Training products derived from islamlab/islamic-sciences

README.md

@@ -58,16 +58,16 @@ model.
 
 | config | split | rows | what it is |
 |---|---|---:|---|
-| `cpt` | train | 775,173 | Apparatus-free, deduplicated text with a provenance header — 3.18B characters |
-| `retrieval` | train | 901,285 | A section heading as the query, a passage from that section as the positive, and hard negatives drawn from elsewhere in the same book |
-| `attribution` | train | 130,200 | Citation questions with exact answers |
-| `eval_retrieval` | test | 26,931 | Same shape, held-out works only |
-| `eval_attribution` | test | 33,150 | Same shape, held-out works only |
+| `cpt` | train | 780,154 | Apparatus-free, deduplicated text with a provenance header — 3.21B characters |
+| `retrieval` | train | 906,693 | A section heading as the query, a passage from that section as the positive, and hard negatives drawn from elsewhere in the same book |
+| `attribution` | train | 131,025 | Citation questions with exact answers |
+| `eval_retrieval` | test | 26,933 | Same shape, held-out works only |
+| `eval_attribution` | test | 32,575 | Same shape, held-out works only |
 
 ## The held-out split
 
-The holdout is at the level of the **work**, not the row: 114 of the
-3,862 works are assigned to evaluation by a hash of their identifier,
+The holdout is at the level of the **work**, not the row: 118 of the
+4,022 works are assigned to evaluation by a hash of their identifier,
 and no line of their text appears in `cpt`, `retrieval` or `attribution`. A
 model trained on the train configs has therefore never read the books it is
 evaluated on. The assignment is deterministic, so it is the same for everyone.
@@ -76,11 +76,11 @@ evaluated on. The assignment is deterministic, so it is the same for everyone.
 
 | task | question | rows (train) |
 |---|---|---:|
-| `author` | من مؤلف هذا النص؟ | 26,040 |
-| `title` | من أي كتاب أُخذ هذا النص؟ | 26,040 |
-| `science` | في أي علم يُصنَّف هذا الكتاب؟ | 26,040 |
-| `century` | في أي قرن هجري توفي المؤلف؟ | 26,040 |
-| `locus` | في أي جزء وصفحة يقع هذا النص؟ | 26,040 |
+| `author` | من مؤلف هذا النص؟ | 26,205 |
+| `title` | من أي كتاب أُخذ هذا النص؟ | 26,205 |
+| `science` | في أي علم يُصنَّف هذا الكتاب؟ | 26,205 |
+| `century` | في أي قرن هجري توفي المؤلف؟ | 26,205 |
+| `locus` | في أي جزء وصفحة يقع هذا النص؟ | 26,205 |
 
 `messages` is a JSON-encoded chat turn — parse it with `json.loads`. `answer`
 holds the exact expected string. The `locus` task deliberately includes the
@@ -107,7 +107,7 @@ negatives = json.loads(row["negatives"])
 
 | Dataset | What it holds |
 |---|---|
-| [`islamlab/islamic-sciences`](https://huggingface.co/datasets/islamlab/islamic-sciences) | The corpus itself — eight sciences, 3,862 works |
+| [`islamlab/islamic-sciences`](https://huggingface.co/datasets/islamlab/islamic-sciences) | The corpus itself — eight sciences, 4,022 works |
 | **`islamlab/islamic-sciences-training`** | The training products above |
 | [`quranlab/*`](https://huggingface.co/quranlab) · [`risaleinur/*`](https://huggingface.co/risaleinur) | Qur'an and hadith; the Risale-i Nur corpus |
 

attribution/train-00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bb27b56db32fcc40d39bc22b1e88b60aa5c17a1aa5c604820e76d227d21dfc7d
-size 12985095
+oid sha256:e513dff6357983b971effaf2da5ef88db0dc93e789c8ccfdf9dab874fa6cecbd
+size 13080544

cpt/train-00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eeee03e0ba3f197cbf22eb9ddb3aad311e0d4ac36572fbd75d65ded734d6507a
-size 1256098991
+oid sha256:af707e191550686c174e89411f945c9862ab921c7701f71cd27f7b32d1ff82b4
+size 1267953963

eval_attribution/test-00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17cf5be900b30269c9bce6d37729d308e7affea333730dde13442821e941d9ba
-size 3319614
+oid sha256:3823079e49d3efb5e495a28a572c3aff00076ce72fb841bf61f379f97abb8110
+size 3269040

eval_retrieval/test-00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f7b3f5d681dba1c06cdd5323a5ba5e7a6cd4dc31a2475c48cac51efe9ae55732
-size 44525970
+oid sha256:5721ddcd962501f6e2f918e5c783dfed69f17f3d8e24c7f76203d19c65d2b347
+size 44553244

retrieval/train-00000.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:93da3d3229856e6d20ae73fb3ee46b1d7d08531f78409101a6c1cd1139eaf35c
-size 1250269477
+oid sha256:6dc173f79e7335ed0bb138c42555f3176ae89f763cae19fb276cb3141b7f2553
+size 1259284487