Edit model card
YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/model-cards#model-card-metadata)

Model Details

  • Model type: Question Generation
  • Language(s) (NLP): Korean
  • Finetuned from model : KoBART

How to Get Started with the Model

for training

from transformers import PreTrainedTokenizerFast, BartForConditionalGeneration

tokenizer = PreTrainedTokenizerFast.from_pretrained("zzrng76/AISYSTEM")
model = BartForConditionalGeneration.from_pretrained("zzrng76/AISYSTEM")

for inference

from transformers import pipeline
from MyPipeline import MyPipeline
pipe = pipeline(model)
kwargs= {
    'frame_size':64,
    'hop_length':32,
    'max_length':512,
    'num_beams':6
}
outs = pipe.make_question(sample,**kwargs)

for i,out in enumerate(outs):
  print(f"[{i}] {out}")

Summary

input

sample = """
์งˆ๋ฌธ ์ƒ์„ฑ: ์ฑ—<unused0>์˜คํ”ˆ์—์ด์•„์ด๊ฐ€ ๊ฐœ๋ฐœํ•œ ๋Œ€ํ™” ์ „๋ฌธ ์ธ๊ณต์ง€๋Šฅ ์ฑ—๋ด‡์œผ๋กœ, ์ฑ—์€ ์ฑ„ํŒ…์˜ ์ค„์ž„๋ง์ด๊ณ  
GPT๋Š” 'Generated Pre-trained Transformer'์˜ ์•ž ๊ธ€์ž๋ฅผ ๋”ด ๊ฒƒ์ด๋‹ค. ์ฑ—GPT๋Š” ์‚ฌ์šฉ์ž๊ฐ€ ๋Œ€ํ™”์ฐฝ์— ํ…์ŠคํŠธ๋ฅผ ์ž…๋ ฅํ•˜๋ฉด 
๊ทธ์— ๋งž์ถฐ ๋Œ€ํ™”๋ฅผ ํ•จ๊ป˜ ๋‚˜๋ˆ„๋Š” ์„œ๋น„์Šค๋กœ, ๊ณต๊ฐœ ๋‹จ 5์ผ ๋งŒ์— ํ•˜๋ฃจ ์ด์šฉ์ž๊ฐ€ 100๋งŒ ๋ช…์„ ๋ŒํŒŒํ•˜๋ฉด์„œ ๋Œํ’์„ ์ผ์œผํ‚ค๊ธฐ ์‹œ์ž‘ํ–ˆ๋‹ค. 
ํŠนํžˆ ์งˆ๋ฌธ์— ๋Œ€ํ•œ ๋‹ต๋ณ€์€ ๋ฌผ๋ก  ๋…ผ๋ฌธ ์ž‘์„ฑ, ๋ฒˆ์—ญ, ๋…ธ๋ž˜ ์ž‘์‚ฌยท์ž‘๊ณก, ์ฝ”๋”ฉ ์ž‘์—… ๋“ฑ ๊ด‘๋ฒ”์œ„ํ•œ ๋ถ„์•ผ์˜ ์—…๋ฌด ์ˆ˜ํ–‰๊นŒ์ง€ ๊ฐ€๋Šฅํ•˜๋‹ค๋Š” ์ ์—์„œ 
๊ธฐ์กด AI์™€๋Š” ํ™•์—ฐํžˆ ๋‹ค๋ฅธ ๋ฉด๋ชจ๋ฅผ ๋ณด์ด๊ณ  ์žˆ๋‹ค. ์ฑ—GPT์˜ ์ œ์ž‘์‚ฌ ์˜คํ”ˆAI๋Š” ์ผ๋ก  ๋จธ์Šคํฌ ํ…Œ์Šฌ๋ผ CEO, ์ƒ˜ ์˜ฌํŠธ๋จผ ์™€์ด์ปด๋น„๋„ค์ดํ„ฐ ์‚ฌ์žฅ(ํ˜„ ์˜คํ”ˆAI CEO) ๋“ฑ์ด 
