Add disjoint 10K ligand example

.gitattributes

@@ -72,3 +72,4 @@ crossdock/crossdocked_v1.0_rmsd_1_to_2_all_heavy_pocket_train.lmdb filter=lfs di
 crossdock/crossdocked_v1.0_rmsd_1_to_2_all_heavy_pocket_valid_3000.lmdb filter=lfs diff=lfs merge=lfs -text
 pockets_center_frame_cartesian/train.lmdb filter=lfs diff=lfs merge=lfs -text
 pockets_center_frame_cartesian/valid.lmdb filter=lfs diff=lfs merge=lfs -text
+ligands/example_10000.lmdb filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,5 +1,5 @@
 ---
-pretty_name: MolWeaver Molecular Data
+pretty_name: MolWeaver Ligands 100M
 task_categories:
   - text-generation
   - feature-extraction
@@ -9,47 +9,15 @@ tags:
   - selfies
   - conformers
   - rdkit
-  - protein-ligand
 ---
 
-# MolWeaver Molecular Data
+# MolWeaver Ligands 100M
 
-This repository contains the molecular data used by MolWeaver: a
-100-million-record ligand corpus, preprocessed CrossDocked protein-ligand
-examples, and protein-pocket train/validation data in a centered Cartesian
-frame. The dataset repository ID is `shaipranesh/mol_data`.
+This dataset contains 100,000,000 globally unique heavy-atom molecular
+records split across five LMDB shards. Each shard contains 20,000,000
+records and all records are assigned to the training split.
 
-## Dataset layout
-
-```text
-ligands/
-  shard_1.lmdb
-  shard_1_metadata.json
-  ...
-  shard_5.lmdb
-  shard_5_metadata.json
-crossdock/
-  crossdocked_v1.0_rmsd_1_to_2_all_heavy_pocket_train.lmdb
-  crossdocked_v1.0_rmsd_1_to_2_all_heavy_pocket_train.lmdb.{keys.pkl,meta.json,split.json}
-  crossdocked_v1.0_rmsd_1_to_2_all_heavy_pocket_valid_3000.lmdb
-  crossdocked_v1.0_rmsd_1_to_2_all_heavy_pocket_valid_3000.lmdb.{keys.pkl,meta.json,split.json}
-pockets_center_frame_cartesian/
-  train.lmdb
-  train.lmdb.{keys.pkl,meta.json}
-  valid.lmdb
-  valid.lmdb.{keys.pkl,meta.json}
-```
-
-The sidecar files record key indexes, preprocessing metadata, and split
-information where applicable.
-
-## Ligand corpus
-
-The `ligands/` directory contains 100,000,000 globally unique heavy-atom
-molecular records split across five LMDB shards. Each shard contains
-20,000,000 records, and all records are assigned to the training split.
-
-### Ligand record schema
+## Record schema
 
 Each numeric LMDB key contains a pickled Python dictionary with:
 
@@ -66,12 +34,30 @@ Each numeric LMDB key contains a pickled Python dictionary with:
 Explicit hydrogen atoms and hydrogen coordinates are not included. Standard
 implicit hydrogens are used by RDKit when calculating molecular properties.
 
-Every ligand shard stores numeric keys `b"0"` through `b"19999999"` and a
+## Files
+
+```text
+ligands/shard_1.lmdb
+ligands/shard_1_metadata.json
+...
+ligands/shard_5.lmdb
+ligands/shard_5_metadata.json
+ligands/example_10000.lmdb
+ligands/example_10000_metadata.json
+```
+
+Every `shard_*.lmdb` stores numeric keys `b"0"` through `b"19999999"` and a
 pickled `b"length"` value equal to `20_000_000`.
 
 The local deduplication registries used during generation are not uploaded;
 they are not needed to train from the finalized records.
 
+`example_10000.lmdb` uses the same record schema and contains 10,000 unique
+SELFIES generated from unused source slice 6. All 10,000 entries were checked
+against the finalized training registries and have no overlap with the 100M
+training records. Its numeric keys are `b"0"` through `b"9999"`, and its
+pickled `b"length"` value is `10_000`.
+
 ## Loading
 
 ```python

ligands/example_10000.lmdb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46c306e68c60c8ebb9a170773344b809d87f5a7e03879e8dbf53065517d6d0c3
+size 43323392

ligands/example_10000_metadata.json

@@ -0,0 +1,50 @@
+{
+  "conformers_per_molecule": 10,
+  "content_sha256": "945947c569420d35bfeee9c460370a2b316053d2c41e44d90812de1c4b995d2c",
+  "coordinate_space": "original RDKit Cartesian coordinates; atom-axis permutation only",
+  "disjoint_from_training": true,
+  "elapsed_seconds": 102.98119640350342,
+  "explicit_hydrogens": false,
+  "lmdb_bytes": 43323392,
+  "normalization_skipped_by_reason": {},
+  "property_keys": [
+    "qed",
+    "sa_score",
+    "molecular_weight",
+    "mol_log_p",
+    "tpsa"
+  ],
+  "record_keys": [
+    "atoms",
+    "coordinates",
+    "mol_log_p",
+    "molecular_weight",
+    "qed",
+    "sa_score",
+    "smi",
+    "tpsa"
+  ],
+  "records": 10000,
+  "sample_duplicate_rejected": 0,
+  "source_file_errors": {},
+  "source_raw_slice": 6,
+  "source_records_processed": 10112,
+  "training_overlap_rejected": 0,
+  "training_records_checked": 100000000,
+  "training_registry_files_checked": [
+    "/scratch/shai/data_20m_selfies_original_cartesian_shards/selfies_seen.lmdb",
+    "/scratch/shai/data_20m_selfies_original_cartesian_shards/shard_1/selfies_seen.lmdb",
+    "/scratch/shai/data_20m_selfies_original_cartesian_shards/shard_2/selfies_seen.lmdb",
+    "/scratch/shai/data_20m_selfies_original_cartesian_shards/shard_3/selfies_seen.lmdb",
+    "/scratch/shai/data_20m_selfies_original_cartesian_shards/shard_4/selfies_seen.lmdb",
+    "/scratch/shai/data_20m_selfies_original_cartesian_shards/shard_5/selfies_seen.lmdb"
+  ],
+  "unique_selfies": 10000,
+  "validation": {
+    "all_coordinates_finite_float32_checked": true,
+    "all_properties_finite_checked": true,
+    "all_records_schema_checked": true,
+    "all_selfies_training_disjoint_checked": true,
+    "all_selfies_unique_checked": true
+  }
+}