Spaces:
Runtime error
A newer version of the Gradio SDK is available:
5.5.0
English | ็ฎไฝไธญๆ | ็น้ซไธญๆ | ํ๊ตญ์ด | Espaรฑol | ๆฅๆฌ่ช | เคนเคฟเคจเฅเคฆเฅ
Jax, Pytorch, TensorFlow๋ฅผ ์ํ ์ต์ฒจ๋จ ์์ฐ์ด์ฒ๋ฆฌ
๐ค Transformers๋ ๋ถ๋ฅ, ์ ๋ณด ์ถ์ถ, ์ง๋ฌธ ๋ต๋ณ, ์์ฝ, ๋ฒ์ญ, ๋ฌธ์ฅ ์์ฑ ๋ฑ์ 100๊ฐ ์ด์์ ์ธ์ด๋ก ์ํํ ์ ์๋ ์์ฒ๊ฐ์ ์ฌ์ ํ์ต๋ ๋ชจ๋ธ์ ์ ๊ณตํฉ๋๋ค. ์ฐ๋ฆฌ์ ๋ชฉํ๋ ๋ชจ๋๊ฐ ์ต์ฒจ๋จ์ NLP ๊ธฐ์ ์ ์ฝ๊ฒ ์ฌ์ฉํ๋ ๊ฒ์ ๋๋ค.
๐ค Transformers๋ ์ด๋ฌํ ์ฌ์ ํ์ต ๋ชจ๋ธ์ ๋น ๋ฅด๊ฒ ๋ค์ด๋ก๋ํด ํน์ ํ ์คํธ์ ์ฌ์ฉํ๊ณ , ์ํ๋ ๋ฐ์ดํฐ๋ก fine-tuningํด ์ปค๋ฎค๋ํฐ๋ ์ฐ๋ฆฌ์ ๋ชจ๋ธ ํ๋ธ์ ๊ณต์ ํ ์ ์๋๋ก API๋ฅผ ์ ๊ณตํฉ๋๋ค. ๋ํ, ๋ชจ๋ธ ๊ตฌ์กฐ๋ฅผ ์ ์ํ๋ ๊ฐ ํ์ด์ฌ ๋ชจ๋์ ์์ ํ ๋ ๋ฆฝ์ ์ด์ฌ์ ์ฐ๊ตฌ ์คํ์ ์ํด ์์ฝ๊ฒ ์์ ํ ์ ์์ต๋๋ค.
๐ค Transformers๋ ๊ฐ์ฅ ์ ๋ช ํ 3๊ฐ์ ๋ฅ๋ฌ๋ ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ์ง์ํฉ๋๋ค. ์ด๋ค์ ์๋ก ์๋ฒฝํ ์ฐ๋๋ฉ๋๋ค โ Jax, PyTorch, TensorFlow. ๊ฐ๋จํ๊ฒ ์ด ๋ผ์ด๋ธ๋ฌ๋ฆฌ ์ค ํ๋๋ก ๋ชจ๋ธ์ ํ์ตํ๊ณ , ๋ ๋ค๋ฅธ ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ก ์ถ๋ก ์ ์ํด ๋ชจ๋ธ์ ๋ถ๋ฌ์ฌ ์ ์์ต๋๋ค.
์จ๋ผ์ธ ๋ฐ๋ชจ
๋๋ถ๋ถ์ ๋ชจ๋ธ์ ๋ชจ๋ธ ํ๋ธ ํ์ด์ง์์ ๋ฐ๋ก ํ ์คํธํด๋ณผ ์ ์์ต๋๋ค. ๊ณต๊ฐ ๋ฐ ๋น๊ณต๊ฐ ๋ชจ๋ธ์ ์ํ ๋น๊ณต๊ฐ ๋ชจ๋ธ ํธ์คํ , ๋ฒ์ ๊ด๋ฆฌ, ์ถ๋ก API๋ ์ ๊ณตํฉ๋๋ค.
์์:
- BERT๋ก ๋ง์คํน๋ ๋จ์ด ์์ฑํ๊ธฐ
- Electra๋ฅผ ์ด์ฉํ ๊ฐ์ฒด๋ช ์ธ์
- GPT-2๋ก ํ ์คํธ ์์ฑํ๊ธฐ
- RoBERTa๋ก ์์ฐ์ด ์ถ๋ก ํ๊ธฐ
- BART๋ฅผ ์ด์ฉํ ์์ฝ
- DistilBERT๋ฅผ ์ด์ฉํ ์ง๋ฌธ ๋ต๋ณ
- T5๋ก ๋ฒ์ญํ๊ธฐ
Transformer์ ๊ธ์ฐ๊ธฐ ๋ ์ด ์ ์ฅ์์ ํ ์คํธ ์์ฑ ๋ฅ๋ ฅ์ ๊ดํ Hugging Face ํ์ ๊ณต์ ๋ฐ๋ชจ์ ๋๋ค.
Hugging Face ํ์ ์ปค์คํ ์ง์์ ์ํ๋ค๋ฉด
ํต ํฌ์ด
์ํ๋ ํ
์คํธ์ ๋ฐ๋ก ๋ชจ๋ธ์ ์ฌ์ฉํ ์ ์๋๋ก, ์ฐ๋ฆฌ๋ pipeline
API๋ฅผ ์ ๊ณตํฉ๋๋ค. Pipeline์ ์ฌ์ ํ์ต ๋ชจ๋ธ๊ณผ ๊ทธ ๋ชจ๋ธ์ ํ์ตํ ๋ ์ ์ฉํ ์ ์ฒ๋ฆฌ ๋ฐฉ์์ ํ๋๋ก ํฉ์นฉ๋๋ค. ๋ค์์ ๊ธ์ ์ ์ธ ํ
์คํธ์ ๋ถ์ ์ ์ธ ํ
์คํธ๋ฅผ ๋ถ๋ฅํ๊ธฐ ์ํด pipeline์ ์ฌ์ฉํ ๊ฐ๋จํ ์์์
๋๋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}]
์ฝ๋์ ๋๋ฒ์งธ ์ค์ pipeline์ด ์ฌ์ฉํ๋ ์ฌ์ ํ์ต ๋ชจ๋ธ์ ๋ค์ด๋ก๋ํ๊ณ ์บ์๋ก ์ ์ฅํฉ๋๋ค. ์ธ๋ฒ์งธ ์ค์์ ๊ทธ ๋ชจ๋ธ์ด ์ฃผ์ด์ง ํ ์คํธ๋ฅผ ํ๊ฐํฉ๋๋ค. ์ฌ๊ธฐ์ ๋ชจ๋ธ์ 99.97%์ ํ๋ฅ ๋ก ํ ์คํธ๊ฐ ๊ธ์ ์ ์ด๋ผ๊ณ ํ๊ฐํ์ต๋๋ค.
