Create a single eval set for every difficulty

README.md

@@ -16,6 +16,12 @@ dataset_info:
     - name: solution_pretty
       dtype: string
 configs:
+- config_name: eval
+  default: true
+  data_files:
+    - split: train
+      path:
+        - "eval.parquet"
 - config_name: kids
   default: true
   data_files:
@@ -109,3 +115,9 @@ The dataset is organized into multiple parquet files grouped by difficulty thres
 | `medium`      | 36            | 90000              | Medium difficulty puzzles                  |
 | `hard`        | 27            | 90000              | Hard puzzles                               |
 | `insane`      | 17            | 100000             | Insane difficulty puzzles                  |
+
+Lastly, there's `eval`, which contains the first 200 puzzles of every single
+difficulty from 80 already filled squares to 17 already filled squares. This is
+sampling of a smaller number of puzzles is much more manageable for running
+holistic evals, compared to running 10,000 attempts on every single difficulty
+level.

eval.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9de15e2f191d79efd5a72f722740415aca34d91c2178ab6bed9d0bf7019e2c8d
+size 4308495