์ธ๋ฅ˜์—๊ฒŒ ๋„์›€์ด ๋  ๋””์ง€ํ„ธ ์ง€๋Šฅ ๊ฐœ๋ฐœ์„ ๋ชฉํ‘œ๋กœ 2015๋…„ ์„ค๋ฆฝํ•œ ๋น„์˜๋ฆฌ ๋ฒ•์ธ์ด๋‹ค. ๊ทธ๋Ÿฌ๋‹ค 2019๋…„ ์˜๋ฆฌ ์ถ”๊ตฌ๋ฅผ ์œ„ํ•œ ์žํšŒ์‚ฌ๋ฅผ ์ถ”๊ฐ€ ์„ค๋ฆฝํ•˜๋ฉด์„œ 
AI ์‚ฌ์—…์„ ๋ณธ๊ฒฉํ™”ํ–ˆ๋Š”๋ฐ, ๊ทธ๋™์•ˆ ์ธ๊ณต์ง€๋Šฅ ์–ธ์–ด๋ชจ๋ธ โ€˜์ง€ํ”ผํ‹ฐ-3โ€™, ๊ทธ๋ฆผ์„ ๊ทธ๋ฆฌ๋Š” ์ธ๊ณต์ง€๋Šฅ โ€˜๋‹ฌ๋ฆฌ2โ€™ ๋‹ค๊ตญ์–ด ์Œ์„ฑ์ธ์‹ ์ธ๊ณต์ง€๋Šฅ โ€˜์œ„์Šคํผ(Whisper)โ€™ ๋“ฑ์„ 
์„ ๋ณด์—ฌ ์™”๋‹ค. ํŠนํžˆ ์–ธ์–ด์— ํŠนํ™”๋œ ์ธ๊ณต์ง€๋Šฅ์ธ GPT์˜ ๊ฒฝ์šฐ 2018๋…„ GPT-1 ์ถœ์‹œ ์ดํ›„ 2019๋…„ GPT-2, 2020๋…„ GPT-3์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ๋ฒ„์ „์„ ๋†’์—ฌ ์™”์œผ๋ฉฐ, 
2022๋…„ 11์›”์—๋Š” GPT-3.5์— ํ•ด๋‹นํ•˜๋Š” ์ฑ—GPT๋ฅผ ๊ณต๊ฐœํ•˜๋ฉด์„œ ํ™”์ œ์˜ ์ค‘์‹ฌ์— ์„ฐ๋‹ค. GPT ์„ฑ๋Šฅ์€ ๋งค๊ฐœ๋ณ€์ˆ˜(ํŒŒ๋ผ๋ฏธํ„ฐ) ๊ฐœ์ˆ˜๊ฐ€ ์ค‘์š”ํ•œ๋ฐ, 
GPT-3๋Š” GPT-1๋ณด๋‹ค 1500๋ฐฐ ๋งŽ์€ ๋งค๊ฐœ๋ณ€์ˆ˜(1750์–ต ๊ฐœ)๋ฅผ ํ™œ์šฉํ•œ ๊ฒƒ์ด๋‹ค. ์ฑ—GPT๋Š” ์ด GPT-3์— ๊ฐ•ํ™”ํ•™์Šต์„ ์ ์šฉํ•ด ๋”์šฑ ์—…๊ทธ๋ ˆ์ด๋“œํ•œ GPT-3.5๋ฅผ 
๊ธฐ๋ฐ˜์œผ๋กœ ๊ฐœ๋ฐœ๋๋Š”๋ฐ,  ์˜คํ”ˆAI๋Š” 2023๋…„ ์ธ๊ฐ„์˜ ์‹œ๋ƒ…์Šค ์ˆ˜์™€ ๋น„์Šทํ•œ ์ˆ˜์ค€์˜ 100์กฐ ๊ฐœ ๋งค๊ฐœ๋ณ€์ˆ˜๋ฅผ ๊ฐ–์ถ˜ GPT-4๋ฅผ ๋‚ด๋†“๋Š”๋‹ค๋Š” ๊ณ„ํš์„ ๊ณต๊ฐœํ•œ ๋ฐ 
์ด์–ด GPT-4๋ฅผ 2023๋…„ 3์›” 14์ผ ๊ณต๊ฐœํ–ˆ๋‹ค. ํ•œํŽธ, ์˜คํ”ˆAI ์„ค๋ฆฝ์ž ๋จธ์Šคํฌ๋Š” 2018๋…„ ์˜คํ”ˆAI ์ด์‚ฌํšŒ์—์„œ ์ „๊ฒฉ ์‚ฌ์ž„ํ–ˆ์œผ๋ฉฐ, ์ด๋•Œ ๋ณด์œ ํ•˜๊ณ  ์žˆ๋˜ ์ง€๋ถ„๋„ 
๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ์— ๋งค๊ฐํ•œ ๊ฒƒ์œผ๋กœ ์•Œ๋ ค์กŒ๋‹ค. MS๋Š” 2019๋…„ ์˜คํ”ˆAI์— 10์–ต ๋‹ฌ๋Ÿฌ(1์กฐ 2000์–ต ์›)๋ฅผ ํˆฌ์žํ–ˆ๊ณ , 2023๋…„ 1์›”์—๋Š” 100์–ต ๋‹ฌ๋Ÿฌ(12์กฐ ์›)๋กœ 
์ถ”์ •๋˜๋Š” ๊ธˆ์•ก์„ ์ถ”๊ฐ€ ํˆฌ์žํ•œ๋‹ค๊ณ  ๋ฐœํ‘œํ–ˆ๋Š”๋ฐ ์ด๋ ‡๊ฒŒ ๋  ๊ฒฝ์šฐ ์˜คํ”ˆAI์˜ MS ์ง€๋ถ„์€ 49%์— ์ด๋ฅด๊ฒŒ ๋œ๋‹ค.  
"""