๋ง์ NLP ๊ณผ์ ๋ค์ pipeline
์ผ๋ก ๋ฐ๋ก ์ํํ ์ ์์ต๋๋ค. ์๋ฅผ ๋ค์ด, ์ง๋ฌธ๊ณผ ๋ฌธ๋งฅ์ด ์ฃผ์ด์ง๋ฉด ์์ฝ๊ฒ ๋ต๋ณ์ ์ถ์ถํ ์ ์์ต๋๋ค:
>>> 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'}
๋ต๋ณ๋ฟ๋ง ์๋๋ผ, ์ฌ๊ธฐ์ ์ฌ์ฉ๋ ์ฌ์ ํ์ต ๋ชจ๋ธ์ ํ์ ๋์ ํ ํฌ๋์ด์ฆ๋ ๋ฌธ์ฅ ์ ๋ต๋ณ์ ์์์ , ๋์ ๊น์ง ๋ฐํํฉ๋๋ค. ์ด ํํ ๋ฆฌ์ผ์์ pipeline
API๊ฐ ์ง์ํ๋ ๋ค์ํ ๊ณผ์ ๋ฅผ ํ์ธํ ์ ์์ต๋๋ค.
์ฝ๋ 3์ค๋ก ์ํ๋ ๊ณผ์ ์ ๋ง๊ฒ ์ฌ์ ํ์ต ๋ชจ๋ธ์ ๋ค์ด๋ก๋ ๋ฐ๊ณ ์ฌ์ฉํ ์ ์์ต๋๋ค. ๋ค์์ PyTorch ๋ฒ์ ์ ๋๋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)
๋ค์์ TensorFlow ๋ฒ์ ์ ๋๋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)
ํ ํฌ๋์ด์ ๋ ์ฌ์ ํ์ต ๋ชจ๋ธ์ ๋ชจ๋ ์ ์ฒ๋ฆฌ๋ฅผ ์ฑ ์์ง๋๋ค. ๊ทธ๋ฆฌ๊ณ (์์ ์์์ฒ๋ผ) 1๊ฐ์ ์คํธ๋ง์ด๋ ๋ฆฌ์คํธ๋ ์ฒ๋ฆฌํ ์ ์์ต๋๋ค. ํ ํฌ๋์ด์ ๋ ๋์ ๋๋ฆฌ๋ฅผ ๋ฐํํ๋๋ฐ, ์ด๋ ๋ค์ด์คํธ๋ฆผ ์ฝ๋์ ์ฌ์ฉํ๊ฑฐ๋ ์ธํจํน ์ฐ์ฐ์ ** ๋ฅผ ์ด์ฉํด ๋ชจ๋ธ์ ๋ฐ๋ก ์ ๋ฌํ ์๋ ์์ต๋๋ค.
๋ชจ๋ธ ์์ฒด๋ ์ผ๋ฐ์ ์ผ๋ก ์ฌ์ฉ๋๋ Pytorch nn.Module
๋ TensorFlow tf.keras.Model
์
๋๋ค. ์ด ํํ ๋ฆฌ์ผ์ ์ด๋ฌํ ๋ชจ๋ธ์ ํ์ค์ ์ธ PyTorch๋ TensorFlow ํ์ต ๊ณผ์ ์์ ์ฌ์ฉํ๋ ๋ฐฉ๋ฒ, ๋๋ ์๋ก์ด ๋ฐ์ดํฐ๋ก fine-tuneํ๊ธฐ ์ํด Trainer
API๋ฅผ ์ฌ์ฉํ๋ ๋ฐฉ๋ฒ์ ์ค๋ช
ํด์ค๋๋ค.
์ transformers๋ฅผ ์ฌ์ฉํด์ผ ํ ๊น์?
์์ฝ๊ฒ ์ฌ์ฉํ ์ ์๋ ์ต์ฒจ๋จ ๋ชจ๋ธ:
- NLU์ NLG ๊ณผ์ ์์ ๋ฐ์ด๋ ์ฑ๋ฅ์ ๋ณด์ ๋๋ค.
- ๊ต์ก์ ์ค๋ฌด์์๊ฒ ์ง์ ์ฅ๋ฒฝ์ด ๋ฎ์ต๋๋ค.
- 3๊ฐ์ ํด๋์ค๋ง ๋ฐฐ์ฐ๋ฉด ๋ฐ๋ก ์ฌ์ฉํ ์ ์์ต๋๋ค.
- ํ๋์ API๋ก ๋ชจ๋ ์ฌ์ ํ์ต ๋ชจ๋ธ์ ์ฌ์ฉํ ์ ์์ต๋๋ค.
๋ ์ ์ ๊ณ์ฐ ๋น์ฉ, ๋ ์ ์ ํ์ ๋ฐ์๊ตญ:
- ์ฐ๊ตฌ์๋ค์ ๋ชจ๋ธ์ ๊ณ์ ๋ค์ ํ์ต์ํค๋ ๋์ ํ์ต๋ ๋ชจ๋ธ์ ๊ณต์ ํ ์ ์์ต๋๋ค.
