Prepare training datasets for large language models with GPU-assisted cleaning. Covers duplicate removal, quality filtering, and personal information redaction, with workflows for text, images, video, and audio using NeMo Curator.
한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.
GPU-accelerated data curation for LLM training. Supports text/image/video/audio. Features fuzzy deduplication (16× faster), quality filtering (30+ heuristics), semantic deduplication, PII redaction, NSFW detection. Scales across GPUs with RAPIDS. Use for preparing high-quality training datasets, cleaning web data, or deduplicating large corpora.
설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.
NVIDIA's toolkit for preparing high-quality training data for LLMs.
Use NeMo Curator when:
Performance:
Use alternatives instead:
# Text curation (CUDA 12)
uv pip install "nemo-curator[text_cuda12]"
# All modalities
uv pip install "nemo-curator[all_cuda12]"
# CPU-only (slower)
uv pip install "nemo-curator[cpu]"
from nemo_curator import ScoreFilter, Modify
from nemo_curator.datasets import DocumentDataset
import pandas as pd
# Load data
df = pd.DataFrame({"text": ["Good document", "Bad doc", "Excellent text"]})
dataset = DocumentDataset(df)
# Quality filtering
def quality_score(doc):
return len(doc["text"].split()) > 5 # Filter short docs
filtered = ScoreFilter(quality_score)(dataset)
# Deduplication
from nemo_curator.modules import ExactDuplicates
deduped = ExactDuplicates()(filtered)
# Save
deduped.to_parquet("curated_data/")
from nemo_curator.filters import (
WordCountFilter,
RepeatedLinesFilter,
UrlRatioFilter,
NonAlphaNumericFilter
)
# Apply 30+ heuristic filters
from nemo_curator import ScoreFilter
# Word count filter
dataset = dataset.filter(WordCountFilter(min_words=50, max_words=100000))
# Remove repetitive content
dataset = dataset.filter(RepeatedLinesFilter(max_repeated_line_fraction=0.3))
# URL ratio filter
dataset = dataset.filter(UrlRatioFilter(max_url_ratio=0.2))
Exact deduplication:
from nemo_curator.modules import ExactDuplicates
# Remove exact duplicates
deduped = ExactDuplicates(id_field="id", text_field="text")(dataset)
Fuzzy deduplication (16× faster on GPU):
from nemo_curator.modules import FuzzyDuplicates
# MinHash + LSH deduplication
fuzzy_dedup = FuzzyDuplicates(
id_field="id",
text_field="text",
num_hashes=260, # MinHash parameters
num_buckets=20,
hash_method="md5"
)
deduped = fuzzy_dedup(dataset)
Semantic deduplication:
from nemo_curator.modules import SemanticDuplicates
# Embedding-based deduplication
semantic_dedup = SemanticDuplicates(
id_field="id",
text_field="text",
embedding_model="sentence-transformers/all-MiniLM-L6-v2",
threshold=0.8 # Cosine similarity threshold
)
deduped = semantic_dedup(dataset)
from nemo_curator.modules import Modify
from nemo_curator.modifiers import PIIRedactor
# Redact personally identifiable information
pii_redactor = PIIRedactor(
supported_entities=["EMAIL_ADDRESS", "PHONE_NUMBER", "PERSON", "LOCATION"],
anonymize_action="replace" # or "redact"
)
redacted = Modify(pii_redactor)(dataset)
from nemo_curator.classifiers import QualityClassifier
# Quality classification
quality_clf = QualityClassifier(
model_path="nvidia/quality-classifier-deberta",
batch_size=256,
device="cuda"
)
# Filter low-quality documents
high_quality = dataset.filter(lambda doc: quality_clf(doc["text"]) > 0.5)
| Operation | CPU (16 cores) | GPU (A100) | Speedup |
|---|---|---|---|
| Fuzzy dedup (8TB) | 120 hours | 7.5 hours | 16× |
| Exact dedup (1TB) | 8 hours | 0.5 hours | 16× |
| Quality filtering | 2 hours | 0.2 hours | 10× |
from nemo_curator import get_client
import dask_cuda
# Initialize GPU cluster
client = get_client(cluster_type="gpu", n_workers=8)
# Process with 8 GPUs
deduped = FuzzyDuplicates(...)(dataset)
from nemo_curator.image import (
AestheticFilter,
NSFWFilter,
CLIPEmbedder
)
# Aesthetic scoring
aesthetic_filter = AestheticFilter(threshold=5.0)
filtered_images = aesthetic_filter(image_dataset)
# NSFW detection
nsfw_filter = NSFWFilter(threshold=0.9)
safe_images = nsfw_filter(filtered_images)
# Generate CLIP embeddings
clip_embedder = CLIPEmbedder(model="openai/clip-vit-base-patch32")
image_embeddings = clip_embedder(safe_images)