output

[0]  ์˜คํ”ˆ์—์ด์•„์ด๊ฐ€ ๊ฐœ๋ฐœํ•œ ๋Œ€ํ™” ์ „๋ฌธ ์ธ๊ณต์ง€๋Šฅ ์ฑ—๋ด‡์˜ ์ด๋ฆ„์€ ๋ฌด์—‡์ธ๊ฐ€?
[1]  ์ฑ—GPT์˜ ์ œ์ž‘์‚ฌ ์˜คํ”ˆAI๋Š” ์–ด๋–ค ํšŒ์‚ฌ์ธ๊ฐ€?
[2]  ์ผ๋ก  ๋จธ์Šคํฌ ํ…Œ์Šฌ๋ผ CEO์™€ ์ƒ˜ ์˜ฌํŠธ๋จผ ์™€์ด์ปด๋น„๋„ค์ดํ„ฐ ์‚ฌ์žฅํ˜„ ์˜คํ”ˆAI CEO๋Š” ๋ˆ„๊ตฌ์ธ๊ฐ€?
[3]  ์ฑ—GPT์˜ ๋งค๊ฐœ๋ณ€์ˆ˜ํŒŒ๋ผ๋ฏธํ„ฐ ๊ฐœ์ˆ˜๋Š” ๋ช‡ ๊ฐœ์ธ๊ฐ€?
[4]  ์˜คํ”ˆAI๊ฐ€ ์ฑ—GPT๋ฅผ ๊ณต๊ฐœํ•œ ๋‚ ์งœ๋Š” ์–ธ์ œ์ธ๊ฐ€?
[5]  ์˜คํ”ˆAI๊ฐ€ ๊ณต๊ฐœํ•œ GPT4์˜ ๋งค๊ฐœ๋ณ€์ˆ˜๋Š” ๋ฌด์—‡์ธ๊ฐ€?
Downloads last month
0
Safetensors
Model size
124M params
Tensor type
F32
ยท
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.