5 days ago
-
release v0.2 (#2)
- release v0.2 (08d51b30e6a2b98cd61d2aab6a609572d8348206)
- release v0.2 (2d6c7a1dffd0cafa8500739a340c90aab9d2baa1)
- release v0.2 (de50a4b6fa946656214433e17a044584e48cc39a)
26 days ago
-
presubmit: align has_structure to SMILES (6); publish CC-BY-ND other (5); card cc-by-nc-4.0
Quick Sci Data pre-submit fixes on fc238e3 lineage:
- Flip has_structure=true on 6 commercial/resolved/train rows with nonempty SMILES (Table 1).
- Publish data/irexp_other.jsonl.gz (5 CC-BY-ND) as companion; commercial DoR stays n=88,519.
- Dataset card YAML adds cc-by-nc-4.0; note that per-record NC policy licences still apply.
SA/NC JSONL bytes unchanged. Empty/unknown still off Hub. -
release: reapply band-count filter and dedup integer bands
Commercial DoR 88545→88519 (−26 rows with <3 bands after thousands-separator reparse). Order-preserving unique of integer cm−1 within lists (1432 commercial, 319 NC). Rebuilt resolved=commercial∩smiles (29255) and train_no_bench (29109). Densified ir_refetch_* bools. ND (5) not included. Updated README/NOTICE counts.
