2026-02-15 02:31:16,784 - INFO - === Loading & Creating Swap Pairs === 2026-02-15 02:31:20,238 - INFO - Swap pair creation stats: 2026-02-15 02:31:20,239 - INFO - left: 616/616 2026-02-15 02:31:20,239 - INFO - right: 620/620 2026-02-15 02:31:20,239 - INFO - above: 596/596 2026-02-15 02:31:20,239 - INFO - under: 602/602 2026-02-15 02:31:20,239 - INFO - far: 594/594 2026-02-15 02:31:20,239 - INFO - close: 612/612 2026-02-15 02:31:20,239 - INFO - Total pairs: 3640 2026-02-15 02:31:20,484 - INFO - PyTorch version 2.9.1 available. 2026-02-15 02:31:20,485 - INFO - TensorFlow version 2.20.0 available. 2026-02-15 02:31:20,487 - INFO - JAX version 0.6.2 available. 2026-02-15 02:31:20,837 - INFO - Loading HF dataset: FlagEval/EmbSpatial-Bench 2026-02-15 02:31:42,710 - INFO - Built bbox cache: 3640 entries (sample keys: ['mp3d_0', 'mp3d_1', 'mp3d_2', 'mp3d_3', 'mp3d_4']) 2026-02-15 02:31:42,743 - INFO - Matched 0/1206 indices between TSV and HF dataset 2026-02-15 02:31:42,744 - INFO - Cross-group quads: 0/1206 (ambiguous=0, no_bbox=1206) 2026-02-15 02:31:42,744 - INFO - ============================================================ 2026-02-15 02:31:42,744 - INFO - Processing molmo - vanilla 2026-02-15 02:31:42,744 - INFO - Model path: allenai/Molmo-7B-O-0924 2026-02-15 02:31:42,744 - INFO - ============================================================ 2026-02-15 02:31:46.346949: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`. 2026-02-15 02:31:46.461419: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. Loading checkpoint shards: 0%| | 0/7 [00:00