- ์ค๋ฌด์๋ค์ ํ์ต์ ํ์ํ ์๊ฐ๊ณผ ๋น์ฉ์ ์ ์ฝํ ์ ์์ต๋๋ค.
- ์์ญ๊ฐ์ ๋ชจ๋ธ ๊ตฌ์กฐ, 2,000๊ฐ ์ด์์ ์ฌ์ ํ์ต ๋ชจ๋ธ, 100๊ฐ ์ด์์ ์ธ์ด๋ก ํ์ต๋ ๋ชจ๋ธ ๋ฑ.
๋ชจ๋ธ์ ๊ฐ ์์ ์ฃผ๊ธฐ์ ์ ํฉํ ํ๋ ์์ํฌ:
- ์ฝ๋ 3์ค๋ก ์ต์ฒจ๋จ ๋ชจ๋ธ์ ํ์ตํ์ธ์.
- ์์ ๋กญ๊ฒ ๋ชจ๋ธ์ TF2.0๋ PyTorch ํ๋ ์์ํฌ๋ก ๋ณํํ์ธ์.
- ํ์ต, ํ๊ฐ, ๊ณต๊ฐ ๋ฑ ๊ฐ ๋จ๊ณ์ ๋ง๋ ํ๋ ์์ํฌ๋ฅผ ์ํ๋๋๋ก ์ ํํ์ธ์.
ํ์ํ ๋๋ก ๋ชจ๋ธ์ด๋ ์์๋ฅผ ์ปค์คํฐ๋ง์ด์ฆํ์ธ์:
- ์ฐ๋ฆฌ๋ ์ ์๊ฐ ๊ณต๊ฐํ ๊ฒฐ๊ณผ๋ฅผ ์ฌํํ๊ธฐ ์ํด ๊ฐ ๋ชจ๋ธ ๊ตฌ์กฐ์ ์์๋ฅผ ์ ๊ณตํฉ๋๋ค.
- ๋ชจ๋ธ ๋ด๋ถ ๊ตฌ์กฐ๋ ๊ฐ๋ฅํ ์ผ๊ด์ ์ผ๋ก ๊ณต๊ฐ๋์ด ์์ต๋๋ค.
- ๋น ๋ฅธ ์คํ์ ์ํด ๋ชจ๋ธ ํ์ผ์ ๋ผ์ด๋ธ๋ฌ๋ฆฌ์ ๋ ๋ฆฝ์ ์ผ๋ก ์ฌ์ฉ๋ ์ ์์ต๋๋ค.
์ transformers๋ฅผ ์ฌ์ฉํ์ง ๋ง์์ผ ํ ๊น์?
- ์ด ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ ์ ๊ฒฝ๋ง ๋ธ๋ก์ ๋ง๋ค๊ธฐ ์ํ ๋ชจ๋์ด ์๋๋๋ค. ์ฐ๊ตฌ์๋ค์ด ์ฌ๋ฌ ํ์ผ์ ์ดํด๋ณด์ง ์๊ณ ๋ฐ๋ก ๊ฐ ๋ชจ๋ธ์ ์ฌ์ฉํ ์ ์๋๋ก, ๋ชจ๋ธ ํ์ผ ์ฝ๋์ ์ถ์ํ ์์ค์ ์ ์ ํ๊ฒ ์ ์งํ์ต๋๋ค.
- ํ์ต API๋ ๋ชจ๋ ๋ชจ๋ธ์ ์ ์ฉํ ์ ์๋๋ก ๋ง๋ค์ด์ง์ง ์์์ง๋ง, ๋ผ์ด๋ธ๋ฌ๋ฆฌ๊ฐ ์ ๊ณตํ๋ ๋ชจ๋ธ๋ค์ ์ ์ฉํ ์ ์๋๋ก ์ต์ ํ๋์์ต๋๋ค. ์ผ๋ฐ์ ์ธ ๋จธ์ ๋ฌ๋์ ์ํด์ , ๋ค๋ฅธ ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ์ฌ์ฉํ์ธ์.
- ๊ฐ๋ฅํ ๋ง์ ์ฌ์ฉ ์์๋ฅผ ๋ณด์ฌ๋๋ฆฌ๊ณ ์ถ์ด์, ์์ ํด๋์ ์คํฌ๋ฆฝํธ๋ฅผ ์ค๋นํ์ต๋๋ค. ์ด ์คํฌ๋ฆฝํธ๋ค์ ์์ ์์ด ํน์ ํ ๋ฌธ์ ์ ๋ฐ๋ก ์ ์ฉํ์ง ๋ชปํ ์ ์์ต๋๋ค. ํ์์ ๋ง๊ฒ ์ผ๋ถ ์ฝ๋๋ฅผ ์์ ํด์ผ ํ ์ ์์ต๋๋ค.
์ค์น
pip๋ก ์ค์นํ๊ธฐ
์ด ์ ์ฅ์๋ Python 3.6+, Flax 0.3.2+, PyTorch 1.3.1+, TensorFlow 2.3+์์ ํ ์คํธ ๋์์ต๋๋ค.
๊ฐ์ ํ๊ฒฝ์ ๐ค Transformers๋ฅผ ์ค์นํ์ธ์. Python ๊ฐ์ ํ๊ฒฝ์ ์ต์ํ์ง ์๋ค๋ฉด, ์ฌ์ฉ์ ๊ฐ์ด๋๋ฅผ ํ์ธํ์ธ์.
์ฐ์ , ์ฌ์ฉํ Python ๋ฒ์ ์ผ๋ก ๊ฐ์ ํ๊ฒฝ์ ๋ง๋ค๊ณ ์คํํ์ธ์.
