model |
5,352 |
모델 5344 · 예측 모델 4 · 학습 모델 4 · 모형 · 수학 모형 · 통계 모형 |
바로가기 |
feature |
2,499 |
함수 841 · 변수 653 · 기능 496 · 속성 162 · 특징 154 · 능력 111 · 특성 76 · 피처 6 |
바로가기 |
training |
2,444 |
학습 1631 · 훈련 794 · 트레이닝 19 |
바로가기 |
inference |
1,009 |
추론 679 · 예측 327 · 인퍼런스 3 |
바로가기 |
kernel |
813 |
커널 777 · 필터 36 · 커널 함수 · 컨볼루션 커널 · 핵 함수 |
바로가기 |
parameter |
765 |
매개변수 525 · 파라미터 232 · 모델 파라미터 8 |
바로가기 |
token |
751 |
토큰 749 · 토큰 단위 2 |
바로가기 |
tensor |
739 |
텐서 534 · 행렬 203 · 다차원 배열 2 |
바로가기 |
layer |
695 |
계층 372 · 레이어 249 · 층 69 · 추상화 수준 5 |
바로가기 |
dataset |
687 |
데이터셋 431 · 데이터 세트 207 · 학습 데이터 43 · 데이터 집합 6 · 자료 집합 |
바로가기 |
agent |
605 |
에이전트 602 · 행위자 2 · 자율 에이전트 1 · 대리인 · 지능형 대리인 · 학습 주체 |
바로가기 |
exploitation |
574 |
활용 385 · 이용 189 |
바로가기 |
architecture |
562 |
구조 290 · 아키텍처 272 |
바로가기 |
attention |
562 |
어텐션 473 · 주의 47 · 집중 41 · 주의 메커니즘 1 · 집중 메커니즘 |
바로가기 |
batch |
561 |
배치 550 · 일괄 처리 11 · 배치 샘플 |
바로가기 |
loss |
428 |
손실 308 · 감소 62 · 오차 29 · 로스 18 · 손실값 11 |
바로가기 |
weight |
412 |
가중치 396 · 가중 7 · 중요도 4 · 비중 3 · 웨이트 2 · 연결 강도 |
바로가기 |
fine-tuning |
411 |
미세 조정 296 · 파인튜닝 68 · 미세조정 23 · 미세 튜닝 14 · 전이학습 10 |
바로가기 |
quantization |
374 |
양자화 374 · 저비트 변환 · 정수화 · 퀀타이제이션 |
바로가기 |
gradient |
371 |
변화도 278 · 기울기 39 · 그래디언트 29 · 경사 25 |
바로가기 |
environment |
351 |
환경 344 · 학습 환경 7 |
바로가기 |
precision |
335 |
정밀도 180 · 정확도 155 · 프리시전 |
바로가기 |
neural network |
309 |
신경망 309 · 뉴럴 네트워크 · 인공 신경망 |
바로가기 |
language model |
308 |
언어 모델 308 |
바로가기 |
checkpoint |
294 |
체크포인트 278 · 스냅샷 16 · 저장점 |
바로가기 |
label |
293 |
레이블 184 · 정답 61 · 라벨 48 |
바로가기 |
classification |
292 |
분류 292 · 범주화 · 카테고리화 |
바로가기 |
pipeline |
256 |
파이프라인 256 · 작업 관로 · 처리 흐름 |
바로가기 |
iteration |
247 |
반복 224 · 학습 단계 23 · 이터레이션 |
바로가기 |
initialization |
240 |
초기화 237 · 가중치 초기화 3 |
바로가기 |
optimizer |
237 |
옵티마이저 226 · 최적화 알고리즘 8 · 최적화기 3 |
바로가기 |
benchmark |
229 |
벤치마크 229 · 성능 평가 기준 |
바로가기 |
embedding |
228 |
임베딩 226 · 매장 2 · 내장 표현 · 벡터 표현 · 분산 표현 |
바로가기 |
segmentation |
216 |
분할 214 · 세그멘테이션 2 |
바로가기 |
tokenizer |
201 |
토크나이저 200 · 토큰화기 1 |
바로가기 |
context |
