Switch all files to Parquet so every config (incl. eval) works in the viewer; same rows

README.md

@@ -21,11 +21,11 @@ configs:
   default: true
   data_files:
   - split: train
-    path: data/sft.jsonl
+    path: data/sft.parquet
 - config_name: all
   data_files:
   - split: train
-    path: data/traces_all.jsonl
+    path: data/traces_all.parquet
 - config_name: eval
   data_files:
   - split: test
@@ -37,7 +37,7 @@ configs:
 - config_name: eval_problems
   data_files:
   - split: test
-    path: eval/problems.jsonl
+    path: eval/problems.parquet
 ---
 
 # Beam Reasoning Traces

data/{sft.jsonl → sft.parquet}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0c12a4daf400c1928129f6ed13fc5d12bce11aed3ca76de0611e812853169331
-size 308228420
+oid sha256:0714dd7b11d57856c92aac693345bf2611fea51e1a7cb53e3556e31ad1509e74
+size 109643115

data/{traces_all.jsonl → traces_all.parquet}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:84be386e189cf26a9a10022fff593b094bf36cf41ae4afdb518a7c4be1b4aa5f
-size 921089884
+oid sha256:08f97680eee011f8ec7ef054e97302be77305d872638a6b81424a22b1da892ca
+size 283152580

eval/problems.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9911654ad54f9ba53ad8b4a74d2de026e881e7c8b3ff1135ba8324f29c33a50f
+size 136933