5GB: update README for sharded parquet 50M

README.md

@@ -7,7 +7,7 @@ language:
 - en
 pretty_name: Arithmetic-Reasoning
 size_categories:
-- 100K<n<1M
+- 10M<n<100M
 tags:
 - math
 - reasoning
@@ -19,11 +19,11 @@ configs:
 - config_name: default
   data_files:
   - split: train
-    path: parquet/train.parquet
+    path: parquet/train-*.parquet
   - split: validation
-    path: parquet/val.parquet
+    path: parquet/validation-*.parquet
   - split: test
-    path: parquet/test.parquet
+    path: parquet/test-*.parquet
 ---
 
 ![Banner](best_3440x1440.png)
@@ -32,7 +32,7 @@ configs:
 
 > A high-quality synthetic arithmetic and elementary mathematics reasoning dataset for training and evaluating small language models - not an "ultimate math" claim, but a clean, verified, tiered reasoning dataset where every answer is programmatically checked.
 
-This dataset was built to train **100M-ish models** that benefit disproportionately from clean, unambiguous examples. At 150K examples (135K train / 7.5K val / 7.5K test) it is designed as a *reasoning* dataset: every sample has an instruction, a step-by-step reasoning trace, and a final answer, all verified to be mathematically correct.
+This dataset was built to train **100M-ish models** that benefit disproportionately from clean, unambiguous examples. At ~50M examples (~45M train / 2.5M val / 2.5M test, ~5GB parquet) it is designed as a *reasoning* dataset: every sample has an instruction, a step-by-step reasoning trace, and a final answer, all verified to be mathematically correct.
 
 ### Why this dataset?
 
@@ -56,7 +56,7 @@ We cover 18+ operation families. Each `instruction` is sampled from 20+ phrasing
 | 10 | **Geometry** | `What is the area of a rectangle 884 by 752?` | `884*752=664768` | `664768` |
 | | *+ 8 more:* `ratios, powers/roots, decimals, halving/doubling, sequences, multi-step, hard arithmetic, quantitative reasoning` - see `operation` field for full list. | | | |
 
-Full per-operation counts in `150K`:
+Full per-operation counts in `50M`:
 `word problems 13.6% | multi-step 10.7% | ratios 9.7% | hard arithmetic 5.7% | algebra 7.4% | fractions 4.6% | sequences 5.3% | geometry 5.1% | ...` (see `Per-operation` in generation log)
 
 
@@ -80,27 +80,12 @@ from huggingface_hub import snapshot_download
 snapshot_download("SagheerLab/Arithmetic-Reasoning", repo_type="dataset", local_dir="./arithmetic-reasoning")
 ```
 
-Both `JSON` and `Parquet` are provided (`parquet/train.parquet` + `json/train.json`) - parquet is the default Viewer config, JSON via `data_files` as below.
 
-```python
-# Parquet (default) - 12.6 MB
-from datasets import load_dataset
-ds = load_dataset("SagheerLab/Arithmetic-Reasoning")  # loads parquet/train.parquet
-
-# JSON - 42 MB - JSON Lines, load via json builder
-from datasets import load_dataset
-ds_json = load_dataset("SagheerLab/Arithmetic-Reasoning", data_files={
-    "train": "json/train.json",
-    "validation": "json/val.json",
-    "test": "json/test.json"
-})
-# or directly: load_dataset("json", data_files="hf://datasets/SagheerLab/Arithmetic-Reasoning/json/train.json")
-```
 
 
 # Dataset Summary
 
-- **Size:** 150,000 examples (135K train / 7.5K val / 7.5K test), 100% programmatically verified
+- **Size:** ~50M examples (~45M train / 2.5M val / 2.5M test, ~5GB parquet), 100% programmatically verified
 - **Generation:** deterministic `src/generate_data_v4.py` with `NAMES 120 + OBJECTS 150 + VERBS 100` for lexical variation, de-duplicated (<1% dupes)
 - **Tiers:** `easy 30% (1-2 digits, 1 step) | intermediate 35% (3-4 digits, 2-3 steps) | hard 25% (5-6 digits, 3-8 steps) | very hard 10% (competition/GSM8K-style)`
 - **License:** Apache 2.0 - see [Licensing](#licensing-information)
@@ -134,11 +119,11 @@ ds_json = load_dataset("SagheerLab/Arithmetic-Reasoning", data_files={
 ### Data Splits
 | Split | Size | Path |
 |-------|------|------|
-| train | 135,000 | `parquet/train.parquet` + `json/train.json` |
-| validation | 7,500 | `parquet/val.parquet` + `json/val.json` |
-| test | 7,500 | `parquet/test.parquet` + `json/test.json` |
+| train | ~45M | `parquet/train-*.parquet` (sharded, ~4.5GB) |
+| validation | ~2.5M | `parquet/validation-*.parquet` |
+| test | ~2.5M | `parquet/test-*.parquet` |
 
-All splits are IID from same generator, `seed 42`, `balanced-sizes`, `max_digits 6`.
+All splits are IID from same generator, `seed 42`, `balanced-sizes`, `max_digits 6`, sharded for Viewer (`row_group_size=10000`, `write_page_index=True`).
 
 # Dataset Creation
 
@@ -164,7 +149,7 @@ Generation uses:
    # else drop & regenerate (seen set ensures <1% dupes)
    ```
 3. Deduplicate `instruction` strings via `seen` set
-4. Shuffle and split 90/5/5, write `JSONL` + `Parquet`, copy `best_3440x1440.png` banner
+4. Shuffle and split 90/5/5, write `Parquet` shards (`row_group_size=10000`, `write_page_index=True`), copy `best_3440x1440.png` banner
 
 ### Personal and Sensitive Information
 None - all synthetic. Names are sampled from common first names, no real PII.