199 |
컨텍스트 148 · 문맥 51 |
바로가기 |
head |
198 |
헤드 198 |
바로가기 |
prompt |
191 |
프롬프트 184 · 입력 프롬프트 7 |
바로가기 |
query |
183 |
쿼리 96 · 질의 87 |
바로가기 |
normalization |
180 |
정규화 173 · 표준화 7 · 노멀라이제이션 |
바로가기 |
backpropagation |
177 |
역전파 176 · 역전파 알고리즘 1 · 오차 역전파 |
바로가기 |
regularization |
173 |
정규화 173 · 규제화 · 레귤라라이제이션 |
바로가기 |
preprocessing |
172 |
전처리 172 |
바로가기 |
pre-training |
168 |
사전 훈련 133 · 사전학습 35 · 프리트레이닝 |
바로가기 |
exploration |
166 |
탐색 166 · 탐험 |
바로가기 |
transformer |
166 |
트랜스포머 134 · 트랜스포머 모델 30 · 변환기 2 |
바로가기 |
tile |
149 |
타일 149 |
바로가기 |
tokenization |
147 |
토큰화 147 · 토큰 분리 |
바로가기 |
node |
134 |
노드 109 · 유닛 22 · 뉴런 3 · 버텍스 · 정점 |
바로가기 |
convolution |
128 |
합성곱 104 · 컨볼루션 18 · 합성곱 연산 6 |
바로가기 |
large language model |
111 |
대규모 언어 모델 111 |
바로가기 |
metric |
111 |
지표 53 · 메트릭 27 · 평가 지표 25 · 평가지표 6 |
바로가기 |
validation |
109 |
검증 108 · 유효성 검사 1 · 밸리데이션 |
바로가기 |
padding |
101 |
패딩 94 · 채우기 7 |
바로가기 |
forward pass |
96 |
순전파 96 · 순방향 전파 · 전방 전달 |
바로가기 |
vocabulary |
96 |
어휘 94 · 어휘 사전 2 |
바로가기 |
pruning |
94 |
가지치기 94 · 프루닝 |
바로가기 |
workflow |
91 |
워크플로우 79 · 워크플로 7 · 작업 흐름 5 · 작업흐름 |
바로가기 |
training loop |
87 |
학습 루프 49 · 훈련 루프 36 · 학습 반복 2 |
바로가기 |
tool calling |
83 |
도구 호출 49 · 도구 사용 32 · 툴 호출 2 · 툴 콜링 |
바로가기 |
channel |
82 |
채널 82 |
바로가기 |
hyperparameter |
81 |
하이퍼파라미터 79 · 초매개변수 1 · 하이퍼 파라미터 1 |
바로가기 |
loss function |
76 |
손실 함수 76 |
바로가기 |
derivative |
72 |
미분 57 · 도함수 14 · 미분값 1 |
바로가기 |
collective |
71 |
집합 통신 71 |
바로가기 |
question answering |
68 |
질의 응답 31 · 질의응답 30 · 질문 답변 7 |
바로가기 |
hidden state |
63 |
은닉 상태 63 |
바로가기 |
image classification |
63 |
이미지 분류 54 · 영상 분류 9 |
바로가기 |
deep learning |
62 |
딥러닝 61 · 심층 학습 1 |
바로가기 |
matrix multiplication |
62 |
행렬 곱셈 27 · 행렬곱 21 · 행렬 곱 14 |
바로가기 |
zero-shot learning |
59 |
제로샷 59 · 무예시 학습 · 제로샷 학습 |
바로가기 |
mini-batch |
58 |
미니 배치 38 · 미니배치 20 |
바로가기 |
image processor |
57 |
이미지 프로세서 57 · 이미지 전처리기 |
바로가기 |
linear layer |
57 |
선형 계층 38 · 선형 레이어 12 · FC 레이어 7 |
바로가기 |
mixed precision |
54 |
혼합 정밀도 50 · 혼합 정밀도 학습 4 |
바로가기 |
momentum |
53 |
관성 37 · 모멘텀 16 · 운동량 |
