Skip to content

Latest commit

ย 

History

History
442 lines (377 loc) ยท 83.9 KB

README_ko.md

File metadata and controls

442 lines (377 loc) ยท 83.9 KB



Build GitHub Documentation GitHub release Contributor Covenant DOI

Jax, Pytorch, TensorFlow๋ฅผ ์œ„ํ•œ ์ตœ์ฒจ๋‹จ ์ž์—ฐ์–ด์ฒ˜๋ฆฌ

๐Ÿค— Transformers๋Š” ๋ถ„๋ฅ˜, ์ •๋ณด ์ถ”์ถœ, ์งˆ๋ฌธ ๋‹ต๋ณ€, ์š”์•ฝ, ๋ฒˆ์—ญ, ๋ฌธ์žฅ ์ƒ์„ฑ ๋“ฑ์„ 100๊ฐœ ์ด์ƒ์˜ ์–ธ์–ด๋กœ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ๋Š” ์ˆ˜์ฒœ๊ฐœ์˜ ์‚ฌ์ „ํ•™์Šต๋œ ๋ชจ๋ธ์„ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค. ์šฐ๋ฆฌ์˜ ๋ชฉํ‘œ๋Š” ๋ชจ๋‘๊ฐ€ ์ตœ์ฒจ๋‹จ์˜ NLP ๊ธฐ์ˆ ์„ ์‰ฝ๊ฒŒ ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.

๐Ÿค— Transformers๋Š” ์ด๋Ÿฌํ•œ ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ์„ ๋น ๋ฅด๊ฒŒ ๋‹ค์šด๋กœ๋“œํ•ด ํŠน์ • ํ…์ŠคํŠธ์— ์‚ฌ์šฉํ•˜๊ณ , ์›ํ•˜๋Š” ๋ฐ์ดํ„ฐ๋กœ fine-tuningํ•ด ์ปค๋ฎค๋‹ˆํ‹ฐ๋‚˜ ์šฐ๋ฆฌ์˜ ๋ชจ๋ธ ํ—ˆ๋ธŒ์— ๊ณต์œ ํ•  ์ˆ˜ ์žˆ๋„๋ก API๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค. ๋˜ํ•œ, ๋ชจ๋ธ ๊ตฌ์กฐ๋ฅผ ์ •์˜ํ•˜๋Š” ๊ฐ ํŒŒ์ด์ฌ ๋ชจ๋“ˆ์€ ์™„์ „ํžˆ ๋…๋ฆฝ์ ์ด์—ฌ์„œ ์—ฐ๊ตฌ ์‹คํ—˜์„ ์œ„ํ•ด ์†์‰ฝ๊ฒŒ ์ˆ˜์ •ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

๐Ÿค— Transformers๋Š” ๊ฐ€์žฅ ์œ ๋ช…ํ•œ 3๊ฐœ์˜ ๋”ฅ๋Ÿฌ๋‹ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ์ง€์›ํ•ฉ๋‹ˆ๋‹ค. ์ด๋“ค์€ ์„œ๋กœ ์™„๋ฒฝํžˆ ์—ฐ๋™๋ฉ๋‹ˆ๋‹ค โ€” Jax, PyTorch, TensorFlow. ๊ฐ„๋‹จํ•˜๊ฒŒ ์ด ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ ์ค‘ ํ•˜๋‚˜๋กœ ๋ชจ๋ธ์„ ํ•™์Šตํ•˜๊ณ , ๋˜ ๋‹ค๋ฅธ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋กœ ์ถ”๋ก ์„ ์œ„ํ•ด ๋ชจ๋ธ์„ ๋ถˆ๋Ÿฌ์˜ฌ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์˜จ๋ผ์ธ ๋ฐ๋ชจ

๋Œ€๋ถ€๋ถ„์˜ ๋ชจ๋ธ์„ ๋ชจ๋ธ ํ—ˆ๋ธŒ ํŽ˜์ด์ง€์—์„œ ๋ฐ”๋กœ ํ…Œ์ŠคํŠธํ•ด๋ณผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ณต๊ฐœ ๋ฐ ๋น„๊ณต๊ฐœ ๋ชจ๋ธ์„ ์œ„ํ•œ ๋น„๊ณต๊ฐœ ๋ชจ๋ธ ํ˜ธ์ŠคํŒ…, ๋ฒ„์ „ ๊ด€๋ฆฌ, ์ถ”๋ก  API๋„ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.

์˜ˆ์‹œ:

Transformer์™€ ๊ธ€์“ฐ๊ธฐ ๋Š” ์ด ์ €์žฅ์†Œ์˜ ํ…์ŠคํŠธ ์ƒ์„ฑ ๋Šฅ๋ ฅ์— ๊ด€ํ•œ Hugging Face ํŒ€์˜ ๊ณต์‹ ๋ฐ๋ชจ์ž…๋‹ˆ๋‹ค.

Hugging Face ํŒ€์˜ ์ปค์Šคํ…€ ์ง€์›์„ ์›ํ•œ๋‹ค๋ฉด

HuggingFace Expert Acceleration Program

ํ€ต ํˆฌ์–ด

์›ํ•˜๋Š” ํ…์ŠคํŠธ์— ๋ฐ”๋กœ ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก, ์šฐ๋ฆฌ๋Š” pipeline API๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค. Pipeline์€ ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ๊ณผ ๊ทธ ๋ชจ๋ธ์„ ํ•™์Šตํ•  ๋•Œ ์ ์šฉํ•œ ์ „์ฒ˜๋ฆฌ ๋ฐฉ์‹์„ ํ•˜๋‚˜๋กœ ํ•ฉ์นฉ๋‹ˆ๋‹ค. ๋‹ค์Œ์€ ๊ธ์ •์ ์ธ ํ…์ŠคํŠธ์™€ ๋ถ€์ •์ ์ธ ํ…์ŠคํŠธ๋ฅผ ๋ถ„๋ฅ˜ํ•˜๊ธฐ ์œ„ํ•ด pipeline์„ ์‚ฌ์šฉํ•œ ๊ฐ„๋‹จํ•œ ์˜ˆ์‹œ์ž…๋‹ˆ๋‹ค:

>>> from transformers import pipeline

# Allocate a pipeline for sentiment-analysis
>>> classifier = pipeline('sentiment-analysis')
>>> classifier('We are very happy to introduce pipeline to the transformers repository.')
[{'label': 'POSITIVE', 'score': 0.9996980428695679}]

์ฝ”๋“œ์˜ ๋‘๋ฒˆ์งธ ์ค„์€ pipeline์ด ์‚ฌ์šฉํ•˜๋Š” ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ์„ ๋‹ค์šด๋กœ๋“œํ•˜๊ณ  ์บ์‹œ๋กœ ์ €์žฅํ•ฉ๋‹ˆ๋‹ค. ์„ธ๋ฒˆ์งธ ์ค„์—์„  ๊ทธ ๋ชจ๋ธ์ด ์ฃผ์–ด์ง„ ํ…์ŠคํŠธ๋ฅผ ํ‰๊ฐ€ํ•ฉ๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ ๋ชจ๋ธ์€ 99.97%์˜ ํ™•๋ฅ ๋กœ ํ…์ŠคํŠธ๊ฐ€ ๊ธ์ •์ ์ด๋ผ๊ณ  ํ‰๊ฐ€ํ–ˆ์Šต๋‹ˆ๋‹ค.

