LEM/training/lem
Snider d998dd252b chore: zstd compress all JSONL training data (145MB → 29MB)
Compress 108 JSONL files with zstd level 3 and remove originals.
Add training/.gitignore to exclude decompressed .jsonl files.
Remove broken symlinks to previously deleted p4 duplicates.

Co-Authored-By: Virgil <virgil@lethean.io>
2026-02-28 12:17:38 +00:00
..
books feat(distill): add Metal memory limit config fields 2026-02-22 17:59:11 +00:00
composure chore: zstd compress all JSONL training data (145MB → 29MB) 2026-02-28 12:17:38 +00:00
creative feat: add training data to repo + make paths repo-relative 2026-02-21 23:49:12 +00:00
deepseek-western-soak chore: zstd compress all JSONL training data (145MB → 29MB) 2026-02-28 12:17:38 +00:00
ethics chore: remove 13 duplicate/empty JSONL files from training data 2026-02-28 12:16:53 +00:00
eval feat: add training data to repo + make paths repo-relative 2026-02-21 23:49:12 +00:00
model/gemma3 chore: zstd compress all JSONL training data (145MB → 29MB) 2026-02-28 12:17:38 +00:00
russian-bridge chore: zstd compress all JSONL training data (145MB → 29MB) 2026-02-28 12:17:38 +00:00
tension feat: add training data to repo + make paths repo-relative 2026-02-21 23:49:12 +00:00
validation chore: zstd compress all JSONL training data (145MB → 29MB) 2026-02-28 12:17:38 +00:00
western-fresh chore: zstd compress all JSONL training data (145MB → 29MB) 2026-02-28 12:17:38 +00:00
zen chore: zstd compress all JSONL training data (145MB → 29MB) 2026-02-28 12:17:38 +00:00