바로가기 |
pretrained model |
53 |
사전 훈련된 모델 52 · 사전학습 모델 1 |
바로가기 |
sparsity |
53 |
희소성 49 · 희소화 4 |
바로가기 |
computational graph |
52 |
계산 그래프 26 · 연산 그래프 26 |
바로가기 |
profiling |
52 |
프로파일링 46 · 성능 분석 6 |
바로가기 |
MCP server |
51 |
MCP 서버 51 |
바로가기 |
epoch |
50 |
에폭 45 · 에포크 3 · 학습 반복 2 · 학습 주기 |
바로가기 |
bias |
48 |
편향 41 · 바이어스 3 · 절편 3 · 데이터 편향 1 · 알고리즘 편향 · 편향 항 |
바로가기 |
chunk |
48 |
조각 22 · 묶음 16 · 청크 10 |
바로가기 |
distributed training |
48 |
분산 학습 48 |
바로가기 |
modality |
48 |
모달리티 38 · 양식 10 |
바로가기 |
augmentation |
47 |
증강 38 · 데이터 증강 9 |
바로가기 |
sequence-to-sequence |
46 |
시퀀스-투-시퀀스 24 · 인코더-디코더 22 |
바로가기 |
learning rate |
44 |
학습률 24 · 학습 속도 17 · 러닝 레이트 3 |
바로가기 |
object detection |
44 |
객체 탐지 37 · 객체 감지 6 · 객체 검출 1 · 물체 검출 |
바로가기 |
self-attention |
44 |
셀프 어텐션 43 · 자기 어텐션 1 · 자가 어텐션 |
바로가기 |
baseline |
42 |
베이스라인 22 · 기준선 20 |
바로가기 |
KV cache |
41 |
키-값 캐시 24 · KV 캐시 17 |
바로가기 |
leaderboard |
40 |
리더보드 39 · 순위표 1 |
바로가기 |
convolutional neural network |
39 |
합성곱 신경망 32 · 합성곱 네트워크 4 · 컨볼루션 신경망 3 |
바로가기 |
parsing |
39 |
파싱 36 · 구문 분석 3 |
바로가기 |
convergence |
37 |
수렴 37 · 수렴 상태 · 수렴점 |
바로가기 |
rollout |
37 |
롤아웃 37 · 궤적 생성 · 롤아웃 샘플 · 전개 |
바로가기 |
autograd |
36 |
자동 미분 34 · 자동 미분 엔진 2 · 자동 그래디언트 |
바로가기 |
batch normalization |
35 |
배치 정규화 35 · 배치 놈 |
바로가기 |
data parallel |
34 |
데이터 병렬 처리 21 · 데이터 병렬화 13 |
바로가기 |
likelihood |
33 |
우도 29 · 가능도 4 |
바로가기 |
natural language processing |
33 |
자연어 처리 32 · 자연어처리 1 |
바로가기 |
computer vision |
31 |
컴퓨터 비전 31 · 컴퓨터 시각 |
바로가기 |
tensor parallelism |
31 |
텐서 병렬 처리 16 · 텐서 병렬화 15 |
바로가기 |
kernel launch |
28 |
커널 실행 28 |
바로가기 |
data collator |
27 |
데이터 콜레이터 27 · 데이터 취합기 · 데이터 취합자 |
바로가기 |
speech recognition |
27 |
자동 음성 인식 15 · 음성 인식 12 |
바로가기 |
distributed data parallel |
25 |
분산 데이터 병렬 14 · 분산 데이터 병렬 처리 11 |
바로가기 |
reduction |
25 |
리덕션 25 |
바로가기 |
supervised learning |
25 |
지도 학습 23 · 지도학습 2 |
바로가기 |
generative model |
24 |
생성 모델 24 · 생성형 모델 · 제너레이티브 모델 |
바로가기 |
memory bandwidth |
24 |
메모리 대역폭 24 |
바로가기 |
softmax |
24 |
소프트맥스 24 · 소프트맥스 함수 |
바로가기 |
backbone |
23 |
백본 23 |