๋งŽ์€ NLP ๊ณผ์ œ๋“ค์„ pipeline์œผ๋กœ ๋ฐ”๋กœ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, ์งˆ๋ฌธ๊ณผ ๋ฌธ๋งฅ์ด ์ฃผ์–ด์ง€๋ฉด ์†์‰ฝ๊ฒŒ ๋‹ต๋ณ€์„ ์ถ”์ถœํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค:

>>> from transformers import pipeline

# Allocate a pipeline for question-answering
>>> question_answerer = pipeline('question-answering')
>>> question_answerer({
...     'question': 'What is the name of the repository ?',
...     'context': 'Pipeline has been included in the huggingface/transformers repository'
... })
{'score': 0.30970096588134766, 'start': 34, 'end': 58, 'answer': 'huggingface/transformers'}

๋‹ต๋ณ€๋ฟ๋งŒ ์•„๋‹ˆ๋ผ, ์—ฌ๊ธฐ์— ์‚ฌ์šฉ๋œ ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ์€ ํ™•์‹ ๋„์™€ ํ† ํฌ๋‚˜์ด์ฆˆ๋œ ๋ฌธ์žฅ ์† ๋‹ต๋ณ€์˜ ์‹œ์ž‘์ , ๋์ ๊นŒ์ง€ ๋ฐ˜ํ™˜ํ•ฉ๋‹ˆ๋‹ค. ์ด ํŠœํ† ๋ฆฌ์–ผ์—์„œ pipeline API๊ฐ€ ์ง€์›ํ•˜๋Š” ๋‹ค์–‘ํ•œ ๊ณผ์ œ๋ฅผ ํ™•์ธํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์ฝ”๋“œ 3์ค„๋กœ ์›ํ•˜๋Š” ๊ณผ์ œ์— ๋งž๊ฒŒ ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ์„ ๋‹ค์šด๋กœ๋“œ ๋ฐ›๊ณ  ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๋‹ค์Œ์€ PyTorch ๋ฒ„์ „์ž…๋‹ˆ๋‹ค:

>>> from transformers import AutoTokenizer, AutoModel

>>> tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
>>> model = AutoModel.from_pretrained("bert-base-uncased")

>>> inputs = tokenizer("Hello world!", return_tensors="pt")
>>> outputs = model(**inputs)

๋‹ค์Œ์€ TensorFlow ๋ฒ„์ „์ž…๋‹ˆ๋‹ค:

>>> from transformers import AutoTokenizer, TFAutoModel

>>> tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
>>> model = TFAutoModel.from_pretrained("bert-base-uncased")

>>> inputs = tokenizer("Hello world!", return_tensors="tf")
>>> outputs = model(**inputs)

ํ† ํฌ๋‚˜์ด์ €๋Š” ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ์˜ ๋ชจ๋“  ์ „์ฒ˜๋ฆฌ๋ฅผ ์ฑ…์ž„์ง‘๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  (์œ„์˜ ์˜ˆ์‹œ์ฒ˜๋Ÿผ) 1๊ฐœ์˜ ์ŠคํŠธ๋ง์ด๋‚˜ ๋ฆฌ์ŠคํŠธ๋„ ์ฒ˜๋ฆฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ํ† ํฌ๋‚˜์ด์ €๋Š” ๋”•์…”๋„ˆ๋ฆฌ๋ฅผ ๋ฐ˜ํ™˜ํ•˜๋Š”๋ฐ, ์ด๋Š” ๋‹ค์šด์ŠคํŠธ๋ฆผ ์ฝ”๋“œ์— ์‚ฌ์šฉํ•˜๊ฑฐ๋‚˜ ์–ธํŒจํ‚น ์—ฐ์‚ฐ์ž ** ๋ฅผ ์ด์šฉํ•ด ๋ชจ๋ธ์— ๋ฐ”๋กœ ์ „๋‹ฌํ•  ์ˆ˜๋„ ์žˆ์Šต๋‹ˆ๋‹ค.

๋ชจ๋ธ ์ž์ฒด๋Š” ์ผ๋ฐ˜์ ์œผ๋กœ ์‚ฌ์šฉ๋˜๋Š” Pytorch nn.Module๋‚˜ TensorFlow tf.keras.Model์ž…๋‹ˆ๋‹ค. ์ด ํŠœํ† ๋ฆฌ์–ผ์€ ์ด๋Ÿฌํ•œ ๋ชจ๋ธ์„ ํ‘œ์ค€์ ์ธ PyTorch๋‚˜ TensorFlow ํ•™์Šต ๊ณผ์ •์—์„œ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ๋ฒ•, ๋˜๋Š” ์ƒˆ๋กœ์šด ๋ฐ์ดํ„ฐ๋กœ fine-tuneํ•˜๊ธฐ ์œ„ํ•ด Trainer API๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ์„ค๋ช…ํ•ด์ค๋‹ˆ๋‹ค.

์™œ transformers๋ฅผ ์‚ฌ์šฉํ•ด์•ผ ํ• ๊นŒ์š”?

  1. ์†์‰ฝ๊ฒŒ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ์ตœ์ฒจ๋‹จ ๋ชจ๋ธ:

    • NLU์™€ NLG ๊ณผ์ œ์—์„œ ๋›ฐ์–ด๋‚œ ์„ฑ๋Šฅ์„ ๋ณด์ž…๋‹ˆ๋‹ค.
    • ๊ต์œก์ž ์‹ค๋ฌด์ž์—๊ฒŒ ์ง„์ž… ์žฅ๋ฒฝ์ด ๋‚ฎ์Šต๋‹ˆ๋‹ค.
    • 3๊ฐœ์˜ ํด๋ž˜์Šค๋งŒ ๋ฐฐ์šฐ๋ฉด ๋ฐ”๋กœ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
    • ํ•˜๋‚˜์˜ API๋กœ ๋ชจ๋“  ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
  2. ๋” ์ ์€ ๊ณ„์‚ฐ ๋น„์šฉ, ๋” ์ ์€ ํƒ„์†Œ ๋ฐœ์ž๊ตญ:

    • ์—ฐ๊ตฌ์ž๋“ค์€ ๋ชจ๋ธ์„ ๊ณ„์† ๋‹ค์‹œ ํ•™์Šต์‹œํ‚ค๋Š” ๋Œ€์‹  ํ•™์Šต๋œ ๋ชจ๋ธ์„ ๊ณต์œ ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
    • ์‹ค๋ฌด์ž๋“ค์€ ํ•™์Šต์— ํ•„์š”ํ•œ ์‹œ๊ฐ„๊ณผ ๋น„์šฉ์„ ์ ˆ์•ฝํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
    • ์ˆ˜์‹ญ๊ฐœ์˜ ๋ชจ๋ธ ๊ตฌ์กฐ, 2,000๊ฐœ ์ด์ƒ์˜ ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ, 100๊ฐœ ์ด์ƒ์˜ ์–ธ์–ด๋กœ ํ•™์Šต๋œ ๋ชจ๋ธ ๋“ฑ.
  3. ๋ชจ๋ธ์˜ ๊ฐ ์ƒ์• ์ฃผ๊ธฐ์— ์ ํ•ฉํ•œ ํ”„๋ ˆ์ž„์›Œํฌ:

    • ์ฝ”๋“œ 3์ค„๋กœ ์ตœ์ฒจ๋‹จ ๋ชจ๋ธ์„ ํ•™์Šตํ•˜์„ธ์š”.
    • ์ž์œ ๋กญ๊ฒŒ ๋ชจ๋ธ์„ TF2.0๋‚˜ PyTorch ํ”„๋ ˆ์ž„์›Œํฌ๋กœ ๋ณ€ํ™˜ํ•˜์„ธ์š”.
    • ํ•™์Šต, ํ‰๊ฐ€, ๊ณต๊ฐœ ๋“ฑ ๊ฐ ๋‹จ๊ณ„์— ๋งž๋Š” ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์›ํ•˜๋Š”๋Œ€๋กœ ์„ ํƒํ•˜์„ธ์š”.
  4. ํ•„์š”ํ•œ ๋Œ€๋กœ ๋ชจ๋ธ์ด๋‚˜ ์˜ˆ์‹œ๋ฅผ ์ปค์Šคํ„ฐ๋งˆ์ด์ฆˆํ•˜์„ธ์š”:

    • ์šฐ๋ฆฌ๋Š” ์ €์ž๊ฐ€ ๊ณต๊ฐœํ•œ ๊ฒฐ๊ณผ๋ฅผ ์žฌํ˜„ํ•˜๊ธฐ ์œ„ํ•ด ๊ฐ ๋ชจ๋ธ ๊ตฌ์กฐ์˜ ์˜ˆ์‹œ๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.
    • ๋ชจ๋ธ ๋‚ด๋ถ€ ๊ตฌ์กฐ๋Š” ๊ฐ€๋Šฅํ•œ ์ผ๊ด€์ ์œผ๋กœ ๊ณต๊ฐœ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
    • ๋น ๋ฅธ ์‹คํ—˜์„ ์œ„ํ•ด ๋ชจ๋ธ ํŒŒ์ผ์€ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ์™€ ๋…๋ฆฝ์ ์œผ๋กœ ์‚ฌ์šฉ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์™œ transformers๋ฅผ ์‚ฌ์šฉํ•˜์ง€ ๋ง์•„์•ผ ํ• ๊นŒ์š”?