๊ทธ ๋ค์, Flax, PyTorch, TensorFlow ์ค ์ ์ด๋ ํ๋๋ ์ค์นํด์ผ ํฉ๋๋ค. ํ๋ซํผ์ ๋ง๋ ์ค์น ๋ช ๋ น์ด๋ฅผ ํ์ธํ๊ธฐ ์ํด TensorFlow ์ค์น ํ์ด์ง, PyTorch ์ค์น ํ์ด์ง, Flax ์ค์น ํ์ด์ง๋ฅผ ํ์ธํ์ธ์.
์ด๋ค ์ค ์ ์ด๋ ํ๋๊ฐ ์ค์น๋์๋ค๋ฉด, ๐ค Transformers๋ ๋ค์๊ณผ ๊ฐ์ด pip์ ์ด์ฉํด ์ค์นํ ์ ์์ต๋๋ค:
pip install transformers
์์๋ค์ ์ฒดํํด๋ณด๊ณ ์ถ๊ฑฐ๋, ์ต์ต์ต์ฒจ๋จ ์ฝ๋๋ฅผ ์ํ๊ฑฐ๋, ์๋ก์ด ๋ฒ์ ์ด ๋์ฌ ๋๊น์ง ๊ธฐ๋ค๋ฆด ์ ์๋ค๋ฉด ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ์์ค์์ ๋ฐ๋ก ์ค์นํ์ ์ผ ํฉ๋๋ค.
conda๋ก ์ค์นํ๊ธฐ
Transformers ๋ฒ์ v4.0.0๋ถํฐ, conda ์ฑ๋์ด ์๊ฒผ์ต๋๋ค: huggingface
.
๐ค Transformers๋ ๋ค์๊ณผ ๊ฐ์ด conda๋ก ์ค์นํ ์ ์์ต๋๋ค:
conda install -c huggingface transformers
Flax, PyTorch, TensorFlow ์ค์น ํ์ด์ง์์ ์ด๋ค์ conda๋ก ์ค์นํ๋ ๋ฐฉ๋ฒ์ ํ์ธํ์ธ์.
๋ชจ๋ธ ๊ตฌ์กฐ
๐ค Transformers๊ฐ ์ ๊ณตํ๋ ๋ชจ๋ ๋ชจ๋ธ ์ฒดํฌํฌ์ธํธ ๋ huggingface.co ๋ชจ๋ธ ํ๋ธ์ ์๋ฒฝํ ์ฐ๋๋์ด ์์ต๋๋ค. ๊ฐ์ธ๊ณผ ๊ธฐ๊ด์ด ๋ชจ๋ธ ํ๋ธ์ ์ง์ ์ ๋ก๋ํ ์ ์์ต๋๋ค.
ํ์ฌ ์ฌ์ฉ ๊ฐ๋ฅํ ๋ชจ๋ธ ์ฒดํฌํฌ์ธํธ์ ๊ฐ์:
๐ค Transformers๋ ๋ค์ ๋ชจ๋ธ๋ค์ ์ ๊ณตํฉ๋๋ค (๊ฐ ๋ชจ๋ธ์ ์์ฝ์ ์ฌ๊ธฐ์ ํ์ธํ์ธ์):
- 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.
- ALIGN (Google Research ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- 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.
- Audio Spectrogram Transformer (from MIT) released with the paper AST: Audio Spectrogram Transformer by Yuan Gong, Yu-An Chung, James Glass.
- 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.
- BARThez (from รcole polytechnique) released with the paper BARThez: a Skilled Pretrained French Sequence-to-Sequence Model by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
- 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.
- BEiT (from Microsoft) released with the paper BEiT: BERT Pre-Training of Image Transformers by Hangbo Bao, Li Dong, Furu Wei.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- 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.
- BLIP-2 (Salesforce ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- BLOOM (from BigScience workshop) released by the BigScience Workshop.
- BORT (Alexa ์์) Adrian de Wynter and Daniel J. Perry ์ Optimal Subarchitecture Extraction For BERT ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- 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.
- ByT5 (Google Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- CamemBERT (Inria/Facebook/Sorbonne ์์) Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suรกrez*, Yoann Dupont, Laurent Romary, รric Villemonte de la Clergerie, Djamรฉ Seddah and Benoรฎt Sagot ์ CamemBERT: a Tasty French Language Model ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- CANINE (Google Research ์์) Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting ์ CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Chinese-CLIP (OFA-Sys ์์) An Yang, Junshu Pan, Junyang Lin, Rui Men, Yichang Zhang, Jingren Zhou, Chang Zhou ์ Chinese CLIP: Contrastive Vision-Language Pretraining in Chinese ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- CLAP (LAION-AI ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- CLIP (OpenAI ์์) 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 ๋ฐํํ์ต๋๋ค.
- CLIPSeg (University of Gรถttingen ์์) Timo Lรผddecke and Alexander Ecker ์ Image Segmentation Using Text and Image Prompts ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- CodeGen (Salesforce ์์) Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong ์ A Conversational Paradigm for Program Synthesis ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Conditional DETR (Microsoft Research Asia ์์) Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang ์ Conditional DETR for Fast Training Convergence ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ConvBERT (YituTech ์์) Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan ์ ConvBERT: Improving BERT with Span-based Dynamic Convolution ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ConvNeXT (Facebook AI ์์) Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie ์ A ConvNet for the 2020s ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- 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.
- CPM (Tsinghua University ์์) 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 ๋ฐํํ์ต๋๋ค.
- CPM-Ant (from OpenBMB) released by the OpenBMB.
- CTRL (Salesforce ์์) Nitish Shirish Keskar*, Bryan McCann*, Lav R. Varshney, Caiming Xiong and Richard Socher ์ CTRL: A Conditional Transformer Language Model for Controllable Generation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- CvT (Microsoft ์์) Haiping Wu, Bin Xiao, Noel Codella, Mengchen Liu, Xiyang Dai, Lu Yuan, Lei Zhang ์ CvT: Introducing Convolutions to Vision Transformers ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Data2Vec (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- DeBERTa (Microsoft ์์) Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen ์ DeBERTa: Decoding-enhanced BERT with Disentangled Attention ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DeBERTa-v2 (Microsoft ์์) Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen ์ DeBERTa: Decoding-enhanced BERT with Disentangled Attention ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Decision Transformer (Berkeley/Facebook/Google ์์) 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 ๋ฐํํ์ต๋๋ค.
