J & J - 정성태의 닷넷 이야기

사용자

메뉴

최근 덧글

[정성태] 그냥 RSS Reader 기능과 약간의 UI 편의성 때문에 사용...[이종효] 오래된 소프트웨어는 보안 위협이 되기도 합니다. 혹시 어떤 기능...[정성태] @Keystroke IEEE의 문서를 소개해 주시다니... +_...[손민수 (Keystroke)] 괜히 듀얼채널 구성할 때 한번에 같은 제품 사라고 하는 것이 아...[정성태] 전각(Full-width)/반각(Half-width) 기능을 토...[정성태] Vector에 대한 내용은 없습니다. Vector가 닷넷 BCL...[orion] 글 읽고 찾아보니 디자인 타임에는 InitializeCompon...[orion] 연휴 전에 재현 프로젝트 올리자 생각해 놓고 여의치 않아서 못 ...[정성태] 아래의 글에 정리했으니 참고하세요. C# - Typed D...[정성태] 간단한 재현 프로젝트라도 있을까요? 저런 식으로 설명만 해...

글쓰기

제목

이름

암호

전자우편

HTML

홈페이지

유형

내용

<div style='display: inline'>
<h1 style='font-family: Malgun Gothic, Consolas; font-size: 20pt; color: #006699; text-align: center; font-weight: bold'>파이썬 - "Transformers (신경망 언어모델 라이브러리) 강좌" - 1장 2절 코드 실습</h1>

다음의 강좌에서, 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
Transformers (신경망 언어모델 라이브러리) 강좌
; <a target='tab' href='https://wikidocs.net/book/8056'>https://wikidocs.net/book/8056</a>
</pre>
 
1장 2절의 내용에, 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
2. 🤗Transformers가 할 수 있는 일들
; <a target='tab' href='https://wikidocs.net/166787'>https://wikidocs.net/166787</a>
</pre>
 
포함된 코드를 구글 Colab에서 수행한 결과를 나열해 봅니다. ^^ 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
!pip install transformers
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

classifier = pipeline("sentiment-analysis")

classifier("I've been waiting for a HuggingFace course my whole life.")

classifier(["I've been waiting for a HuggingFace course my whole life.", "I hate this so much!"])

# 실행 결과
[{'label': 'POSITIVE', 'score': 0.9598048329353333},
 {'label': 'NEGATIVE', 'score': 0.9994558691978455}]
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

classifier = pipeline("zero-shot-classification")
classifier(
    "This is a course about the Transformers library",
    candidate_labels=["education", "politics", "business"],
)

# 실행 결과
{'sequence': 'This is a course about the Transformers library',
 'labels': ['education', 'business', 'politics'],
 'scores': [0.8445989489555359, 0.11197412759065628, 0.04342695698142052]}
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

generator = pipeline("text-generation")
generator("In this course, we will teach you how to")

# 실행 결과
[{'generated_text': "In this course, we will teach you how to use NLP with the following tasks. In this course, we will work with a computer running NLP. I'm using the npc-get system to find your NPM scripts and to start"}]
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

generator = pipeline("text-generation", model="distilgpt2")    # distilgpt2 모델을 로드한다.
generator(
    "In this course, we will teach you how to",
    max_length=30,
    num_return_sequences=2,
)

# 실행 결과
[{'generated_text': 'In this course, we will teach you how to create a simple and fun web design using Photoshop for building a simple website.\n\n\n\nThe'},
 {'generated_text': 'In this course, we will teach you how to apply the following basic concepts to your life (see below). This course aims to help you to choose'}]
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

unmasker = pipeline("fill-mask")
unmasker("This course will teach you all about <mask> models.", top_k=3)

# 실행 결과
[{'score': 0.19619806110858917,
 'token': 30412,
 'token_str': ' mathematical',
 'sequence': 'This course will teach you all about mathematical models.'},
 {'score': 0.04052723944187164,
 'token': 38163,
 'token_str': ' computational',
 'sequence': 'This course will teach you all about computational models.'},
 {'score': 0.03301795944571495,
 'token': 27930,
 'token_str': ' predictive',
 'sequence': 'This course will teach you all about predictive models.'}]
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

ner = pipeline("ner", grouped_entities=True)
ner("My name is Sylvain and I work at Hugging Face in Brooklyn.")

# 실행 결과
[{'entity_group': 'PER',
 'score': 0.9981694,
 'word': 'Sylvain',
 'start': 11,
 'end': 18},
 {'entity_group': 'ORG',
 'score': 0.9796019,
 'word': 'Hugging Face',
 'start': 33,
 'end': 45},
 {'entity_group': 'LOC',
 'score': 0.9932106,
 'word': 'Brooklyn',
 'start': 49,
 'end': 57}]
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

question_answerer = pipeline("question-answering")
question_answerer(
    question="Where do I work?",
    context="My name is Sylvain and I work at Hugging Face in Brooklyn",
)

# 실행 결과
{'score': 0.6949767470359802, 'start': 33, 'end': 45, 'answer': 'Hugging Face'}
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

summarizer = pipeline("summarization")
summarizer(
    """
    America has changed dramatically during recent years. Not only has the number of 
    graduates in traditional engineering disciplines such as mechanical, civil, 
    electrical, chemical, and aeronautical engineering declined, but in most of 
    the premier American universities engineering curricula now concentrate on 
    and encourage largely the study of engineering science. As a result, there 
    are declining offerings in engineering subjects dealing with infrastructure, 
    the environment, and related issues, and greater concentration on high 
    technology subjects, largely supporting increasingly complex scientific 
    developments. While the latter is important, it should not be at the expense 
    of more traditional engineering.