  • ์ด ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋Š” ์‹ ๊ฒฝ๋ง ๋ธ”๋ก์„ ๋งŒ๋“ค๊ธฐ ์œ„ํ•œ ๋ชจ๋“ˆ์ด ์•„๋‹™๋‹ˆ๋‹ค. ์—ฐ๊ตฌ์ž๋“ค์ด ์—ฌ๋Ÿฌ ํŒŒ์ผ์„ ์‚ดํŽด๋ณด์ง€ ์•Š๊ณ  ๋ฐ”๋กœ ๊ฐ ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก, ๋ชจ๋ธ ํŒŒ์ผ ์ฝ”๋“œ์˜ ์ถ”์ƒํ™” ์ˆ˜์ค€์„ ์ ์ •ํ•˜๊ฒŒ ์œ ์ง€ํ–ˆ์Šต๋‹ˆ๋‹ค.
  • ํ•™์Šต API๋Š” ๋ชจ๋“  ๋ชจ๋ธ์— ์ ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ๋งŒ๋“ค์–ด์ง€์ง„ ์•Š์•˜์ง€๋งŒ, ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๊ฐ€ ์ œ๊ณตํ•˜๋Š” ๋ชจ๋ธ๋“ค์— ์ ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ์ตœ์ ํ™”๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ์ผ๋ฐ˜์ ์ธ ๋จธ์‹  ๋Ÿฌ๋‹์„ ์œ„ํ•ด์„ , ๋‹ค๋ฅธ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ์‚ฌ์šฉํ•˜์„ธ์š”.
  • ๊ฐ€๋Šฅํ•œ ๋งŽ์€ ์‚ฌ์šฉ ์˜ˆ์‹œ๋ฅผ ๋ณด์—ฌ๋“œ๋ฆฌ๊ณ  ์‹ถ์–ด์„œ, ์˜ˆ์‹œ ํด๋”์˜ ์Šคํฌ๋ฆฝํŠธ๋ฅผ ์ค€๋น„ํ–ˆ์Šต๋‹ˆ๋‹ค. ์ด ์Šคํฌ๋ฆฝํŠธ๋“ค์„ ์ˆ˜์ • ์—†์ด ํŠน์ •ํ•œ ๋ฌธ์ œ์— ๋ฐ”๋กœ ์ ์šฉํ•˜์ง€ ๋ชปํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ํ•„์š”์— ๋งž๊ฒŒ ์ผ๋ถ€ ์ฝ”๋“œ๋ฅผ ์ˆ˜์ •ํ•ด์•ผ ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์„ค์น˜

pip๋กœ ์„ค์น˜ํ•˜๊ธฐ

์ด ์ €์žฅ์†Œ๋Š” Python 3.6+, Flax 0.3.2+, PyTorch 1.3.1+, TensorFlow 2.3+์—์„œ ํ…Œ์ŠคํŠธ ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.

๊ฐ€์ƒ ํ™˜๊ฒฝ์— ๐Ÿค— Transformers๋ฅผ ์„ค์น˜ํ•˜์„ธ์š”. Python ๊ฐ€์ƒ ํ™˜๊ฒฝ์— ์ต์ˆ™ํ•˜์ง€ ์•Š๋‹ค๋ฉด, ์‚ฌ์šฉ์ž ๊ฐ€์ด๋“œ๋ฅผ ํ™•์ธํ•˜์„ธ์š”.

์šฐ์„ , ์‚ฌ์šฉํ•  Python ๋ฒ„์ „์œผ๋กœ ๊ฐ€์ƒ ํ™˜๊ฒฝ์„ ๋งŒ๋“ค๊ณ  ์‹คํ–‰ํ•˜์„ธ์š”.

๊ทธ ๋‹ค์Œ, Flax, PyTorch, TensorFlow ์ค‘ ์ ์–ด๋„ ํ•˜๋‚˜๋Š” ์„ค์น˜ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ํ”Œ๋žซํผ์— ๋งž๋Š” ์„ค์น˜ ๋ช…๋ น์–ด๋ฅผ ํ™•์ธํ•˜๊ธฐ ์œ„ํ•ด TensorFlow ์„ค์น˜ ํŽ˜์ด์ง€, PyTorch ์„ค์น˜ ํŽ˜์ด์ง€, Flax ์„ค์น˜ ํŽ˜์ด์ง€๋ฅผ ํ™•์ธํ•˜์„ธ์š”.

์ด๋“ค ์ค‘ ์ ์–ด๋„ ํ•˜๋‚˜๊ฐ€ ์„ค์น˜๋˜์—ˆ๋‹ค๋ฉด, ๐Ÿค— Transformers๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด pip์„ ์ด์šฉํ•ด ์„ค์น˜ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค:

pip install transformers

์˜ˆ์‹œ๋“ค์„ ์ฒดํ—˜ํ•ด๋ณด๊ณ  ์‹ถ๊ฑฐ๋‚˜, ์ตœ์ตœ์ตœ์ฒจ๋‹จ ์ฝ”๋“œ๋ฅผ ์›ํ•˜๊ฑฐ๋‚˜, ์ƒˆ๋กœ์šด ๋ฒ„์ „์ด ๋‚˜์˜ฌ ๋•Œ๊นŒ์ง€ ๊ธฐ๋‹ค๋ฆด ์ˆ˜ ์—†๋‹ค๋ฉด ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ์†Œ์Šค์—์„œ ๋ฐ”๋กœ ์„ค์น˜ํ•˜์…”์•ผ ํ•ฉ๋‹ˆ๋‹ค.

conda๋กœ ์„ค์น˜ํ•˜๊ธฐ

Transformers ๋ฒ„์ „ v4.0.0๋ถ€ํ„ฐ, conda ์ฑ„๋„์ด ์ƒ๊ฒผ์Šต๋‹ˆ๋‹ค: huggingface.

๐Ÿค— Transformers๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด conda๋กœ ์„ค์น˜ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค:

conda install -c huggingface transformers

Flax, PyTorch, TensorFlow ์„ค์น˜ ํŽ˜์ด์ง€์—์„œ ์ด๋“ค์„ conda๋กœ ์„ค์น˜ํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ํ™•์ธํ•˜์„ธ์š”.