- Deformable DETR (SenseTime Research ์์) Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai ์ Deformable DETR: Deformable Transformers for End-to-End Object Detection ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DeiT (Facebook ์์) Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervรฉ Jรฉgou ์ Training data-efficient image transformers & distillation through attention ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DePlot (Google AI ์์ ์ ๊ณต)์ Fangyu Liu, Julian Martin Eisenschlos, Francesco Piccinno, Syrine Krichene, Chenxi Pang, Kenton Lee, Mandar Joshi, Wenhu Chen, Nigel Collier, Yasemin Altun.์ DePlot: One-shot visual language reasoning by plot-to-table translation๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DETA (The University of Texas at Austin ์์ ์ ๊ณต)์ Jeffrey Ouyang-Zhang, Jang Hyun Cho, Xingyi Zhou, Philipp Krรคhenbรผhl.์ NMS Strikes Back๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DETR (Facebook ์์) Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko ์ End-to-End Object Detection with Transformers ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DialoGPT (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- DiNAT (SHI Labs ์์) Ali Hassani and Humphrey Shi ์ Dilated Neighborhood Attention Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DistilBERT (HuggingFace ์์) 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 ๋ฐํํ์ต๋๋ค.
- DiT (Microsoft Research ์์) Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei ์ DiT: Self-supervised Pre-training for Document Image Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Donut (NAVER ์์) 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 ๋ฐํํ์ต๋๋ค.
- DPR (Facebook ์์) Vladimir Karpukhin, Barlas Oฤuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih ์ Dense Passage Retrieval for Open-Domain Question Answering ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- DPT (Intel Labs ์์) Renรฉ Ranftl, Alexey Bochkovskiy, Vladlen Koltun ์ Vision Transformers for Dense Prediction ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- 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.
- EfficientNet (from Google Brain) released with the paper EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks by Mingxing Tan, Quoc V. Le.
- ELECTRA (Google Research/Stanford University ์์) Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning ์ ELECTRA: Pre-training text encoders as discriminators rather than generators ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- EncoderDecoder (Google Research ์์) Sascha Rothe, Shashi Narayan, Aliaksei Severyn ์ Leveraging Pre-trained Checkpoints for Sequence Generation Tasks ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ERNIE (Baidu ์์) 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 ๋ฐํํ์ต๋๋ค.
- ErnieM (Baidu ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- 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.
- 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
- 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
- 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.
- 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.
- FNet (from Google Research) released with the paper FNet: Mixing Tokens with Fourier Transforms by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
- 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.
- 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.
- 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.
- GPT (from OpenAI) released with the paper Improving Language Understanding by Generative Pre-Training by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
- GPT Neo (from EleutherAI) released in the repository EleutherAI/gpt-neo by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
- GPT NeoX (EleutherAI ์์) 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 ๋ฐํํ์ต๋๋ค.
- GPT NeoX Japanese (from ABEJA) released by Shinya Otani, Takayoshi Makabe, Anuj Arora, and Kyo Hattori.
- GPT-2 (OpenAI ์์) Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever** ์ Language Models are Unsupervised Multitask Learners ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- GPT-J (from EleutherAI) released in the repository kingoflolz/mesh-transformer-jax by Ben Wang and Aran Komatsuzaki.
- GPT-Sw3 (AI-Sweden ์์) Ariel Ekgren, Amaru Cuba Gyllensten, Evangelia Gogoulou, Alice Heiman, Severine Verlinden, Joey รhman, Fredrik Carlsson, Magnus Sahlgren. ์ Lessons Learned from GPT-SW3: Building the First Large-Scale Generative Language Model for Swedish ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- GPTBigCode (BigCode ์์ ์ ๊ณต)์ Loubna Ben Allal, Raymond Li, Denis Kocetkov, Chenghao Mou, Christopher Akiki, Carlos Munoz Ferrandis, Niklas Muennighoff, Mayank Mishra, Alex Gu, Manan Dey, Logesh Kumar Umapathi, Carolyn Jane Anderson, Yangtian Zi, Joel Lamy Poirier, Hailey Schoelkopf, Sergey Troshin, Dmitry Abulkhanov, Manuel Romero, Michael Lappert, Francesco De Toni, Bernardo Garcรญa del Rรญo, Qian Liu, Shamik Bose, Urvashi Bhattacharyya, Terry Yue Zhuo, Ian Yu, Paulo Villegas, Marco Zocca, Sourab Mangrulkar, David Lansky, Huu Nguyen, Danish Contractor, Luis Villa, Jia Li, Dzmitry Bahdanau, Yacine Jernite, Sean Hughes, Daniel Fried, Arjun Guha, Harm de Vries, Leandro von Werra.์ SantaCoder: don't reach for the stars!๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- GPTSAN-japanese released in the repository tanreinama/GPTSAN by Toshiyuki Sakamoto(tanreinama).
- 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 ๋ฐํํ์ต๋๋ค.
- GroupViT (UCSD, NVIDIA ์์) Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang ์ GroupViT: Semantic Segmentation Emerges from Text Supervision ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Hubert (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- I-BERT (Berkeley ์์) Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer ์ I-BERT: Integer-only BERT Quantization ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ImageGPT (OpenAI ์์) Mark Chen, Alec Radford, Rewon Child, Jeffrey Wu, Heewoo Jun, David Luan, Ilya Sutskever ์ Generative Pretraining from Pixels ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- 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.
- Jukebox (OpenAI ์์) Prafulla Dhariwal, Heewoo Jun, Christine Payne, Jong Wook Kim, Alec Radford, Ilya Sutskever ์ Jukebox: A Generative Model for Music ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- LayoutLM (Microsoft Research Asia ์์) 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 ๋ฐํํ์ต๋๋ค.
