lewald/typst
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
No description
1 commit 1 branch 0 tags 319 KiB
  • Typst 90.1%
  • Python 9.9%
Find a file
Exact
lewald 97fd5e9fe0 Initial commit
2026-05-27 14:45:22 +02:00
data Initial commit 2026-05-27 14:45:22 +02:00
data_backup Initial commit 2026-05-27 14:45:22 +02:00
add_accessibility.py Initial commit 2026-05-27 14:45:22 +02:00
add_missing_features.py Initial commit 2026-05-27 14:45:22 +02:00
add_more_features.py Initial commit 2026-05-27 14:45:22 +02:00
AGENTS.md Initial commit 2026-05-27 14:45:22 +02:00
analyze_data.py Initial commit 2026-05-27 14:45:22 +02:00
analyze_gaps.py Initial commit 2026-05-27 14:45:22 +02:00
analyze_true_duplicates.py Initial commit 2026-05-27 14:45:22 +02:00
balance_dataset.py Initial commit 2026-05-27 14:45:22 +02:00
check_169.py Initial commit 2026-05-27 14:45:22 +02:00
check_duplicates.py Initial commit 2026-05-27 14:45:22 +02:00
compare_versions.py Initial commit 2026-05-27 14:45:22 +02:00
expand_short.py Initial commit 2026-05-27 14:45:22 +02:00
final_report.py Initial commit 2026-05-27 14:45:22 +02:00
find_dupes.py Initial commit 2026-05-27 14:45:22 +02:00
find_duplicates.py Initial commit 2026-05-27 14:45:22 +02:00
find_remaining.py Initial commit 2026-05-27 14:45:22 +02:00
fix_169.py Initial commit 2026-05-27 14:45:22 +02:00
fix_data.py Initial commit 2026-05-27 14:45:22 +02:00
fix_remaining.py Initial commit 2026-05-27 14:45:22 +02:00
JL-SPARK.md Initial commit 2026-05-27 14:45:22 +02:00
README.md Initial commit 2026-05-27 14:45:22 +02:00
requirements.txt Initial commit 2026-05-27 14:45:22 +02:00
SESSION_SUMMARY.md Initial commit 2026-05-27 14:45:22 +02:00
view_access.py Initial commit 2026-05-27 14:45:22 +02:00
view_others.py Initial commit 2026-05-27 14:45:22 +02:00

README.md

Typst-Training-Projekt

Dieses Projekt dient dazu, ein kleines LLM (z.B. Gemma-2-2b-it oder Llama-3.2-1B) mit Unsloth auf Typst zu fine-tunen.

Projektstruktur

typst-training/
├── data/          # Trainingsdaten (JSONL)
├── src/           # Skripte (Daten-Extraktion, Training)
├── models/        # Gespeicherte Modelle
└── output/        # Trainings-Output

Schnellstart

  1. Installiere Dependencies: pip install -r requirements.txt
  2. Trainiere: python src/train_typst_model.py

Hinweis: Das Dataset (data/typst_train.jsonl) ist bereits fertig (175 Samples). Die Extraktions-Skripte wurden nicht benötigt.

Anforderungen

  • GPU mit mindestens 8GB VRAM (4-bit Quantisierung empfohlen)
  • Python 3.10+
  • PyTorch mit CUDA-Support