๋ชจ๋ธ ๊ตฌ์กฐ

๐Ÿค— Transformers๊ฐ€ ์ œ๊ณตํ•˜๋Š” ๋ชจ๋“  ๋ชจ๋ธ ์ฒดํฌํฌ์ธํŠธ ๋Š” huggingface.co ๋ชจ๋ธ ํ—ˆ๋ธŒ์— ์™„๋ฒฝํžˆ ์—ฐ๋™๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ๊ฐœ์ธ๊ณผ ๊ธฐ๊ด€์ด ๋ชจ๋ธ ํ—ˆ๋ธŒ์— ์ง์ ‘ ์—…๋กœ๋“œํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

ํ˜„์žฌ ์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ๋ชจ๋ธ ์ฒดํฌํฌ์ธํŠธ์˜ ๊ฐœ์ˆ˜:

๐Ÿค— Transformers๋Š” ๋‹ค์Œ ๋ชจ๋ธ๋“ค์„ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค (๊ฐ ๋ชจ๋ธ์˜ ์š”์•ฝ์€ ์—ฌ๊ธฐ์„œ ํ™•์ธํ•˜์„ธ์š”):

  1. ALBERT (from Google Research and the Toyota Technological Institute at Chicago) released with the paper ALBERT: A Lite BERT for Self-supervised Learning of Language Representations, by Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut.
  2. ALIGN (Google Research ์—์„œ ์ œ๊ณต)์€ Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig.์˜ Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  3. AltCLIP (from BAAI) released with the paper AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities by Chen, Zhongzhi and Liu, Guang and Zhang, Bo-Wen and Ye, Fulong and Yang, Qinghong and Wu, Ledell.
  4. Audio Spectrogram Transformer (from MIT) released with the paper AST: Audio Spectrogram Transformer by Yuan Gong, Yu-An Chung, James Glass.
  5. BART (from Facebook) released with the paper BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer.
  6. BARThez (from ร‰cole polytechnique) released with the paper BARThez: a Skilled Pretrained French Sequence-to-Sequence Model by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
  7. BARTpho (from VinAI Research) released with the paper BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese by Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen.
  8. BEiT (from Microsoft) released with the paper BEiT: BERT Pre-Training of Image Transformers by Hangbo Bao, Li Dong, Furu Wei.
  9. BERT (from Google) released with the paper BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding by Jacob Devlin, Ming-Wei Chang, Kenton Lee and Kristina Toutanova.
  10. BERT For Sequence Generation (from Google) released with the paper Leveraging Pre-trained Checkpoints for Sequence Generation Tasks by Sascha Rothe, Shashi Narayan, Aliaksei Severyn.
  11. BERTweet (from VinAI Research) released with the paper BERTweet: A pre-trained language model for English Tweets by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen.
  12. BigBird-Pegasus (from Google Research) released with the paper Big Bird: Transformers for Longer Sequences by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed.
  13. BigBird-RoBERTa (from Google Research) released with the paper Big Bird: Transformers for Longer Sequences by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed.
  14. BioGpt (from Microsoft Research AI4Science) released with the paper BioGPT: generative pre-trained transformer for biomedical text generation and mining by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu.
  15. BiT (from Google AI) released with the paper [Big Transfer (BiT) by Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Joan Puigcerver, Jessica Yung, Sylvain Gelly, Neil Houlsby.
  16. Blenderbot (from Facebook) released with the paper Recipes for building an open-domain chatbot by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
  17. BlenderbotSmall (from Facebook) released with the paper Recipes for building an open-domain chatbot by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
  18. BLIP (from Salesforce) released with the paper BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation by Junnan Li, Dongxu Li, Caiming Xiong, Steven Hoi.
  19. BLIP-2 (Salesforce ์—์„œ ์ œ๊ณต)์€ Junnan Li, Dongxu Li, Silvio Savarese, Steven Hoi.์˜ BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  20. BLOOM (from BigScience workshop) released by the BigScience Workshop.
  21. BORT (Alexa ์—์„œ) Adrian de Wynter and Daniel J. Perry ์˜ Optimal Subarchitecture Extraction For BERT ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  22. BridgeTower (from Harbin Institute of Technology/Microsoft Research Asia/Intel Labs) released with the paper BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning by Xiao Xu, Chenfei Wu, Shachar Rosenman, Vasudev Lal, Wanxiang Che, Nan Duan.
  23. ByT5 (Google Research ์—์„œ) Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel ์˜ ByT5: Towards a token-free future with pre-trained byte-to-byte models ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  24. CamemBERT (Inria/Facebook/Sorbonne ์—์„œ) Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suรกrez*, Yoann Dupont, Laurent Romary, ร‰ric Villemonte de la Clergerie, Djamรฉ Seddah and Benoรฎt Sagot ์˜ CamemBERT: a Tasty French Language Model ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  25. CANINE (Google Research ์—์„œ) Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting ์˜ CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  26. Chinese-CLIP (OFA-Sys ์—์„œ) An Yang, Junshu Pan, Junyang Lin, Rui Men, Yichang Zhang, Jingren Zhou, Chang Zhou ์˜ Chinese CLIP: Contrastive Vision-Language Pretraining in Chinese ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  27. CLAP (LAION-AI ์—์„œ ์ œ๊ณต)์€ Yusong Wu, Ke Chen, Tianyu Zhang, Yuchen Hui, Taylor Berg-Kirkpatrick, Shlomo Dubnov.์˜ [Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation]https://arxiv.org/abs/2211.06687)๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  28. CLIP (OpenAI ์—์„œ) Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever ์˜ Learning Transferable Visual Models From Natural Language Supervision ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  29. CLIPSeg (University of Gรถttingen ์—์„œ) Timo Lรผddecke and Alexander Ecker ์˜ Image Segmentation Using Text and Image Prompts ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  30. CodeGen (Salesforce ์—์„œ) Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong ์˜ A Conversational Paradigm for Program Synthesis ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  31. Conditional DETR (Microsoft Research Asia ์—์„œ) Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang ์˜ Conditional DETR for Fast Training Convergence ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  32. ConvBERT (YituTech ์—์„œ) Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan ์˜ ConvBERT: Improving BERT with Span-based Dynamic Convolution ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  33. ConvNeXT (Facebook AI ์—์„œ) Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie ์˜ A ConvNet for the 2020s ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  34. ConvNeXTV2 (from Facebook AI) released with the paper ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders by Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon, Saining Xie.
  35. CPM (Tsinghua University ์—์„œ) Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun ์˜ CPM: A Large-scale Generative Chinese Pre-trained Language Model ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  36. CTRL (Salesforce ์—์„œ) Nitish Shirish Keskar*, Bryan McCann*, Lav R. Varshney, Caiming Xiong and Richard Socher ์˜ CTRL: A Conditional Transformer Language Model for Controllable Generation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  37. CvT (Microsoft ์—์„œ) Haiping Wu, Bin Xiao, Noel Codella, Mengchen Liu, Xiyang Dai, Lu Yuan, Lei Zhang ์˜ CvT: Introducing Convolutions to Vision Transformers ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  38. Data2Vec (Facebook ์—์„œ) Alexei Baevski, Wei-Ning Hsu, Qiantong Xu, Arun Babu, Jiatao Gu, Michael Auli ์˜ Data2Vec: A General Framework for Self-supervised Learning in Speech, Vision and Language ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  39. DeBERTa (Microsoft ์—์„œ) Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen ์˜ DeBERTa: Decoding-enhanced BERT with Disentangled Attention ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  40. DeBERTa-v2 (Microsoft ์—์„œ) Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen ์˜ DeBERTa: Decoding-enhanced BERT with Disentangled Attention ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  41. Decision Transformer (Berkeley/Facebook/Google ์—์„œ) Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch ์˜ Decision Transformer: Reinforcement Learning via Sequence Modeling ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  42. Deformable DETR (SenseTime Research ์—์„œ) Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai ์˜ Deformable DETR: Deformable Transformers for End-to-End Object Detection ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  43. DeiT (Facebook ์—์„œ) Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervรฉ Jรฉgou ์˜ Training data-efficient image transformers & distillation through attention ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  44. DETA (The University of Texas at Austin ์—์„œ ์ œ๊ณต)์€ Jeffrey Ouyang-Zhang, Jang Hyun Cho, Xingyi Zhou, Philipp Krรคhenbรผhl.์˜ NMS Strikes Back๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  45. DETR (Facebook ์—์„œ) Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko ์˜ End-to-End Object Detection with Transformers ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  46. DialoGPT (Microsoft Research ์—์„œ) Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan ์˜ DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  47. DiNAT (SHI Labs ์—์„œ) Ali Hassani and Humphrey Shi ์˜ Dilated Neighborhood Attention Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  48. DistilBERT (HuggingFace ์—์„œ) Victor Sanh, Lysandre Debut and Thomas Wolf. The same method has been applied to compress GPT2 into DistilGPT2, RoBERTa into DistilRoBERTa, Multilingual BERT into DistilmBERT and a German version of DistilBERT ์˜ DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  49. DiT (Microsoft Research ์—์„œ) Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei ์˜ DiT: Self-supervised Pre-training for Document Image Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  50. Donut (NAVER ์—์„œ) Geewook Kim, Teakgyu Hong, Moonbin Yim, Jeongyeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park ์˜ OCR-free Document Understanding Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  51. DPR (Facebook ์—์„œ) Vladimir Karpukhin, Barlas OฤŸuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih ์˜ Dense Passage Retrieval for Open-Domain Question Answering ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  52. DPT (Intel Labs ์—์„œ) Renรฉ Ranftl, Alexey Bochkovskiy, Vladlen Koltun ์˜ Vision Transformers for Dense Prediction ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  53. EfficientFormer (from Snap Research) released with the paper EfficientFormer: Vision Transformers at MobileNetSpeed by Yanyu Li, Geng Yuan, Yang Wen, Ju Hu, Georgios Evangelidis, Sergey Tulyakov, Yanzhi Wang, Jian Ren.
  54. EfficientNet (from Google Brain) released with the paper EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks by Mingxing Tan, Quoc V. Le.
  55. ELECTRA (Google Research/Stanford University ์—์„œ) Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning ์˜ ELECTRA: Pre-training text encoders as discriminators rather than generators ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  56. EncoderDecoder (Google Research ์—์„œ) Sascha Rothe, Shashi Narayan, Aliaksei Severyn ์˜ Leveraging Pre-trained Checkpoints for Sequence Generation Tasks ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  57. ERNIE (Baidu ์—์„œ) Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian, Hua Wu ์˜ ERNIE: Enhanced Representation through Knowledge Integration ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  58. ErnieM (Baidu ์—์„œ ์ œ๊ณต)์€ Xuan Ouyang, Shuohuan Wang, Chao Pang, Yu Sun, Hao Tian, Hua Wu, Haifeng Wang.์˜ ERNIE-M: Enhanced Multilingual Representation by Aligning Cross-lingual Semantics with Monolingual Corpora๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  59. ESM (from Meta AI) are transformer protein language models. ESM-1b was released with the paper Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences by Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus. ESM-1v was released with the paper Language models enable zero-shot prediction of the effects of mutations on protein function by Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu and Alexander Rives. ESM-2 was released with the paper Language models of protein sequences at the scale of evolution enable accurate structure prediction by Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Allan dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Sal Candido, Alexander Rives.
  60. FLAN-T5 (from Google AI) released in the repository google-research/t5x by Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, Albert Webson, Shixiang Shane Gu, Zhuyun Dai, Mirac Suzgun, Xinyun Chen, Aakanksha Chowdhery, Sharan Narang, Gaurav Mishra, Adams Yu, Vincent Zhao, Yanping Huang, Andrew Dai, Hongkun Yu, Slav Petrov, Ed H. Chi, Jeff Dean, Jacob Devlin, Adam Roberts, Denny Zhou, Quoc V. Le, and Jason Wei