- LayoutLMv2 (Microsoft Research Asia ์์) 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 ๋ฐํํ์ต๋๋ค.
- LayoutLMv3 (Microsoft Research Asia ์์) Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei ์ LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- LayoutXLM (Microsoft Research Asia ์์) 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 ๋ฐํํ์ต๋๋ค.
- LED (AllenAI ์์) Iz Beltagy, Matthew E. Peters, Arman Cohan ์ Longformer: The Long-Document Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- LeViT (Meta AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- LiLT (South China University of Technology ์์) Jiapeng Wang, Lianwen Jin, Kai Ding ์ LiLT: A Simple yet Effective Language-Independent Layout Transformer for Structured Document Understanding ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- LLaMA (The FAIR team of Meta AI ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- Longformer (AllenAI ์์) Iz Beltagy, Matthew E. Peters, Arman Cohan ์ Longformer: The Long-Document Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- LongT5 (Google AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- LUKE (Studio Ousia ์์) Ikuya Yamada, Akari Asai, Hiroyuki Shindo, Hideaki Takeda, Yuji Matsumoto ์ LUKE: Deep Contextualized Entity Representations with Entity-aware Self-attention ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- LXMERT (UNC Chapel Hill ์์) Hao Tan and Mohit Bansal ์ LXMERT: Learning Cross-Modality Encoder Representations from Transformers for Open-Domain Question Answering ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- M-CTC-T (Facebook ์์) Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert ์ Pseudo-Labeling For Massively Multilingual Speech Recognition ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- M2M100 (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- MarianMT Machine translation models trained using OPUS data by Jรถrg Tiedemann. The Marian Framework is being developed by the Microsoft Translator Team.
- MarkupLM (Microsoft Research Asia ์์) Junlong Li, Yiheng Xu, Lei Cui, Furu Wei ์ MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Mask2Former (FAIR and UIUC ์์ ์ ๊ณต)์ Bowen Cheng, Ishan Misra, Alexander G. Schwing, Alexander Kirillov, Rohit Girdhar.์ Masked-attention Mask Transformer for Universal Image Segmentation๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- MaskFormer (Meta and UIUC ์์) Bowen Cheng, Alexander G. Schwing, Alexander Kirillov ์ Per-Pixel Classification is Not All You Need for Semantic Segmentation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- MatCha (Google AI ์์ ์ ๊ณต)์ Fangyu Liu, Francesco Piccinno, Syrine Krichene, Chenxi Pang, Kenton Lee, Mandar Joshi, Yasemin Altun, Nigel Collier, Julian Martin Eisenschlos.์ MatCha: Enhancing Visual Language Pretraining with Math Reasoning and Chart Derendering๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- mBART (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- mBART-50 (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- MEGA (Facebook ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- Megatron-BERT (NVIDIA ์์) 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 ๋ฐํํ์ต๋๋ค.
- Megatron-GPT2 (NVIDIA ์์) 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 ๋ฐํํ์ต๋๋ค.
- MGP-STR (Alibaba Research ์์ ์ ๊ณต)์ Peng Wang, Cheng Da, and Cong Yao.์ Multi-Granularity Prediction for Scene Text Recognition๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- mLUKE (Studio Ousia ์์) Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka ์ mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- MobileBERT (CMU/Google Brain ์์) 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 ๋ฐํํ์ต๋๋ค.
- MobileNetV1 (Google Inc. ์์) 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 ๋ฐํํ์ต๋๋ค.
- MobileNetV2 (Google Inc. ์์) Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, Liang-Chieh Chen ์ MobileNetV2: Inverted Residuals and Linear Bottlenecks ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- MobileViT (Apple ์์) Sachin Mehta and Mohammad Rastegari ์ MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- MPNet (Microsoft Research ์์) Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu ์ MPNet: Masked and Permuted Pre-training for Language Understanding ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- MT5 (Google AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- MVP (RUC AI Box ์์) Tianyi Tang, Junyi Li, Wayne Xin Zhao and Ji-Rong Wen ์ MVP: Multi-task Supervised Pre-training for Natural Language Generation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- NAT (SHI Labs ์์) Ali Hassani, Steven Walton, Jiachen Li, Shen Li, and Humphrey Shi ์ Neighborhood Attention Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Nezha (Huawei Noahโs Ark Lab ์์) 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 ๋ฐํํ์ต๋๋ค.
- NLLB (Meta ์์) the NLLB team ์ No Language Left Behind: Scaling Human-Centered Machine Translation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- NLLB-MOE (Meta ์์ ์ ๊ณต)์ the NLLB team.์ No Language Left Behind: Scaling Human-Centered Machine Translation๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Nystrรถmformer (the University of Wisconsin - Madison ์์) 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 ๋ฐํํ์ต๋๋ค.
- OneFormer (SHI Labs ์์) Jitesh Jain, Jiachen Li, MangTik Chiu, Ali Hassani, Nikita Orlov, Humphrey Shi ์ OneFormer: One Transformer to Rule Universal Image Segmentation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- OPT (Meta AI ์์) Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al ์ OPT: Open Pre-trained Transformer Language Models ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- OWL-ViT (Google AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- Pegasus (Google ์์) Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu ์ PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- PEGASUS-X (Google ์์) Jason Phang, Yao Zhao, Peter J. Liu ์ Investigating Efficiently Extending Transformers for Long Input Summarization ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Perceiver IO (Deepmind ์์) 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 ๋ฐํํ์ต๋๋ค.
- PhoBERT (VinAI Research ์์) Dat Quoc Nguyen and Anh Tuan Nguyen ์ PhoBERT: Pre-trained language models for Vietnamese ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Pix2Struct (Google ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- PLBart (UCLA NLP ์์) Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang ์ Unified Pre-training for Program Understanding and Generation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- PoolFormer (Sea AI Labs ์์) 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 ๋ฐํํ์ต๋๋ค.