from nemo_curator.video import (
SceneDetector,
ClipExtractor,
InternVideo2Embedder
)
# Detect scenes
scene_detector = SceneDetector(threshold=27.0)
scenes = scene_detector(video_dataset)
# Extract clips
clip_extractor = ClipExtractor(min_duration=2.0, max_duration=10.0)
clips = clip_extractor(scenes)
# Generate embeddings
video_embedder = InternVideo2Embedder()
video_embeddings = video_embedder(clips)
from nemo_curator.audio import (
ASRInference,
WERFilter,
DurationFilter
)
# ASR transcription
asr = ASRInference(model="nvidia/stt_en_fastconformer_hybrid_large_pc")
transcribed = asr(audio_dataset)
# Filter by WER (word error rate)
wer_filter = WERFilter(max_wer=0.3)
high_quality_audio = wer_filter(transcribed)
# Duration filtering
duration_filter = DurationFilter(min_duration=1.0, max_duration=30.0)
filtered_audio = duration_filter(high_quality_audio)
from nemo_curator import ScoreFilter, Modify
from nemo_curator.filters import *
from nemo_curator.modules import *
from nemo_curator.datasets import DocumentDataset
# Load Common Crawl data
dataset = DocumentDataset.read_parquet("common_crawl/*.parquet")
# Pipeline
pipeline = [
# 1. Quality filtering
WordCountFilter(min_words=100, max_words=50000),
RepeatedLinesFilter(max_repeated_line_fraction=0.2),
SymbolToWordRatioFilter(max_symbol_to_word_ratio=0.3),
UrlRatioFilter(max_url_ratio=0.3),
# 2. Language filtering
LanguageIdentificationFilter(target_languages=["en"]),
# 3. Deduplication
ExactDuplicates(id_field="id", text_field="text"),
FuzzyDuplicates(id_field="id", text_field="text", num_hashes=260),
# 4. PII redaction
PIIRedactor(),
# 5. NSFW filtering
NSFWClassifier(threshold=0.8)
]
# Execute
for stage in pipeline:
dataset = stage(dataset)
# Save
dataset.to_parquet("curated_common_crawl/")
from nemo_curator import get_client
from dask_cuda import LocalCUDACluster
# Multi-GPU cluster
cluster = LocalCUDACluster(n_workers=8)
client = get_client(cluster=cluster)
# Process large dataset
dataset = DocumentDataset.read_parquet("s3://large_dataset/*.parquet")
deduped = FuzzyDuplicates(...)(dataset)
# Cleanup
client.close()
cluster.close()
CPU-based curation (AWS c5.18xlarge × 10):
GPU-based curation (AWS p4d.24xlarge × 2):
Savings: 89% reduction ($3,828 saved)
Production deployments:
아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.
로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.
아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.
Create interactive 3D scenes for websites, including product views and scroll-driven animations. Covers choosing Three.js, React Three Fiber, or Spline, preparing lightweight models, showing loading progress, and adapting experiences for mobile devices.
한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.
Plan A/B tests that compare page or product changes, with clear hypotheses, sample sizes, success metrics, and traffic allocation. Get a structured test plan and guidance for interpreting results and deciding what to try next.
한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.
Prepare an academic CV for faculty, postdoctoral, or research applications. Organize publications, grants, teaching, and research experience, with section order and formatting adapted to the role and academic field.
한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.
Audit and improve web accessibility using WCAG 2.1 guidelines. Covers image descriptions, color contrast, keyboard controls, forms, and screen reader support, with code examples and automated and manual checks to guide fixes.
한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.
Audit websites for accessibility barriers and improve image descriptions, color contrast, forms, keyboard navigation, and screen reader support. Provides code examples and automated and manual testing guidance for work toward WCAG 2.1 AA or AAA accessibility standards.
한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.
Design autonomous agents for Claude Code plugins, including when they should run, what tools they may use, and how they should respond. Provides configuration templates, prompt guidance, and validation steps for creating agent files.
한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.