  61. FLAN-UL2 (from Google AI) released in the repository google-research/t5x by Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, Albert Webson, Shixiang Shane Gu, Zhuyun Dai, Mirac Suzgun, Xinyun Chen, Aakanksha Chowdhery, Sharan Narang, Gaurav Mishra, Adams Yu, Vincent Zhao, Yanping Huang, Andrew Dai, Hongkun Yu, Slav Petrov, Ed H. Chi, Jeff Dean, Jacob Devlin, Adam Roberts, Denny Zhou, Quoc V. Le, and Jason Wei
  62. FlauBERT (from CNRS) released with the paper FlauBERT: Unsupervised Language Model Pre-training for French by Hang Le, Loรฏc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoรฎt Crabbรฉ, Laurent Besacier, Didier Schwab.
  63. FLAVA (from Facebook AI) released with the paper FLAVA: A Foundational Language And Vision Alignment Model by Amanpreet Singh, Ronghang Hu, Vedanuj Goswami, Guillaume Couairon, Wojciech Galuba, Marcus Rohrbach, and Douwe Kiela.
  64. FNet (from Google Research) released with the paper FNet: Mixing Tokens with Fourier Transforms by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
  65. Funnel Transformer (from CMU/Google Brain) released with the paper Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing by Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le.
  66. GIT (from Microsoft Research) released with the paper GIT: A Generative Image-to-text Transformer for Vision and Language by Jianfeng Wang, Zhengyuan Yang, Xiaowei Hu, Linjie Li, Kevin Lin, Zhe Gan, Zicheng Liu, Ce Liu, Lijuan Wang.
  67. GLPN (from KAIST) released with the paper Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
  68. GPT (from OpenAI) released with the paper Improving Language Understanding by Generative Pre-Training by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
  69. GPT Neo (from EleutherAI) released in the repository EleutherAI/gpt-neo by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
  70. GPT NeoX (EleutherAI ์—์„œ) Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbac ์˜ GPT-NeoX-20B: An Open-Source Autoregressive Language Model ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  71. GPT NeoX Japanese (from ABEJA) released by Shinya Otani, Takayoshi Makabe, Anuj Arora, and Kyo Hattori.
  72. GPT-2 (OpenAI ์—์„œ) Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever** ์˜ Language Models are Unsupervised Multitask Learners ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  73. GPT-J (from EleutherAI) released in the repository kingoflolz/mesh-transformer-jax by Ben Wang and Aran Komatsuzaki.
  74. GPT-Sw3 (AI-Sweden ์—์„œ) Ariel Ekgren, Amaru Cuba Gyllensten, Evangelia Gogoulou, Alice Heiman, Severine Verlinden, Joey ร–hman, Fredrik Carlsson, Magnus Sahlgren. ์˜ Lessons Learned from GPT-SW3: Building the First Large-Scale Generative Language Model for Swedish ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  75. GPTSAN-japanese released in the repository tanreinama/GPTSAN by Toshiyuki Sakamoto(tanreinama).
  76. Graphormer (from Microsoft) Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, Tie-Yan Liu ์˜ Do Transformers Really Perform Bad for Graph Representation? ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  77. GroupViT (UCSD, NVIDIA ์—์„œ) Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang ์˜ GroupViT: Semantic Segmentation Emerges from Text Supervision ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  78. Hubert (Facebook ์—์„œ) Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed ์˜ HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  79. I-BERT (Berkeley ์—์„œ) Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer ์˜ I-BERT: Integer-only BERT Quantization ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  80. ImageGPT (OpenAI ์—์„œ) Mark Chen, Alec Radford, Rewon Child, Jeffrey Wu, Heewoo Jun, David Luan, Ilya Sutskever ์˜ Generative Pretraining from Pixels ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  81. Informer (from Beihang University, UC Berkeley, Rutgers University, SEDD Company) released with the paper Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting by Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wancai Zhang.
  82. Jukebox (OpenAI ์—์„œ) Prafulla Dhariwal, Heewoo Jun, Christine Payne, Jong Wook Kim, Alec Radford, Ilya Sutskever ์˜ Jukebox: A Generative Model for Music ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  83. LayoutLM (Microsoft Research Asia ์—์„œ) Yiheng Xu, Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou ์˜ LayoutLM: Pre-training of Text and Layout for Document Image Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  84. LayoutLMv2 (Microsoft Research Asia ์—์„œ) Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou ์˜ LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  85. LayoutLMv3 (Microsoft Research Asia ์—์„œ) Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei ์˜ LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  86. LayoutXLM (Microsoft Research Asia ์—์„œ) Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei ์˜ LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  87. LED (AllenAI ์—์„œ) Iz Beltagy, Matthew E. Peters, Arman Cohan ์˜ Longformer: The Long-Document Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  88. LeViT (Meta AI ์—์„œ) Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervรฉ Jรฉgou, Matthijs Douze ์˜ LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  89. LiLT (South China University of Technology ์—์„œ) Jiapeng Wang, Lianwen Jin, Kai Ding ์˜ LiLT: A Simple yet Effective Language-Independent Layout Transformer for Structured Document Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  90. LLaMA (The FAIR team of Meta AI ์—์„œ ์ œ๊ณต)์€ Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothรฉe Lacroix, Baptiste Roziรจre, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, Guillaume Lample.์˜ LLaMA: Open and Efficient Foundation Language Models๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  91. Longformer (AllenAI ์—์„œ) Iz Beltagy, Matthew E. Peters, Arman Cohan ์˜ Longformer: The Long-Document Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  92. LongT5 (Google AI ์—์„œ) Mandy Guo, Joshua Ainslie, David Uthus, Santiago Ontanon, Jianmo Ni, Yun-Hsuan Sung, Yinfei Yang ์˜ LongT5: Efficient Text-To-Text Transformer for Long Sequences ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  93. LUKE (Studio Ousia ์—์„œ) Ikuya Yamada, Akari Asai, Hiroyuki Shindo, Hideaki Takeda, Yuji Matsumoto ์˜ LUKE: Deep Contextualized Entity Representations with Entity-aware Self-attention ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  94. LXMERT (UNC Chapel Hill ์—์„œ) Hao Tan and Mohit Bansal ์˜ LXMERT: Learning Cross-Modality Encoder Representations from Transformers for Open-Domain Question Answering ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  95. M-CTC-T (Facebook ์—์„œ) Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert ์˜ Pseudo-Labeling For Massively Multilingual Speech Recognition ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  96. M2M100 (Facebook ์—์„œ) Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, Armand Joulin ์˜ Beyond English-Centric Multilingual Machine Translation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  97. MarianMT Machine translation models trained using OPUS data by Jรถrg Tiedemann. The Marian Framework is being developed by the Microsoft Translator Team.
  98. MarkupLM (Microsoft Research Asia ์—์„œ) Junlong Li, Yiheng Xu, Lei Cui, Furu Wei ์˜ MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  99. Mask2Former (FAIR and UIUC ์—์„œ ์ œ๊ณต)์€ Bowen Cheng, Ishan Misra, Alexander G. Schwing, Alexander Kirillov, Rohit Girdhar.์˜ Masked-attention Mask Transformer for Universal Image Segmentation๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  100. MaskFormer (Meta and UIUC ์—์„œ) Bowen Cheng, Alexander G. Schwing, Alexander Kirillov ์˜ Per-Pixel Classification is Not All You Need for Semantic Segmentation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  101. mBART (Facebook ์—์„œ) Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer ์˜ Multilingual Denoising Pre-training for Neural Machine Translation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  102. mBART-50 (Facebook ์—์„œ) Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan ์˜ Multilingual Translation with Extensible Multilingual Pretraining and Finetuning ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  103. MEGA (Facebook ์—์„œ ์ œ๊ณต)์€ Xuezhe Ma, Chunting Zhou, Xiang Kong, Junxian He, Liangke Gui, Graham Neubig, Jonathan May, and Luke Zettlemoyer.์˜ Mega: Moving Average Equipped Gated Attention๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  104. Megatron-BERT (NVIDIA ์—์„œ) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro ์˜ Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  105. Megatron-GPT2 (NVIDIA ์—์„œ) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro ์˜ Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  106. MGP-STR (Alibaba Research ์—์„œ ์ œ๊ณต)์€ Peng Wang, Cheng Da, and Cong Yao.์˜ Multi-Granularity Prediction for Scene Text Recognition๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  107. mLUKE (Studio Ousia ์—์„œ) Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka ์˜ mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  108. MobileBERT (CMU/Google Brain ์—์„œ) Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou ์˜ MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  109. MobileNetV1 (Google Inc. ์—์„œ) Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, Hartwig Adam ์˜ MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  110. MobileNetV2 (Google Inc. ์—์„œ) Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, Liang-Chieh Chen ์˜ MobileNetV2: Inverted Residuals and Linear Bottlenecks ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  111. MobileViT (Apple ์—์„œ) Sachin Mehta and Mohammad Rastegari ์˜ MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  112. MPNet (Microsoft Research ์—์„œ) Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu ์˜ MPNet: Masked and Permuted Pre-training for Language Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  113. MT5 (Google AI ์—์„œ) Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel ์˜ mT5: A massively multilingual pre-trained text-to-text transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  114. MVP (RUC AI Box ์—์„œ) Tianyi Tang, Junyi Li, Wayne Xin Zhao and Ji-Rong Wen ์˜ MVP: Multi-task Supervised Pre-training for Natural Language Generation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  115. NAT (SHI Labs ์—์„œ) Ali Hassani, Steven Walton, Jiachen Li, Shen Li, and Humphrey Shi ์˜ Neighborhood Attention Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  116. Nezha (Huawei Noahโ€™s Ark Lab ์—์„œ) Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu ์˜ NEZHA: Neural Contextualized Representation for Chinese Language Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  117. NLLB (Meta ์—์„œ) the NLLB team ์˜ No Language Left Behind: Scaling Human-Centered Machine Translation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  118. NLLB-MOE (Meta ์—์„œ ์ œ๊ณต)์€ the NLLB team.์˜ No Language Left Behind: Scaling Human-Centered Machine Translation๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  119. Nystrรถmformer (the University of Wisconsin - Madison ์—์„œ) Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh ์˜ Nystrรถmformer: A Nystrรถm-Based Algorithm for Approximating Self-Attention ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  120. OneFormer (SHI Labs ์—์„œ) Jitesh Jain, Jiachen Li, MangTik Chiu, Ali Hassani, Nikita Orlov, Humphrey Shi ์˜ OneFormer: One Transformer to Rule Universal Image Segmentation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  121. OPT (Meta AI ์—์„œ) Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al ์˜ OPT: Open Pre-trained Transformer Language Models ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  122. OWL-ViT (Google AI ์—์„œ) Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby ์˜ Simple Open-Vocabulary Object Detection with Vision Transformers ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  123. Pegasus (Google ์—์„œ) Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu ์˜ PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  124. PEGASUS-X (Google ์—์„œ) Jason Phang, Yao Zhao, Peter J. Liu ์˜ Investigating Efficiently Extending Transformers for Long Input Summarization ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  125. Perceiver IO (Deepmind ์—์„œ) Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hรฉnaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, Joรฃo Carreira ์˜ Perceiver IO: A General Architecture for Structured Inputs & Outputs ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  126. PhoBERT (VinAI Research ์—์„œ) Dat Quoc Nguyen and Anh Tuan Nguyen ์˜ PhoBERT: Pre-trained language models for Vietnamese ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  127. Pix2Struct (Google ์—์„œ ์ œ๊ณต)์€ Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova.์˜ Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  128. PLBart (UCLA NLP ์—์„œ) Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang ์˜ Unified Pre-training for Program Understanding and Generation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  129. PoolFormer (Sea AI Labs ์—์„œ) Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng ์˜ MetaFormer is Actually What You Need for Vision ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  130. ProphetNet (Microsoft Research ์—์„œ) Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou ์˜ ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  131. QDQBert (NVIDIA ์—์„œ) Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius ์˜ Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  132. RAG (Facebook ์—์„œ) Patrick Lewis, Ethan Perez, Aleksandara Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Kรผttler, Mike Lewis, Wen-tau Yih, Tim Rocktรคschel, Sebastian Riedel, Douwe Kiela ์˜ Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  133. REALM (Google Research ์—์„œ) Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat and Ming-Wei Chang ์˜ REALM: Retrieval-Augmented Language Model Pre-Training ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  134. Reformer (Google Research ์—์„œ) Nikita Kitaev, ลukasz Kaiser, Anselm Levskaya ์˜ Reformer: The Efficient Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  135. RegNet (META Research ์—์„œ) Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollรกr ์˜ Designing Network Design Space ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  136. RemBERT (Google Research ์—์„œ) Hyung Won Chung, Thibault Fรฉvry, Henry Tsai, M. Johnson, Sebastian Ruder ์˜ Rethinking embedding coupling in pre-trained language models ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  137. ResNet (Microsoft Research ์—์„œ) Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun ์˜ Deep Residual Learning for Image Recognition ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  138. RoBERTa (Facebook ์—์„œ) Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov ์˜ a Robustly Optimized BERT Pretraining Approach ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  139. RoBERTa-PreLayerNorm (Facebook ์—์„œ) Myle Ott, Sergey Edunov, Alexei Baevski, Angela Fan, Sam Gross, Nathan Ng, David Grangier, Michael Auli ์˜ fairseq: A Fast, Extensible Toolkit for Sequence Modeling ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  140. RoCBert (WeChatAI ์—์„œ) HuiSu, WeiweiShi, XiaoyuShen, XiaoZhou, TuoJi, JiaruiFang, JieZhou ์˜ RoCBert: Robust Chinese Bert with Multimodal Contrastive Pretraining ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  141. RoFormer (ZhuiyiTechnology ์—์„œ) Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu ์˜ a RoFormer: Enhanced Transformer with Rotary Position Embedding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  142. SegFormer (NVIDIA ์—์„œ) Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo ์˜ SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  143. SEW (ASAPP ์—์„œ) Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi ์˜ Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  144. SEW-D (ASAPP ์—์„œ) Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi ์˜ Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  145. SpeechT5 (Microsoft Research ์—์„œ ์ œ๊ณต)์€ Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei.์˜ SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  146. SpeechToTextTransformer (Facebook ์—์„œ) Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino ์˜ fairseq S2T: Fast Speech-to-Text Modeling with fairseq ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  147. SpeechToTextTransformer2 (Facebook ์—์„œ) Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau ์˜ Large-Scale Self- and Semi-Supervised Learning for Speech Translation ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  148. Splinter (Tel Aviv University ์—์„œ) Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy ์˜ Few-Shot Question Answering by Pretraining Span Selection ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  149. SqueezeBERT (Berkeley ์—์„œ) Forrest N. Iandola, Albert E. Shaw, Ravi Krishna, and Kurt W. Keutzer ์˜ SqueezeBERT: What can computer vision teach NLP about efficient neural networks? ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  150. Swin Transformer (Microsoft ์—์„œ) Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, Baining Guo ์˜ Swin Transformer: Hierarchical Vision Transformer using Shifted Windows ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  151. Swin Transformer V2 (Microsoft ์—์„œ) Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo ์˜ Swin Transformer V2: Scaling Up Capacity and Resolution ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  152. Swin2SR (University of Wรผrzburg ์—์„œ) Marcos V. Conde, Ui-Jin Choi, Maxime Burchi, Radu Timofte ์˜ Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  153. SwitchTransformers (Google ์—์„œ) William Fedus, Barret Zoph, Noam Shazeer. ์˜ Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  154. T5 (Google AI ์—์„œ) Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu ์˜ Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  155. T5v1.1 (from Google AI) released in the repository google-research/text-to-text-transfer-transformer by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
  156. Table Transformer (Microsoft Research ์—์„œ) Brandon Smock, Rohith Pesala, Robin Abraham ์˜ PubTables-1M: Towards Comprehensive Table Extraction From Unstructured Documents ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  157. TAPAS (Google AI ์—์„œ) Jonathan Herzig, Paweล‚ Krzysztof Nowak, Thomas Mรผller, Francesco Piccinno and Julian Martin Eisenschlos ์˜ TAPAS: Weakly Supervised Table Parsing via Pre-training ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  158. TAPEX (Microsoft Research ์—์„œ) Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou ์˜ TAPEX: Table Pre-training via Learning a Neural SQL Executor ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  159. Time Series Transformer (from HuggingFace).
  160. TimeSformer (Facebook ์—์„œ) Gedas Bertasius, Heng Wang, Lorenzo Torresani ์˜ Is Space-Time Attention All You Need for Video Understanding? ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  161. Trajectory Transformer (the University of California at Berkeley ์—์„œ) Michael Janner, Qiyang Li, Sergey Levin ์˜ Offline Reinforcement Learning as One Big Sequence Modeling Problem ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  162. Transformer-XL (Google/CMU ์—์„œ) Zihang Dai*, Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov ์˜ Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  163. TrOCR (Microsoft ์—์„œ) Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei ์˜ TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  164. TVLT (from UNC Chapel Hill ์—์„œ) Zineng Tang, Jaemin Cho, Yixin Nie, Mohit Bansal ์˜ TVLT: Textless Vision-Language Transformer ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  165. UL2 (Google Research ์—์„œ) Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzle ์˜ Unifying Language Learning Paradigms ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  166. UniSpeech (Microsoft Research ์—์„œ) Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang ์˜ UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  167. UniSpeechSat (Microsoft Research ์—์„œ) Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu ์˜ UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  168. UPerNet (Peking University ์—์„œ ์ œ๊ณต)์€ Tete Xiao, Yingcheng Liu, Bolei Zhou, Yuning Jiang, Jian Sun.์˜ Unified Perceptual Parsing for Scene Understanding๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  169. VAN (Tsinghua University and Nankai University ์—์„œ) Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu ์˜ Visual Attention Network ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  170. VideoMAE (Multimedia Computing Group, Nanjing University ์—์„œ) Zhan Tong, Yibing Song, Jue Wang, Limin Wang ์˜ VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  171. ViLT (NAVER AI Lab/Kakao Enterprise/Kakao Brain ์—์„œ) Wonjae Kim, Bokyung Son, Ildoo Kim ์˜ ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  172. Vision Transformer (ViT) (Google AI ์—์„œ) Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby ์˜ An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  173. VisualBERT (UCLA NLP ์—์„œ) Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang ์˜ VisualBERT: A Simple and Performant Baseline for Vision and Language ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  174. ViT Hybrid (Google AI ์—์„œ) Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby ์˜ An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  175. ViTMAE (Meta AI ์—์„œ) Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollรกr, Ross Girshick ์˜ Masked Autoencoders Are Scalable Vision Learners ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  176. ViTMSN (Meta AI ์—์„œ) Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas ์˜ Masked Siamese Networks for Label-Efficient Learning ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  177. Wav2Vec2 (Facebook AI ์—์„œ) Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli ์˜ wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  178. Wav2Vec2-Conformer (Facebook AI ์—์„œ) Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino ์˜ FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  179. Wav2Vec2Phoneme (Facebook AI ์—์„œ) Qiantong Xu, Alexei Baevski, Michael Auli ์˜ Simple and Effective Zero-shot Cross-lingual Phoneme Recognition ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  180. WavLM (Microsoft Research ์—์„œ) Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei ์˜ WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  181. Whisper (OpenAI ์—์„œ) Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever ์˜ Robust Speech Recognition via Large-Scale Weak Supervision ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  182. X-CLIP (Microsoft Research ์—์„œ) Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling ์˜ Expanding Language-Image Pretrained Models for General Video Recognition ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  183. X-MOD (Meta AI ์—์„œ ์ œ๊ณต)์€ Jonas Pfeiffer, Naman Goyal, Xi Lin, Xian Li, James Cross, Sebastian Riedel, Mikel Artetxe.์˜ Lifting the Curse of Multilinguality by Pre-training Modular Transformers๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  184. XGLM (Facebook AI ์—์„œ ์ œ๊ณต) Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li ์˜ Few-shot Learning with Multilingual Language Models ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  185. XLM (Facebook ์—์„œ) Guillaume Lample and Alexis Conneau ์˜ Cross-lingual Language Model Pretraining ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  186. XLM-ProphetNet (Microsoft Research ์—์„œ) Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou ์˜ ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  187. XLM-RoBERTa (Facebook AI ์—์„œ) Alexis Conneau*, Kartikay Khandelwal*, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmรกn, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov ์˜ Unsupervised Cross-lingual Representation Learning at Scale ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  188. XLM-RoBERTa-XL (Facebook AI ์—์„œ) Naman Goyal, Jingfei Du, Myle Ott, Giri Anantharaman, Alexis Conneau ์˜ Larger-Scale Transformers for Multilingual Masked Language Modeling ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  189. XLM-V (Meta AI ์—์„œ) Davis Liang, Hila Gonen, Yuning Mao, Rui Hou, Naman Goyal, Marjan Ghazvininejad, Luke Zettlemoyer, Madian Khabsa ์˜ XLM-V: Overcoming the Vocabulary Bottleneck in Multilingual Masked Language Models ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  190. XLNet (Google/CMU ์—์„œ) Zhilin Yang*, Zihang Dai*, Yiming Yang, Jaime Carbonell, Ruslan Salakhutdinov, Quoc V. Le ์˜ โ€‹XLNet: Generalized Autoregressive Pretraining for Language Understanding ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  191. XLS-R (Facebook AI ์—์„œ) Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, Michael Auli ์˜ XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  192. XLSR-Wav2Vec2 (Facebook AI ์—์„œ) Alexis Conneau, Alexei Baevski, Ronan Collobert, Abdelrahman Mohamed, Michael Auli ์˜ Unsupervised Cross-Lingual Representation Learning For Speech Recognition ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  193. YOLOS (Huazhong University of Science & Technology ์—์„œ) Yuxin Fang, Bencheng Liao, Xinggang Wang, Jiemin Fang, Jiyang Qi, Rui Wu, Jianwei Niu, Wenyu Liu ์˜ You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  194. YOSO (the University of Wisconsin - Madison ์—์„œ) Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh ์˜ [You Only Sample (Almost) ๋…ผ๋ฌธ๊ณผ ํ•จ๊ป˜ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
  195. ์ƒˆ๋กœ์šด ๋ชจ๋ธ์„ ์˜ฌ๋ฆฌ๊ณ  ์‹ถ๋‚˜์š”? ์šฐ๋ฆฌ๊ฐ€ ์ƒ์„ธํ•œ ๊ฐ€์ด๋“œ์™€ ํ…œํ”Œ๋ฆฟ ์œผ๋กœ ์ƒˆ๋กœ์šด ๋ชจ๋ธ์„ ์˜ฌ๋ฆฌ๋„๋ก ๋„์™€๋“œ๋ฆด๊ฒŒ์š”. ๊ฐ€์ด๋“œ์™€ ํ…œํ”Œ๋ฆฟ์€ ์ด ์ €์žฅ์†Œ์˜ templates ํด๋”์—์„œ ํ™•์ธํ•˜์‹ค ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ปจํŠธ๋ฆฌ๋ทฐ์…˜ ๊ฐ€์ด๋“œ๋ผ์ธ์„ ๊ผญ ํ™•์ธํ•ด์ฃผ์‹œ๊ณ , PR์„ ์˜ฌ๋ฆฌ๊ธฐ ์ „์— ๋ฉ”์ธํ…Œ์ด๋„ˆ์—๊ฒŒ ์—ฐ๋ฝํ•˜๊ฑฐ๋‚˜ ์ด์Šˆ๋ฅผ ์˜คํ”ˆํ•ด ํ”ผ๋“œ๋ฐฑ์„ ๋ฐ›์œผ์‹œ๊ธธ ๋ฐ”๋ž๋‹ˆ๋‹ค.

