Datasets:
|
Download cards/README.de.md from pathcosmos/korean-pretrain-code: direct link, hf CLI and curl.
- Browser
- Download file 10.8 kB
-
https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/resolve/main/cards/README.de.md
- Command line
-
hf download hf://datasets/pathcosmos/korean-pretrain-code/cards/README.de.md
-
curl -L -o README.de.md https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/resolve/main/cards/README.de.md
10.8 kB
| # korean-pretrain-code — Deutsch | |
| [한국어](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ko.md) · [English](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.en.md) · [简体中文](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.zh-CN.md) · [繁體中文](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.zh-TW.md) · [日本語](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ja.md) · [Français](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.fr.md) · [Deutsch](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.de.md) · [Español](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.es.md) · [Português (Brasil)](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.pt-BR.md) · [हिन्दी](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.hi.md) · [العربية](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ar.md) · [עברית](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.he.md) · [Русский](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ru.md) | |
| ## Überblick | |
| Dieses Repository enthält die bereinigten und aufgeteilten Daten der unten angegebenen einzelnen Quelle. Dieselben Dateien sind auch im zugehörigen Ordner des Gesamtkorpus verfügbar. Fixierte Datenversion und Dokumentationsrevision sind getrennt. | |
| Übersetzungssprachen bedeuten keine entsprechenden Trainingsdaten. Hauptsprachen sind Koreanisch und Englisch; Web-, Mathematik- und Codeinhalte können weitere Sprachen enthalten. | |
| ## Inhalt und Umfang | |
| | | | | |
| |---|---:| | |
| | Quelle | github_code_v1 | | |
| | Dokumente | 2,148,459 | | |
| | Trainingsdokumente | 2,125,938 | | |
| | Evaluationsdokumente | 22,521 | | |
| | Textbytes | 3,584,084,955 | | |
| | | | | |
| |---|---:| | |
| | tokens | 4,006,737,196 | | |
| | eval tokens | 42,420,481 | | |
| | revision | v1.0.0 | | |
| | upstream | [codeparrot/github-code](https://huggingface.co/datasets/codeparrot/github-code/blob/b5661e6b17396364b2bcf8e68977b0d28e1ebd19/README.md) | | |
| | upstream license | other | | |
| | provenance_status | `repo_path_row_preserved_git_commit_unknown` | | |
| ## Verarbeitung und Aufteilung | |
| Webverarbeitung bereinigt Leerraum, nutzt Längen-/Wiederholungs-/Sprachheuristiken und maskiert ausgewählte personenbezogene Muster. Der anfängliche shardinterne Hash unterscheidet sich vom abschließenden NFC-SHA256. Code behält Großschreibung, Einrückung und ursprüngliches Unicode; Pfade, Lizenzangaben, generierter Code und ausgewählte Geheimnismuster werden geprüft. Verbundene Gruppen gleicher Texte, NFC-äquivalenter Webtexte und Code-Repositories werden train/eval zugeteilt. | |
| ## Verwendung | |
| Die Befehle im Repository-Hauptverzeichnis ausführen. Neue Ausgabe-/Downloadordner verwenden und Umfang, Größe sowie fixierte Revision prüfen. Das synthetische CPU-Beispiel lädt keine externen Daten. Benötigte Dateien aus separaten Veröffentlichungen wählen und Größe/SHA256 anhand CHECKSUMS.json oder Quellmanifests prüfen. | |
| ```bash | |
| python -m pip install "pyarrow>=19,<20" "huggingface_hub>=0.34,<1" "tokenizers>=0.21,<0.23" "transformers==4.55.2" "requests>=2,<3" | |
| ``` | |
| ```python | |
| from huggingface_hub import snapshot_download | |
| root = snapshot_download('pathcosmos/korean-pretrain-code', repo_type="dataset", | |
| revision='v1.0.0', allow_patterns=['text/*.parquet', 'source.json', 'manifest.json', 'release.json', 'LICENSE-SCOPE.md'], | |
| local_dir='korean-pretrain-code', token=False) | |
| ``` | |
| ```python | |
| from pathlib import Path | |
| import subprocess, sys | |
| from huggingface_hub import snapshot_download | |
| reader = snapshot_download("pathcosmos/korean-pretrain-corpus", repo_type="dataset", revision="v1.1.0", | |
| allow_patterns=["examples/*"], local_dir="corpus-reader", token=False) | |
| subprocess.run([sys.executable, str(Path(reader) / "examples/stream_documents.py"), | |
| "--repo", "pathcosmos/korean-pretrain-corpus", "--revision", "v1.1.0", "--source", "github_code_v1", | |
| "--split", "train", "--limit", "10"], check=True) | |
| ``` | |
| [examples/stream_documents.py](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/blob/main/examples/stream_documents.py) · [manifest.json](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/manifest.json) | |
| ## Modell und Training | |
| Der reine PyTorch-Dense-Decoder verwendet RMSNorm, RoPE, GQA und SwiGLU. Der reale3B-Pfad nutzt FSDP2,8GPUs und eager: Kontext4096, Microbatch4, global2,097,152Tokens, Spitzen-LR3e-4 und Ziel57,000Schritte. Das CPU-Microbeispiel unterscheidet sich vom verteilten Training. Exakte Wiederaufnahme verlangt Modell, Optimizer, RNG und Eingabe-/Konfigurationsverträge. | |
| ## Optimierung und Benchmarks | |