- ProphetNet (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- QDQBert (NVIDIA ์์) Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius ์ Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- RAG (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- REALM (Google Research ์์) Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat and Ming-Wei Chang ์ REALM: Retrieval-Augmented Language Model Pre-Training ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Reformer (Google Research ์์) Nikita Kitaev, ลukasz Kaiser, Anselm Levskaya ์ Reformer: The Efficient Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- RegNet (META Research ์์) Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollรกr ์ Designing Network Design Space ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- RemBERT (Google Research ์์) Hyung Won Chung, Thibault Fรฉvry, Henry Tsai, M. Johnson, Sebastian Ruder ์ Rethinking embedding coupling in pre-trained language models ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ResNet (Microsoft Research ์์) Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun ์ Deep Residual Learning for Image Recognition ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- RoBERTa (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- RoBERTa-PreLayerNorm (Facebook ์์) 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 ๋ฐํํ์ต๋๋ค.
- RoCBert (WeChatAI ์์) HuiSu, WeiweiShi, XiaoyuShen, XiaoZhou, TuoJi, JiaruiFang, JieZhou ์ RoCBert: Robust Chinese Bert with Multimodal Contrastive Pretraining ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- RoFormer (ZhuiyiTechnology ์์) Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu ์ a RoFormer: Enhanced Transformer with Rotary Position Embedding ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- SegFormer (NVIDIA ์์) 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 ๋ฐํํ์ต๋๋ค.
- SEW (ASAPP ์์) 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 ๋ฐํํ์ต๋๋ค.
- SEW-D (ASAPP ์์) 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 ๋ฐํํ์ต๋๋ค.
- SpeechT5 (Microsoft Research ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- SpeechToTextTransformer (Facebook ์์) Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino ์ fairseq S2T: Fast Speech-to-Text Modeling with fairseq ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- SpeechToTextTransformer2 (Facebook ์์) Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau ์ Large-Scale Self- and Semi-Supervised Learning for Speech Translation ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Splinter (Tel Aviv University ์์) Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy ์ Few-Shot Question Answering by Pretraining Span Selection ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- SqueezeBERT (Berkeley ์์) 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 ๋ฐํํ์ต๋๋ค.
- Swin Transformer (Microsoft ์์) 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 ๋ฐํํ์ต๋๋ค.
- Swin Transformer V2 (Microsoft ์์) 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 ๋ฐํํ์ต๋๋ค.
- Swin2SR (University of Wรผrzburg ์์) Marcos V. Conde, Ui-Jin Choi, Maxime Burchi, Radu Timofte ์ Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- SwitchTransformers (Google ์์) William Fedus, Barret Zoph, Noam Shazeer. ์ Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- T5 (Google AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- 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.
- Table Transformer (Microsoft Research ์์) Brandon Smock, Rohith Pesala, Robin Abraham ์ PubTables-1M: Towards Comprehensive Table Extraction From Unstructured Documents ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- TAPAS (Google AI ์์) Jonathan Herzig, Paweล Krzysztof Nowak, Thomas Mรผller, Francesco Piccinno and Julian Martin Eisenschlos ์ TAPAS: Weakly Supervised Table Parsing via Pre-training ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- TAPEX (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- Time Series Transformer (from HuggingFace).
- TimeSformer (Facebook ์์) Gedas Bertasius, Heng Wang, Lorenzo Torresani ์ Is Space-Time Attention All You Need for Video Understanding? ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Trajectory Transformer (the University of California at Berkeley ์์) Michael Janner, Qiyang Li, Sergey Levin ์ Offline Reinforcement Learning as One Big Sequence Modeling Problem ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Transformer-XL (Google/CMU ์์) 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 ๋ฐํํ์ต๋๋ค.
- TrOCR (Microsoft ์์) 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 ๋ฐํํ์ต๋๋ค.
- TVLT (from UNC Chapel Hill ์์) Zineng Tang, Jaemin Cho, Yixin Nie, Mohit Bansal ์ TVLT: Textless Vision-Language Transformer ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- UL2 (Google Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- UniSpeech (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- UniSpeechSat (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- UPerNet (Peking University ์์ ์ ๊ณต)์ Tete Xiao, Yingcheng Liu, Bolei Zhou, Yuning Jiang, Jian Sun.์ Unified Perceptual Parsing for Scene Understanding๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- VAN (Tsinghua University and Nankai University ์์) Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu ์ Visual Attention Network ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- VideoMAE (Multimedia Computing Group, Nanjing University ์์) Zhan Tong, Yibing Song, Jue Wang, Limin Wang ์ VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ViLT (NAVER AI Lab/Kakao Enterprise/Kakao Brain ์์) Wonjae Kim, Bokyung Son, Ildoo Kim ์ ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Vision Transformer (ViT) (Google AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- VisualBERT (UCLA NLP ์์) 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 ๋ฐํํ์ต๋๋ค.
- ViT Hybrid (Google AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- ViTMAE (Meta AI ์์) Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollรกr, Ross Girshick ์ Masked Autoencoders Are Scalable Vision Learners ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ViTMSN (Meta AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- Wav2Vec2 (Facebook AI ์์) Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli ์ wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- Wav2Vec2-Conformer (Facebook AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- Wav2Vec2Phoneme (Facebook AI ์์) Qiantong Xu, Alexei Baevski, Michael Auli ์ Simple and Effective Zero-shot Cross-lingual Phoneme Recognition ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- WavLM (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- Whisper (OpenAI ์์) Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever ์ Robust Speech Recognition via Large-Scale Weak Supervision ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- X-CLIP (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- X-MOD (Meta AI ์์ ์ ๊ณต)์ 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 ๋ฐํํ์ต๋๋ค.
- XGLM (Facebook AI ์์ ์ ๊ณต) 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 ๋ฐํํ์ต๋๋ค.