๊ฐ ๋ชจ๋ธ์ด Flax, PyTorch, TensorFlow์œผ๋กœ ๊ตฌํ˜„๋˜์—ˆ๋Š”์ง€ ๋˜๋Š” ๐Ÿค— Tokenizers ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๊ฐ€ ์ง€์›ํ•˜๋Š” ํ† ํฌ๋‚˜์ด์ €๋ฅผ ์‚ฌ์šฉํ•˜๋Š”์ง€ ํ™•์ธํ•˜๋ ค๋ฉด, ์ด ํ‘œ๋ฅผ ํ™•์ธํ•˜์„ธ์š”.

์ด ๊ตฌํ˜„์€ ์—ฌ๋Ÿฌ ๋ฐ์ดํ„ฐ๋กœ ๊ฒ€์ฆ๋˜์—ˆ๊ณ  (์˜ˆ์‹œ ์Šคํฌ๋ฆฝํŠธ๋ฅผ ์ฐธ๊ณ ํ•˜์„ธ์š”) ์˜ค๋ฆฌ์ง€๋„ ๊ตฌํ˜„์˜ ์„ฑ๋Šฅ๊ณผ ๊ฐ™์•„์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๋„ํ๋จผํŠธ์˜ Examples ์„น์…˜์—์„œ ์„ฑ๋Šฅ์— ๋Œ€ํ•œ ์ž์„ธํ•œ ์„ค๋ช…์„ ํ™•์ธํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

