creates training data to fine tune Large-Language Models on!
options:
-h, --help show this help message and exit
--model MODEL, -m MODEL
Selection of an OpenAI model for data generation
--topic TOPIC, -t TOPIC
Topic for data generation, eg - Global Economy
--format {json,parquet}, -f {json,parquet}
The format of the output data produced by the LLM
--engine {native}, -e {native}
The backend used to generate data. More engines coming soon
--threads THREADS, -th THREADS
An integer to indicate how many chats to be created on the topic. A very high thread value may result in an error specially if your Open AI account is at tier 1.
--length LENGTH, -l LENGTH
The conversation length of a chat topic
--system_prompt SYSTEM_PROMPT, -sp SYSTEM_PROMPT
The system prompt that is to be given to the assistant.
Contoh Penggunaan diberikan di bawah ini
python main.py --model " gpt-4-turbo-preview " --topic " Mysteries and Horror stories " --format " json " --engine " native " --threads 2 --length 2 --system_prompt " You are a helpful assistant who has an interest in Mysteries and Horror stories. You are also excellent at articulating such stories "Pada perintah yang ditentukan, model yang dipilih adalah gpt-4-turbo-preview , dan topik yang dituju adalah Misteri dan Cerita Horor . Format untuk data keluaran diatur ke json , dengan mesin yang digunakan adalah native . Perintah ini menunjukkan pembuatan 2 utas obrolan, masing-masing mencakup 2 percakapan bolak-balik yang lengkap, sebagaimana ditentukan oleh parameter panjang . Parameter terakhir yang disediakan berfungsi sebagai prompt sistem untuk asisten LLM .
Di bawah ini adalah hasil potongan yang dihasilkan dari perintah di atas dalam format json :
{
"Chats" : [
{
"messages" : [
{
"role" : " system " ,
"content" : " You are a helpful assistant who has an interest in Mysteries and Horror stories. You are also excellent at articulating such stories "
},
{
"role" : " user " ,
"content" : " What's considered one of the most chilling unsolved mysteries in history? "
},
{
"role" : " assistant " ,
"content" : " Among the pantheon of unsolved mysteries, the chilling tale of the Dyatlov Pass incident stands out, both for its perplexing evidence and the many questions that still haunt its narrative. This enigma unfolded in the bleak expanse of the Ural Mountains in Russia during the icy clutches of winter in 1959. ............. "
},
{
"role" : " user " ,
"content" : " Are there any other unsolved mysteries that are as eerie or confounding as the Dyatlov Pass incident? "
},
{
"role" : " assistant " ,
"content": "Indeed, the world is riddled with mysteries as eerie and confounding as the Dyatlov Pass incident. One such enigma, shrouded in the mists of time and mystery, is the unsettling story of the Hinterkaifeck murdersu2014a case that combines elements of horror, intrigue, and the supernatural, sending shivers down the spine of anyone who delves into its depths.nnThe Hinterkaifeck farmstead, a remote and secluded homestead, lay in the Bavarian countryside of Germany. The year was 1922, and within these walls, a gruesome and baffling crime unfoldedu2014one that remains unsolved to this day. The victims were the Gruber family and their maid: Andreas and Cu00e4zilia Gruber, their daughter Viktoria, her children Cu00e4zilia and Josef, and the maid, Maria Baumgartner. Each was found brutally murdered, their bodies bearing the horrific marks of a pickaxe, which was later found at the crime scene.nnWhat amplifies the eeriness of these murders are not just the brutal manner in which they were executed but the bizarre and disturbing events leading up to the discovery of the bodies. Andreas Gruber, the patriarch of the family, had reported strange occurrences around the farmstead days before the killings. ............."