바로가기 |
word embedding |
23 |
단어 임베딩 15 · 단어 벡터 6 · 워드 임베딩 2 |
바로가기 |
reward |
22 |
보상 22 · 보상 신호 |
바로가기 |
bounding box |
20 |
바운딩 박스 18 · 경계 상자 2 · 검출 상자 |
바로가기 |
causal language modeling |
20 |
인과 언어 모델링 11 · 인과적 언어 모델링 9 |
바로가기 |
decoder model |
20 |
디코더 모델 20 |
바로가기 |
pooling |
20 |
풀링 15 · 다운샘플링 5 · 서브샘플링 |
바로가기 |
activation function |
19 |
활성화 함수 18 · 활성 함수 1 · 비선형 함수 |
바로가기 |
feature extraction |
19 |
특징 추출 15 · 특성 추출 4 |
바로가기 |
image captioning |
19 |
이미지 캡셔닝 19 |
바로가기 |
image segmentation |
19 |
이미지 분할 17 · 세그멘테이션 2 · 영상 분할 |
바로가기 |
mixture of experts |
19 |
전문가 혼합 17 · 전문가 혼합 모델 2 |
바로가기 |
stride |
17 |
스트라이드 15 · 보폭 2 |
바로가기 |
token classification |
17 |
토큰 분류 17 |
바로가기 |
attention mask |
16 |
어텐션 마스크 16 · 주의 마스크 |
바로가기 |
chain of thought |
16 |
사고 과정 10 · 생각의 사슬 6 · 연쇄 추론 |
바로가기 |
system prompt |
16 |
시스템 프롬프트 16 |
바로가기 |
automatic speech recognition |
15 |
자동 음성 인식 15 · 자동 음성인식 |
바로가기 |
logit |
15 |
로짓 15 · 로그 오즈 · 원시 출력 |
바로가기 |
perplexity |
15 |
펄플렉서티 15 · 퍼플렉시티 · 혼잡도 |
바로가기 |
post-training |
15 |
사후 학습 15 · 포스트 트레이닝 |
바로가기 |
retrieval-augmented generation |
15 |
검색 증강 생성 15 |
바로가기 |
TensorBoard |
15 |
텐서보드 15 |
바로가기 |
transpose |
15 |
전치 14 · 전치 행렬 1 · 행렬 전치 |
바로가기 |
evaluation mode |
14 |
평가 모드 9 · 추론 모드 5 |
바로가기 |
inference provider |
14 |
추론 공급자 9 · 추론 제공자 5 · 추론 서비스 사업자 |
바로가기 |
named entity recognition |
14 |
개체명 인식 14 · 개체 인식 |
바로가기 |
pipeline parallelism |
14 |
파이프라인 병렬 처리 8 · 파이프라인 병렬화 6 |
바로가기 |
reasoning model |
14 |
추론 모델 14 |
바로가기 |
autotuner |
13 |
오토튜너 13 |
바로가기 |
model parallelism |
13 |
모델 병렬화 8 · 모델 병렬 처리 5 |
바로가기 |
pixel values |
13 |
픽셀 값 13 |
바로가기 |
teacher model |
13 |
교사 모델 7 · 교사 6 |
바로가기 |
flatten |
12 |
평탄화 12 · 1차원 변환 · 평활화 |
바로가기 |
global memory |
12 |
전역 메모리 12 |
바로가기 |
model serving |
12 |
모델 배포 9 · 모델 서빙 3 · 추론 서빙 |
바로가기 |
reasoning trace |
12 |
추론 과정 12 · 추론 흔적 |
바로가기 |
dropout |
11 |
드롭아웃 11 · 뉴런 비활성화 · 드롭아웃 정규화 |
바로가기 |
element-wise |
11 |
요소별 7 · 원소별 4 · 성분별 |
바로가기 |
ensemble |
11 |
앙상블 11 · 모델 결합 · 앙상블 학습 |
바로가기 |
feature extractor |
11 |
