All Projects → ksjae → Kogpt

ksjae / Kogpt

GPT-2 pretrained on Korean datasets.

Labels

Projects that are alternatives of or similar to Kogpt

Click2analyze Androiddevchallenge
An app to analyze the text and fixing the anomaly of the message that deviates from what is standard, normal, or expected. #AndroidDevChallenge
Stars: ✭ 20 (-35.48%)
Mutual labels:  ai
Bfgs Neldermead Trustregion
Python implementation of some numerical (optimization) methods
Stars: ✭ 8 (-74.19%)
Mutual labels:  ai
Gold Miner
🥇掘金翻译计划,可能是世界最大最好的英译中技术社区,最懂读者和译者的翻译平台:
Stars: ✭ 29,872 (+96261.29%)
Mutual labels:  ai
Game Server
Distributed Java game server, including cluster management server, gateway server, hall server, game logic server, background monitoring server and a running web version of fishing. State machine, behavior tree, A* pathfinding, navigation mesh and other AI tools
Stars: ✭ 916 (+2854.84%)
Mutual labels:  ai
Ai Doctor
Machine learning discord bot , Detects diseases within seconds with just an xray scan
Stars: ✭ 27 (-12.9%)
Mutual labels:  ai
Machine Learning Experiments
🤖 Interactive Machine Learning experiments: 🏋️models training + 🎨models demo
Stars: ✭ 841 (+2612.9%)
Mutual labels:  ai
Openhabai
Train Neuronal networks to automate your home
Stars: ✭ 19 (-38.71%)
Mutual labels:  ai
Sarif
A distributed system serving as a personal assistant
Stars: ✭ 29 (-6.45%)
Mutual labels:  ai
Re Verb
speaker diarization system using an LSTM
Stars: ✭ 27 (-12.9%)
Mutual labels:  ai
Awareness
The new architecture of co-computation for data processing and machine learning.
Stars: ✭ 11 (-64.52%)
Mutual labels:  ai
Sliding puzzle
Swift implementation of the Sliding Puzzle game with Iterative Deepening A* AI Solver.
Stars: ✭ 25 (-19.35%)
Mutual labels:  ai
Workshops mathematics
Share resources, assets and information for the math sessions 🚀
Stars: ✭ 26 (-16.13%)
Mutual labels:  ai
Autodl
Automated Deep Learning without ANY human intervention. 1'st Solution for AutoDL [email protected]
Stars: ✭ 854 (+2654.84%)
Mutual labels:  ai
Tetrisai
The old school Tetris game in addition with an AI that learns evolutionary how to play this game
Stars: ✭ 22 (-29.03%)
Mutual labels:  ai
Machine Learning Open Source
Monthly Series - Machine Learning Top 10 Open Source Projects
Stars: ✭ 943 (+2941.94%)
Mutual labels:  ai
Ncappzoo
Contains examples for the Movidius Neural Compute Stick.
Stars: ✭ 902 (+2809.68%)
Mutual labels:  ai
Elyra
Elyra extends JupyterLab Notebooks with an AI centric approach.
Stars: ✭ 839 (+2606.45%)
Mutual labels:  ai
Openkore
A free/open source client and automation tool for Ragnarok Online
Stars: ✭ 956 (+2983.87%)
Mutual labels:  ai
Ai Programmer
Using artificial intelligence and genetic algorithms to automatically write programs. Tutorial: http://www.primaryobjects.com/cms/article149
Stars: ✭ 948 (+2958.06%)
Mutual labels:  ai
Awesome Ai Books
Some awesome AI related books and pdfs for learning and downloading, also apply some playground models for learning
Stars: ✭ 855 (+2658.06%)
Mutual labels:  ai

language: ko tags:

  • kogpt2
  • gpt2
  • text-generation license: MIT datasets:
  • array of dataset identifiers metrics:
  • array of metric identifiers

Model name

KoGPT2

Buy Me A Coffee

Open In Colab Demo available at : 아무말 대잔치

Model description

GPT2 and GPT3 trained on ~40GB of Korean datasets. see the included json files for hyperparameter details.