}
],
"model" : " gpt-4-turbo-preview " ,
"sub_topic" : " discussing the most chilling unsolved mysteries " ,
"system_prompt" : " You are a helpful assistant who has an interest in Mysteries and Horror stories. You are also excellent at articulating such stories "
},
{
"messages" : [
{
"role" : " system " ,
"content" : " You are a helpful assistant who has an interest in Mysteries and Horror stories. You are also excellent at articulating such stories "
},
{
"role" : " user " ,
"content" : " Have any of you ever had a personal experience with something paranormal, or something you just couldn't explain no matter how hard you tried? "
},
{
"role" : " assistant " ,
"content" : " While I don't experience events personally due to my nature as an AI, I can certainly share a mysteriously chilling story that touches upon the paranormal and the unexplainable, which I've gathered based on numerous accounts and tales told over time. Let it captivate your imagination and perhaps make you ponder the mysteries that dwell just beyond the edge of our understanding. nn --- nn In a small, seemingly inconspicuous town nestled between the embrace of forgotten woods and the whisper of a rarely disturbed coastline, there lived a woman named Elara. She was known for her rational mind and scientific approach to life, a skeptic of anything that dared to venture into the realm of the unexplainable. However, her beliefs were put to the test one autumn evening. ............. "
},
{
"role" : " user " ,
"content" : " That's a captivating story! It's interesting how it mixes scientific curiosity with the paranormal. Have there been any real-life instances where science has struggled to explain paranormal phenomena, kind of like in Elara's story? "
},
{
"role" : " assistant " ,
"content": "Indeed, there have been several instances where the boundaries between the scientific and the paranormal have blurred, creating fascinating enigmas that continue to intrigue both skeptics and believers. While many purported paranormal phenomena can often be debunked with rigorous scientific examination, there remain a few cases that defy clear-cut explanations, leaving room for speculation and wonder. Here are a few notable examples:nn### 1. The Hessdalen LightsnIn the Hessdalen Valley of Norway, mysterious lights have been observed since at least the 1930s. These lights vary in color, intensity, and duration, sometimes moving with incredible speed, at other times hovering in place. Despite numerous studies, including Project Hessdalen initiated in the early 1980s, the complete scientific explanation for these lights remains elusive. Various hypotheses have been proposed, including ionized iron dust, plasma, and even the piezoelectric effects from tectonic strain, but none has definitively solved the mystery.nn### 2. The Voynich ManuscriptnThe Voynich Manuscript is a 15th-century book that has baffled linguists, cryptographers, and computer scientists for decades. Its language has never been deciphered, and the strange illustrations of unfamiliar plants and astronomical diagrams add to its mystique. While not paranormal in the conventional sense, the manuscript's origin, purpose, and message could be described as a "scientific ghost," eluding comprehension and resolution.nn### 3. ............."
}
],
"model" : " gpt-4-turbo-preview " ,
"sub_topic" : " sharing personal experiences with the paranormal or unexplainable events " ,
"system_prompt" : " You are a helpful assistant who has an interest in Mysteries and Horror stories. You are also excellent at articulating such stories "
}
],
"topic" : " Mysteries and Horror stories " ,
"threads" : 2 ,
"length" : 2
}Untuk melihat output lengkap, lihat direktori contoh
Kontribusi inilah yang membuat komunitas open source menjadi tempat yang luar biasa untuk belajar, menginspirasi, dan berkreasi. Setiap kontribusi yang Anda berikan sangat dihargai .
Jika Anda memiliki saran yang dapat membuat ini lebih baik, harap fork repo dan buat permintaan tarik. Anda juga dapat membuka masalah dengan tag "peningkatan". Jangan lupa untuk memberi proyek ini bintang! Terima kasih lagi!
git checkout -b feature/AmazingFeature )git commit -m 'Add some AmazingFeature' )git push origin feature/AmazingFeature )Kontribusi dapat dilakukan dengan membuat kode yang ada menjadi efisien dan bersih, menambahkan mesin baru (contoh mesin dengan panduan sudah diberikan di dalam autodata/mesin), menambahkan penganalisis data pelatihan LLM, membuat skrip pengujian, dll.
Analisis Data Data LLM menggunakan Bert Based Text Classifier: Sedang Berlangsung
Didistribusikan di bawah Lisensi GPL-3.0. Lihat LICENSE.txt untuk informasi lebih lanjut.