특징 추출기 8 · 특성 추출기 3 · 피처 추출기 |
바로가기 |
feature map |
11 |
특징 맵 11 · 특성 맵 · 피처맵 |
바로가기 |
multimodal model |
11 |
멀티모달 모델 11 · 다중양식 모델 |
바로가기 |
overfitting |
11 |
과적합 11 · 과대적합 · 오버피팅 |
바로가기 |
sentiment analysis |
11 |
감정 분석 8 · 감성 분석 3 · 오피니언 마이닝 |
바로가기 |
special token |
11 |
특수 토큰 11 |
바로가기 |
gradient descent |
10 |
경사하강법 10 · 그래디언트 디센트 · 기울기 하강법 |
바로가기 |
MCP client |
10 |
MCP 클라이언트 10 |
바로가기 |
student model |
10 |
학생 모델 6 · 학생 4 |
바로가기 |
transfer learning |
10 |
전이학습 10 · 전달 학습 · 트랜스퍼 러닝 |
바로가기 |
data augmentation |
9 |
데이터 증강 9 · 데이터 어그멘테이션 · 데이터 확장 |
바로가기 |
image patch |
9 |
이미지 패치 8 · 이미지 조각 1 |
바로가기 |
lowering |
9 |
저수준화 9 |
바로가기 |
masked language modeling |
9 |
마스크 언어 모델링 6 · 마스크드 언어 모델링 3 |
바로가기 |
memory traffic |
9 |
메모리 트래픽 9 |
바로가기 |
spectrogram |
9 |
스펙트로그램 9 · 분광도 |
바로가기 |
unsupervised learning |
9 |
비지도 학습 9 · 비지도학습 |
바로가기 |
encoder model |
8 |
인코더 모델 8 |
바로가기 |
audio classification |
7 |
오디오 분류 7 |
바로가기 |
knowledge distillation |
7 |
지식 증류 7 · 모델 증류 · 지식 전이 |
바로가기 |
Microscaling |
7 |
마이크로스케일링 7 |
바로가기 |
sequence classification |
7 |
시퀀스 분류 7 |
바로가기 |
sigmoid |
7 |
시그모이드 7 · 로지스틱 함수 · 시그모이드 함수 |
바로가기 |
code agent |
6 |
코드 에이전트 6 |
바로가기 |
crop |
6 |
크롭 5 · 자르기 1 |
바로가기 |
input IDs |
6 |
입력 ID 6 · 입력 토큰 ID |
바로가기 |
multi-step agent |
6 |
멀티스텝 에이전트 6 |
바로가기 |
reinforcement learning |
6 |
강화학습 6 |
바로가기 |
representation learning |
6 |
표현 학습 6 |
바로가기 |
structured output |
6 |
구조화된 출력 6 |
바로가기 |
vertical fusion |
6 |
수직 융합 6 |
바로가기 |
agency |
5 |
에이전시 5 |
바로가기 |
chain rule |
5 |
연쇄 법칙 5 · 연쇄율 · 체인 룰 |
바로가기 |
dataloader |
5 |
데이터로더 5 · 데이터 적재기 |
바로가기 |
deep neural network |
5 |
심층 신경망 5 |
바로가기 |
GAN |
5 |
적대적 생성 신경망 4 · 생성적 적대 신경망 1 · 갠 |
바로가기 |
hidden layer |
5 |
은닉층 5 · 중간층 · 히든 레이어 |
바로가기 |
kernel fusion |
5 |
커널 융합 5 |
바로가기 |
mean-squared error |
5 |
평균 제곱 오차 4 · 평균제곱오차 1 |
바로가기 |
recall |
5 |
리콜 2 · 재현율 2 · 민감도 1 |
바로가기 |
recurrent neural network |
5 |
순환 신경망 5 · 재귀 신경망 |
바로가기 |
vision-language model |
5 |
비전-언어 모델 5 · 시각언어 모델 |
바로가기 |
agent memory |
4 |
에이전트 메모리 4 |
바로가기 |
cross-entropy |
4 |
