3 Commits (f5f8939a5c97c97ffc04282296f669bfb5cf25bb)

Author	SHA1	Message	Date
songhyeonsu	f5f8939a5c	Wire up wandb logging for Step1 training ... - Global.use_wandb: True + top-level wandb.project=kr_lp_pgnet - Add wandb to setup_server.sh pip install list User must run `docker exec -it kr_lp_pgnet wandb login` once before training so the API key lands in /root/.netrc inside the container. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>	1 month ago
songhyeonsu	82c046522e	Add Step1 training runner and lower default epochs to 50 ... - run_step1.sh: symlinks /workspace/train_data into PaddleOCR, runs tools/train.py with the step1 pretrain checkpoint, supports DRY_RUN=1 for quick smoke test and EPOCHS=N override - epoch_num: 200 -> 50 (matches the 50k synthetic budget) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>	1 month ago
songhyeonsu	3e5f823dd5	Add Korean LP dictionary and PGNet config ... - dict/kr_lp_dict.txt: 67 chars covering 4 plate types (10 digits + 40 usage hangul + 17 region hangul, dedup) - configs/kr_lp_pgnet.yml: PGNet config tuned for Korean LP (pad_num=67, max_text_length=10, valid_set=partvgg, infer_visual_type=CN) - setup_server.sh: symlink dict and config into PaddleOCR tree Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>	1 month ago