chore(scripts): drop ad-hoc verify harnesses, move dataset converters to datasets/

scripts/convert_daily_omni_dataset_format_to_unirl.py → datasets/daily_omni_av/convert_daily_omni_dataset_format_to_unirl.py

@@ -5,7 +5,7 @@
 embedded audio track from that same file when ``use_audio_in_video=true``.
 
 Example:
-    python scripts/convert_video_r1_260k_to_unirl.py \
+    python datasets/daily_omni_av/convert_daily_omni_dataset_format_to_unirl.py \
       --train-input /path/to/daily_omni_av_train.jsonl \
       --val-input /path/to/daily_omni_av_val.jsonl \
       --out-dir datasets/daily_omni_av

scripts/convert_video_r1_260k_to_unirl.py → datasets/video_r1_260k/convert_video_r1_260k_to_unirl.py

@@ -10,9 +10,9 @@
 
 Missing videos are skipped unless ``--keep-missing`` is set. Example::
 
-    python scripts/convert_video_r1_260k_to_unirl.py \
+    python datasets/video_r1_260k/convert_video_r1_260k_to_unirl.py \
         --data-root /path/to/Video-R1-data \
-        --out-dir /path/to/output/video_r1_260k \
+        --out-dir datasets/video_r1_260k \
         --sources CLEVRER,STAR,NeXT-QA,PerceptionTest \
         --max-total 20000 --val-count 200
 """