Rapidly developing economies such as China and India, as well as other 
    industrial countries in Europe and Asia, continue to encourage and advance 
    the teaching of engineering. Both China and India, respectively, graduate 
    six and eight times as many traditional engineers as does the United States. 
    Other industrial countries at minimum maintain their output, while America 
    suffers an increasingly serious decline in the number of engineering graduates 
    and a lack of well-educated engineers.
    """
)

# 실행 결과
[{'summary_text': ' America has changed dramatically during recent years . The number of engineering graduates in the U.S. has declined in traditional engineering disciplines such as mechanical, civil, electrical, chemical, and aeronautical engineering . Rapidly developing economies such as China and India, as well as other industrial countries in Europe and Asia, continue to encourage and advance engineering .'}]
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ko-en")
translator("그동안 너무 잘해 주셔서 감사드립니다.")

# 실행 결과
[{'translation_text': 'Thank you so much for your kindness.'}]
</pre>
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
from transformers import pipeline

unmasker = pipeline("fill-mask", model="bert-base-uncased")
result = unmasker("This man works as a [MASK].")
print([r["token_str"] for r in result])

result = unmasker("This woman works as a [MASK].")
print([r["token_str"] for r in result])

# 실행 결과
['carpenter', 'lawyer', 'farmer', 'businessman', 'doctor']
['nurse', 'maid', 'teacher', 'waitress', 'prostitute']
</pre>
 
<hr style='width: 50%' /> 
 
참고로, Colab이 아닌 Windows에서의 python 환경에서 테스트하고 싶다면 우선 python 3.10으로 설치하고, 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
Python 3.10.0
; <a target='tab' href='https://www.python.org/downloads/release/python-3100/'>https://www.python.org/downloads/release/python-3100/</a>
</pre>
 
제 경우에는 <a target='tab' href='https://www.sysnet.pe.kr/2/0/12861'>"Windows embeddable package (64-bit)"를 다운로드했고 (따라서 _pth 파일과 pip을 별도로 설정한 다음)</a>, virtualenv도 마저 설치합니다. 
 
이후 새로운 virtualenv 환경을 만들고, 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
C:\python\llml&gt; virtualenv test
created virtual environment CPython3.10.0.final.0-64 in 3934ms
 ...[생략]...
</pre>
 
활성화시킨 후, 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
C:\python\llml&gt; cd test
C:\python\llml\test&gt; .\Scripts\activate

(test) C:\python\llml\test&gt;
</pre>
 
transformers를 설치합니다. 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
(test) C:\python\llml\test&gt; python -m pip install "transformers[sentencepiece]"
</pre>
 
그런데, 이것만으로는 pipeline 예제를 실행하는 경우 예외가 발생합니다. 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
Traceback (most recent call last):
 File "C:\python\llml\test\sc1.py", line 3, in <module>
 unmasker = pipeline("fill-mask", model="bert-base-uncased")
 File "C:\python\llml\test\lib\site-packages\transformers\pipelines\__init__.py", line 788, in pipeline
 framework, model = infer_framework_load_model(
 File "C:\python\llml\test\lib\site-packages\transformers\pipelines\base.py", line 222, in infer_framework_load_model
 raise RuntimeError(
RuntimeError: At least one of TensorFlow 2.0 or PyTorch should be installed. To install TensorFlow 2.0, read the instructions at https://www.tensorflow.org/install/ To install PyTorch, read the instructions at https://pytorch.org/.
</pre>
 
메시지에서 의미하듯이 PyTorch (또는 <a target='tab' href='https://www.sysnet.pe.kr/2/0/12817'>tensorflow</a>)를 설치해야 하는데요, 
 
<pre style='margin: 10px 0px 10px 10px; padding: 10px 0px 10px 10px; background-color: #fbedbb; overflow: auto; font-family: Consolas, Verdana;' >
START LOCALLY
; <a target='tab' href='https://pytorch.org/get-started/locally/'>https://pytorch.org/get-started/locally/</a>

// NVidia CUDA 11.8
python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

// CPU
python -m pip install torch torchvision torchaudio
</pre>
 
PyTorch의 경우 지원하는 Compute Platform에 CPU와 CUDA만 있으므로 아쉽게도 <a target='tab' href='https://www.sysnet.pe.kr/0/0/522'>AMD 그래픽 카드</a>에서는 사용할 수 없습니다. 하지만, 이미 이 글에서 실습한 코드들의 경우 Model을 직접 훈련시키는 것이 아닌, 이미 훈련된 Model을 사용하는 것에 불과하므로 CPU로도 문제없이 실습이 가능합니다. (<a target='tab' href='https://wikidocs.net/166802'>3장의 미세 조정 학습</a>까지는!)

<hr />[이 글에 대해서 여러분들과 의견을 공유하고 싶습니다. 틀리거나 미흡한 부분 또는 의문 사항이 있으시면 언제든 댓글 남겨주십시오.]

</div>

첨부파일

스팸 방지용 인증 번호

1856 (왼쪽의 숫자를 입력해야 합니다.)