Available models (Training ATM):

  • KoGPT2-base(117M)
  • KoGPT2-medium(345M)
  • KoGPT2-large(774M)
  • KoGPT2-xlarge(1.5B)
  • KoGPT2-2.7B TBA

Models are available as TF checkpoint files Training script or Huggingface transformers compatible ones

n_ctx available : 1024 2048 384

Intended uses & limitations

Intended for Korean text generation for ai-text-adventure(https://github.com/ksjae/ai-text-adventure) with PPLM.

Downloads

Download files from links at the Releases tab. Alternatively, it is available from my own server(wget-friendly)

How to use

Try out on colab

or go to KoGPT2-train and use scripts/demo.py

Limitations and bias

v0.1 may have faulty tokenizers, producing bad outputs.

v0.2+ be GPT2 with n_ctx of 2048. True form of GPT-3 implementation(alternating layers) will not be available within the year.

If other limitations or errors are found, please open an issue.

Training data

Initialized with GPT(774M,https://github.com/openai/gpt-2/blob/master/model_card.md).

The following data was used, and is available for redistribution here:

  • Namuwiki database dump, Early 2020
  • KCC(Kookmin University Corpus)
  • Dump of Korean Wikipedia
  • NAVER movie reviews
  • Korean news(about 1GB) from Leipzig(a German university)
  • Context data from KorSQUAD questions
  • Parsed Korean CommonCrawl data(WIP)

Please note the completed dataset includes <|endoftext|> tags.

The following data were used, but is unavailable for redistribution:

  • Sejong Corpus
  • '모두의 말뭉치' from corpus.korean.go.kr
  • A PRIVATE collection of korean novels
  • Webcrawl of modern, uploaded text novels('텍본' - If you want to prevent your novel from going in the training set, please contact me and I will blacklist it)
  • Game storylines (with authors' approval)

Training procedure

All hyperparameters are the same as GPT2-large One paragraph per line(TextDataset)

Early models(GPT2-large v0.2 and prior) are trained on 2xTesla V100 for 3~4 weeks. Models up to XL size are trained on v3-8 TPUs.

Eval results

prompt >>> 나는 어두운 숲 속을 거닐고 있다.

v0.1 (xl)

어느새 나는 그 작은 산들을 향해 뛰기 시작했다. 그리고 내 시야에는 이 골짜기에 대한 어떤 동경도, 혹은 동경과 환희조차 엿보였다가 사라졌다가는 사라져 버리고 말았 다. 그러나 나 역시 그것을 믿지 않았다. 아니 그것도 모른다……. 그렇다면 그것은 또 무슨 말인가? 내가 이렇게 말해도 좋을 텐데…… 하지만 어쨌든 이곳엔 왜 그리 먼 곳에서부터 찾아올까 하고 생각해 보았다. 그래서 그는 지금 어느 한 곳만을 헤매면서 자 꾸만 걸어오는 것일까?

v0.2-WEB (xl)

나는 어두운 숲 속을 거닐고 있다. #앨리건트테이블 의 #코코넛젤리 색이 넘 예쁘고 맛나다. #그릭요거트 원물이 들어간 마지막 베이커리 #말차초코케이크 도 맛나고 겉에 초코도 두껍게 씹히고..✨💕💕👍🏻 . . #앨리건트테이블 @eleganttable_ #간식 #카페어니언베이커리베이커 #콩콩볼 #쿠키 #디저트맛집 #디저트카페 #dessert #카페스타그램 #맛스타그램 #먹스타 그램 #먹방 #일상 #daily

v0.2-story (xl)

나는 어두운 숲 속을 거닐고 있다. 정명이 쓴웃음을 지은 채 곳에 걸어나왔다. ... ... ... ... 사망탑 높이까지 올라가야하나?

BibTeX entry and citation info

@unpublished{CitekeyUnpublished,
      author = "Seungjae Kim",
      title  = "Introducing larger KoGPT2",
      year   = 2020
    }
Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].