- XLM (Facebook ์์) Guillaume Lample and Alexis Conneau ์ Cross-lingual Language Model Pretraining ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- XLM-ProphetNet (Microsoft Research ์์) 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 ๋ฐํํ์ต๋๋ค.
- XLM-RoBERTa (Facebook AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- XLM-RoBERTa-XL (Facebook AI ์์) Naman Goyal, Jingfei Du, Myle Ott, Giri Anantharaman, Alexis Conneau ์ Larger-Scale Transformers for Multilingual Masked Language Modeling ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- XLM-V (Meta AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- XLNet (Google/CMU ์์) Zhilin Yang*, Zihang Dai*, Yiming Yang, Jaime Carbonell, Ruslan Salakhutdinov, Quoc V. Le ์ โXLNet: Generalized Autoregressive Pretraining for Language Understanding ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- XLS-R (Facebook AI ์์) 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 ๋ฐํํ์ต๋๋ค.
- XLSR-Wav2Vec2 (Facebook AI ์์) Alexis Conneau, Alexei Baevski, Ronan Collobert, Abdelrahman Mohamed, Michael Auli ์ Unsupervised Cross-Lingual Representation Learning For Speech Recognition ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- YOLOS (Huazhong University of Science & Technology ์์) 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 ๋ฐํํ์ต๋๋ค.
- YOSO (the University of Wisconsin - Madison ์์) Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh ์ [You Only Sample (Almost) ๋ ผ๋ฌธ๊ณผ ํจ๊ป ๋ฐํํ์ต๋๋ค.
- ์๋ก์ด ๋ชจ๋ธ์ ์ฌ๋ฆฌ๊ณ ์ถ๋์? ์ฐ๋ฆฌ๊ฐ ์์ธํ ๊ฐ์ด๋์ ํ
ํ๋ฆฟ ์ผ๋ก ์๋ก์ด ๋ชจ๋ธ์ ์ฌ๋ฆฌ๋๋ก ๋์๋๋ฆด๊ฒ์. ๊ฐ์ด๋์ ํ
ํ๋ฆฟ์ ์ด ์ ์ฅ์์
templates
ํด๋์์ ํ์ธํ์ค ์ ์์ต๋๋ค. ์ปจํธ๋ฆฌ๋ทฐ์ ๊ฐ์ด๋๋ผ์ธ์ ๊ผญ ํ์ธํด์ฃผ์๊ณ , PR์ ์ฌ๋ฆฌ๊ธฐ ์ ์ ๋ฉ์ธํ ์ด๋์๊ฒ ์ฐ๋ฝํ๊ฑฐ๋ ์ด์๋ฅผ ์คํํด ํผ๋๋ฐฑ์ ๋ฐ์ผ์๊ธธ ๋ฐ๋๋๋ค.
๊ฐ ๋ชจ๋ธ์ด Flax, PyTorch, TensorFlow์ผ๋ก ๊ตฌํ๋์๋์ง ๋๋ ๐ค Tokenizers ๋ผ์ด๋ธ๋ฌ๋ฆฌ๊ฐ ์ง์ํ๋ ํ ํฌ๋์ด์ ๋ฅผ ์ฌ์ฉํ๋์ง ํ์ธํ๋ ค๋ฉด, ์ด ํ๋ฅผ ํ์ธํ์ธ์.
์ด ๊ตฌํ์ ์ฌ๋ฌ ๋ฐ์ดํฐ๋ก ๊ฒ์ฆ๋์๊ณ (์์ ์คํฌ๋ฆฝํธ๋ฅผ ์ฐธ๊ณ ํ์ธ์) ์ค๋ฆฌ์ง๋ ๊ตฌํ์ ์ฑ๋ฅ๊ณผ ๊ฐ์์ผ ํฉ๋๋ค. ๋ํ๋จผํธ์ Examples ์น์ ์์ ์ฑ๋ฅ์ ๋ํ ์์ธํ ์ค๋ช ์ ํ์ธํ ์ ์์ต๋๋ค.
๋ ์์๋ณด๊ธฐ
์น์ | ์ค๋ช |
---|---|
๋ํ๋จผํธ | ์ ์ฒด API ๋ํ๋จผํธ์ ํํ ๋ฆฌ์ผ |
๊ณผ์ ์์ฝ | ๐ค Transformers๊ฐ ์ง์ํ๋ ๊ณผ์ ๋ค |
์ ์ฒ๋ฆฌ ํํ ๋ฆฌ์ผ | Tokenizer ํด๋์ค๋ฅผ ์ด์ฉํด ๋ชจ๋ธ์ ์ํ ๋ฐ์ดํฐ ์ค๋นํ๊ธฐ |
ํ์ต๊ณผ fine-tuning | ๐ค Transformers๊ฐ ์ ๊ณตํ๋ ๋ชจ๋ธ PyTorch/TensorFlow ํ์ต ๊ณผ์ ๊ณผ Trainer API์์ ์ฌ์ฉํ๊ธฐ |
ํต ํฌ์ด: Fine-tuning/์ฌ์ฉ ์คํฌ๋ฆฝํธ | ๋ค์ํ ๊ณผ์ ์์ ๋ชจ๋ธ fine-tuningํ๋ ์์ ์คํฌ๋ฆฝํธ |
๋ชจ๋ธ ๊ณต์ ๋ฐ ์ ๋ก๋ | ์ปค๋ฎค๋ํฐ์ fine-tune๋ ๋ชจ๋ธ์ ์ ๋ก๋ ๋ฐ ๊ณต์ ํ๊ธฐ |
๋ง์ด๊ทธ๋ ์ด์ | pytorch-transformers ๋ pytorch-pretrained-bert ์์ ๐ค Transformers๋ก ์ด๋ํ๊ธฐ |
์ธ์ฉ
๐ค Transformers ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ์ธ์ฉํ๊ณ ์ถ๋ค๋ฉด, ์ด ๋ ผ๋ฌธ์ ์ธ์ฉํด ์ฃผ์ธ์:
@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"
}