๋” ์•Œ์•„๋ณด๊ธฐ

์„น์…˜ ์„ค๋ช…
๋„ํ๋จผํŠธ ์ „์ฒด API ๋„ํ๋จผํŠธ์™€ ํŠœํ† ๋ฆฌ์–ผ
๊ณผ์ œ ์š”์•ฝ ๐Ÿค— Transformers๊ฐ€ ์ง€์›ํ•˜๋Š” ๊ณผ์ œ๋“ค
์ „์ฒ˜๋ฆฌ ํŠœํ† ๋ฆฌ์–ผ Tokenizer ํด๋ž˜์Šค๋ฅผ ์ด์šฉํ•ด ๋ชจ๋ธ์„ ์œ„ํ•œ ๋ฐ์ดํ„ฐ ์ค€๋น„ํ•˜๊ธฐ
ํ•™์Šต๊ณผ fine-tuning ๐Ÿค— Transformers๊ฐ€ ์ œ๊ณตํ•˜๋Š” ๋ชจ๋ธ PyTorch/TensorFlow ํ•™์Šต ๊ณผ์ •๊ณผ Trainer API์—์„œ ์‚ฌ์šฉํ•˜๊ธฐ
ํ€ต ํˆฌ์–ด: Fine-tuning/์‚ฌ์šฉ ์Šคํฌ๋ฆฝํŠธ ๋‹ค์–‘ํ•œ ๊ณผ์ œ์—์„œ ๋ชจ๋ธ fine-tuningํ•˜๋Š” ์˜ˆ์‹œ ์Šคํฌ๋ฆฝํŠธ
๋ชจ๋ธ ๊ณต์œ  ๋ฐ ์—…๋กœ๋“œ ์ปค๋ฎค๋‹ˆํ‹ฐ์— fine-tune๋œ ๋ชจ๋ธ์„ ์—…๋กœ๋“œ ๋ฐ ๊ณต์œ ํ•˜๊ธฐ
๋งˆ์ด๊ทธ๋ ˆ์ด์…˜ pytorch-transformers๋‚˜ pytorch-pretrained-bert์—์„œ ๐Ÿค— Transformers๋กœ ์ด๋™ํ•˜๊ธฐ

์ธ์šฉ

๐Ÿค— Transformers ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ์ธ์šฉํ•˜๊ณ  ์‹ถ๋‹ค๋ฉด, ์ด ๋…ผ๋ฌธ์„ ์ธ์šฉํ•ด ์ฃผ์„ธ์š”:

@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 Rรฉmi 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"
}