교차 엔트로피 2 · 교차 엔트로피 손실 2 · 크로스 엔트로피 |
바로가기 |
DAG |
4 |
방향성 비순환 그래프 4 · 비순환 방향 그래프 · 유향 비순환 그래프 |
바로가기 |
device-agnostic |
4 |
장치에 구애받지 않는 4 |
바로가기 |
edge deployment |
4 |
엣지 배포 2 · 엣지 추론 1 · 온디바이스 배포 1 |
바로가기 |
freezing |
4 |
동결 4 · 가중치 동결 · 파라미터 고정 |
바로가기 |
hardware-agnostic |
4 |
하드웨어에 구애받지 않는 4 |
바로가기 |
in-place |
4 |
제자리 4 · 제자리 연산 |
바로가기 |
one-hot encoding |
4 |
원핫 인코딩 4 · 원-핫 인코딩 |
바로가기 |
sampling rate |
4 |
샘플링 레이트 4 · 샘플율 · 표본화율 |
바로가기 |
adversarial example |
3 |
적대적 예제 3 · 공격 샘플 · 적대적 사례 |
바로가기 |
autoregressive model |
3 |
자기회귀 모델 3 |
바로가기 |
classification token |
3 |
[CLS] 토큰 3 · 분류 토큰 |
바로가기 |
downstream task |
3 |
하위 과제 2 · 다운스트림 과제 1 |
바로가기 |
early stopping |
3 |
조기 종료 3 · 조기 중단 |
바로가기 |
epilogue fusion |
3 |
에필로그 융합 3 |
바로가기 |
exact match |
3 |
정확 일치 3 |
바로가기 |
fine-tuned model |
3 |
미세 조정 모델 3 · 미세조정 모델 |
바로가기 |
gradient accumulation |
3 |
그래디언트 누적 3 · 그래디언트 축적 |
바로가기 |
learning rate scheduler |
3 |
학습률 스케줄러 3 · 학습률 조절기 |
바로가기 |
machine translation |
3 |
기계 번역 3 · 신경 기계 번역 |
바로가기 |
pointwise fusion |
3 |
점별 융합 3 |
바로가기 |
synthetic data |
3 |
합성 데이터 3 · 인공 데이터 |
바로가기 |
all-reduce |
2 |
올리듀스 2 · 전체 집계 |
바로가기 |
dense layer |
2 |
완전 연결 계층 2 · 밀집층 · 완전 연결층 · 전결합층 |
바로가기 |
hessian |
2 |
헤시안 2 · 헤세 행렬 · 헤시안 행렬 |
바로가기 |
horizontal fusion |
2 |
수평 융합 2 |
바로가기 |
model context protocol |
2 |
모델 컨텍스트 프로토콜 2 |
바로가기 |
race condition |
2 |
경쟁 조건 2 |
바로가기 |
rotary position embedding |
2 |
회전 위치 임베딩 2 · 로터리 위치 임베딩 |
바로가기 |
semi-structured sparsity |
2 |
반구조화 희소성 2 |
바로가기 |
waveform |
2 |
웨이브폼 1 · 파형 1 · 음성 파형 |
바로가기 |
confusion matrix |
1 |
오차 행렬 1 · 혼동 행렬 |
바로가기 |
direct preference optimization |
1 |
직접 선호 최적화 1 |
바로가기 |
off-policy |
1 |
오프폴리시 1 · 오프-정책 |
바로가기 |
planning step |
1 |
계획 단계 1 |
바로가기 |
preference data |
1 |
선호도 데이터 1 · 선호 데이터 |
바로가기 |
prologue fusion |
1 |
프롤로그 융합 1 |
바로가기 |
rejection sampling |
1 |
거부 샘플링 1 · 리젝션 샘플링 |
바로가기 |
semantic search |
1 |
시맨틱 검색 1 |
바로가기 |
text-to-speech |
1 |
텍스트 음성 변환 1 · 음성 합성 |
바로가기 |
Hugging Face KREW와 PyTorch의 번역 문서를 모두 합쳐, 사전에 등록된 용어의 한국어 표기가 나타난 횟수를 많은 순으로 정리했습니다.