| Verglichen werden Worker, Threads, Batches, SQLite-Cache, RAM, lokal/NFS, eager/compile, Präzision und GPU-Speicher. Gesamtzeit, Durchsatz, Speicherspitze und Inhalts-/numerische Gleichwertigkeit begründen Annahme oder Ablehnung:47 Kampagnen,99 Ergebnis-JSONs. Geschwindigkeit rechtfertigt keinen numerisch fehlerhaften Kandidaten; kleine Stichproben beweisen kein Gesamtoptimum. | |
| ## Validierung | |
| Die GH-Quellversion bestand284 CPU-Tests bei9 Überspringungen; frische GitHub-CPU-CI bestand258 bei35 Überspringungen. Gründe sind fehlende Originalprotokolle oder ein zweites Dateisystem. Der synthetische Quickstart nutzt120Dokumente, train100/eval20, Vokabular272 und108,864Parameter und prüft exakte Modell-/Optimizer-Wiederaufnahme. Datenhashes, Splits und Readerprüfungen sind keine vollständige unabhängige3B-Reproduktion. | |
| [release.json](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/release.json) · [source.json](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/source.json) | |
| ## Quellen und Lizenzen | |
| Originaler Projektcode steht unter MIT, eigener Erklärungstext unter CC BY4.0. Dies ersetzt nicht die Rechte an Quelldaten. source.json, LICENSE-SCOPE.md und Herkunftsdokumente nennen Bedingungen, Zuschreibungen, Hinweise und offene Punkte. Eine unbeschränkte kommerzielle Weitergabe aller Originaltexte wird nicht behauptet. | |
| [codeparrot/github-code](https://huggingface.co/datasets/codeparrot/github-code/blob/b5661e6b17396364b2bcf8e68977b0d28e1ebd19/README.md) · [source.json](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/source.json) · [LICENSE-SCOPE.md](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/LICENSE-SCOPE.md) | |
| ## Grenzen | |
| Globale Fast-Duplikatentfernung, vollständige PII-/Geheimnisentfernung, kontaminationsfreie Bewertung und finale Qualität werden nicht garantiert. Haupttraining lief beim Snapshot; SFT, Export und offizielle Sprachbenchmarks sind nicht abgeschlossen.11 historische Codeverweise/9einzigartige Hashes haben keinen wiedergefundenen Snapshot. Einige Betriebswerkzeuge benötigen Originalmanifests. | |
| ## Verwandte Repositories | |
| [GitHub](https://github.com/pathcosmos/korean-dense-llm/blob/main/cards/README.de.md) · [README](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code) | |
| [GH / Verwendung](https://github.com/pathcosmos/korean-dense-llm/blob/main/docs/PUBLIC_GUIDE.md) · [GH / Optimierung und Benchmarks](https://github.com/pathcosmos/korean-dense-llm/blob/main/experiments/README.md) · [GH / Validierung](https://github.com/pathcosmos/korean-dense-llm/blob/main/docs/PUBLIC_VALIDATION.md) | |
| - [korean-pretrain-corpus](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/blob/main/cards/README.de.md) | |
| - [korean-pretrain-training-data](https://huggingface.co/datasets/pathcosmos/korean-pretrain-training-data/blob/main/cards/README.de.md) | |
| - [korean-pretrain-wikipedia](https://huggingface.co/datasets/pathcosmos/korean-pretrain-wikipedia/blob/main/cards/README.de.md) | |
| - [korean-pretrain-fineweb2](https://huggingface.co/datasets/pathcosmos/korean-pretrain-fineweb2/blob/main/cards/README.de.md) | |
| - [korean-pretrain-hplt2](https://huggingface.co/datasets/pathcosmos/korean-pretrain-hplt2/blob/main/cards/README.de.md) | |
| - [korean-pretrain-fineweb-edu](https://huggingface.co/datasets/pathcosmos/korean-pretrain-fineweb-edu/blob/main/cards/README.de.md) | |
| - [korean-pretrain-math](https://huggingface.co/datasets/pathcosmos/korean-pretrain-math/blob/main/cards/README.de.md) | |
| - [korean-pretrain-code](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.de.md) | |
| - [wiki_ko/](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/tree/main/wiki_ko) | |
| - [fineweb2_ko/](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/tree/main/fineweb2_ko) | |
| - [hplt_ko/](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/tree/main/hplt_ko) | |
| - [fineweb_edu/](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/tree/main/fineweb_edu) | |
| - [owm/](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/tree/main/owm) | |
| - [github_code_v1/](https://huggingface.co/datasets/pathcosmos/korean-pretrain-corpus/tree/main/github_code_v1) | |
| [한국어](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ko.md) · [English](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.en.md) · [简体中文](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.zh-CN.md) · [繁體中文](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.zh-TW.md) · [日本語](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ja.md) · [Français](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.fr.md) · [Deutsch](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.de.md) · [Español](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.es.md) · [Português (Brasil)](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.pt-BR.md) · [हिन्दी](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.hi.md) · [العربية](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ar.md) · [עברית](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.he.md) · [Русский](https://huggingface.co/datasets/pathcosmos/korean-pretrain-code/blob/main/cards/README.ru.md) | |