33f3ccc23826· 생성 2026-09-13T13:25:44+00:00npm run update:usage/ 상세 규칙은 docs/usage-statisticsNote
횟수는 한국어 문자열의 출현 빈도이며, 특정 영문 용어의 번역 횟수나 권장 번역을 뜻하지 않습니다. 사전에 등록된 후보를 그대로 검색하므로 일상어·다의어가 섞입니다. 예를 들어
feature의 후보에는기능의 동의어인함수·능력이 포함되어 있고,exploitation의 후보활용·이용은 일반 문맥에서도 쓰입니다. 순위는 "어떤 표기가 문서에 많이 보이는가"로만 읽어 주세요.빈도순 (237개)
modelfeaturetraininginferencekernelparametertokentensorlayerdatasetagentexploitationarchitectureattentionbatchlossweightfine-tuningquantizationgradientenvironmentprecisionneural networklanguage modelcheckpointlabelclassificationpipelineiterationinitializationoptimizerbenchmarkembeddingsegmentationtokenizercontextheadpromptquerynormalizationbackpropagationregularizationpreprocessingpre-trainingexplorationtransformertiletokenizationnodeconvolutionlarge language modelmetricvalidationpaddingforward passvocabularypruningworkflowtraining looptool callingchannelhyperparameterloss functionderivativecollectivequestion answeringhidden stateimage classificationdeep learningmatrix multiplicationzero-shot learningmini-batchimage processorlinear layermixed precisionmomentumpretrained modelsparsitycomputational graphprofilingMCP serverepochbiaschunkdistributed trainingmodalityaugmentationsequence-to-sequencelearning rateobject detectionself-attentionbaselineKV cacheleaderboardconvolutional neural networkparsingconvergencerolloutautogradbatch normalizationdata parallellikelihoodnatural language processingcomputer visiontensor parallelismkernel launchdata collatorspeech recognitiondistributed data parallelreductionsupervised learninggenerative modelmemory bandwidthsoftmaxbackboneword embeddingrewardbounding boxcausal language modelingdecoder modelpoolingactivation functionfeature extractionimage captioningimage segmentationmixture of expertsstridetoken classificationattention maskchain of thoughtsystem promptautomatic speech recognitionlogitperplexitypost-trainingretrieval-augmented generationTensorBoardtransposeevaluation modeinference providernamed entity recognitionpipeline parallelismreasoning modelautotunermodel parallelismpixel valuesteacher modelflattenglobal memorymodel servingreasoning tracedropoutelement-wiseensemblefeature extractorfeature mapmultimodal modeloverfittingsentiment analysisspecial tokengradient descentMCP clientstudent modeltransfer learningdata augmentationimage patchloweringmasked language modelingmemory trafficspectrogramunsupervised learningencoder modelaudio classificationknowledge distillationMicroscalingsequence classificationsigmoidcode agentcropinput IDsmulti-step agentreinforcement learningrepresentation learningstructured outputvertical fusionagencychain ruledataloaderdeep neural networkGANhidden layerkernel fusionmean-squared errorrecallrecurrent neural networkvision-language modelagent memorycross-entropyDAGdevice-agnosticedge deploymentfreezinghardware-agnosticin-placeone-hot encodingsampling rateadversarial exampleautoregressive modelclassification tokendownstream taskearly stoppingepilogue fusionexact matchfine-tuned modelgradient accumulationlearning rate schedulermachine translationpointwise fusionsynthetic dataall-reducedense layerhessianhorizontal fusionmodel context protocolrace conditionrotary position embeddingsemi-structured sparsitywaveformconfusion matrixdirect preference optimizationoff-policyplanning steppreference dataprologue fusionrejection samplingsemantic searchtext-to-speech출현이 없는 용어 26개
anchor boxBERTCUDAdeep Q-networkdirect alignment algorithmsfeed-forward networkforward KLGEMMgradient synchronizationheld-out setinstruction tuningLLM-as-a-judgemarginMarkov decision processmel-frequencyon-policyONNXpolicy gradientprefixReActReLUreverse KLreward modelsupervised fine-tuningunderfittingvanishing gradient