Upload train_20261004_220214357693/train_smoke_20261004_220154.log with huggingface_hub

train_20261004_220214357693/train_smoke_20261004_220154.log

@@ -0,0 +1,147 @@
+nnsight is not detected. Please install via 'pip install nnsight' for nnsight backend.
+[22:02:15] INFO    __main__:90 → Using bfloat16 for model HyperSteer
+
+Repo card metadata block was not found. Setting CardData to empty.
+[22:02:17] WARNING huggingface_hub.repocard:90 → Repo card metadata block was not found. Setting CardData to empty.
+[22:02:30] INFO    hypersteer.data.axbench:90 → Loading training dataset pyvene/axbench-concept500 with files {'train': '2b/l20/train/data.parquet'}
+[22:02:31] INFO    hypersteer.data.axbench:90 → Loaded dataset with 36216 examples
+[22:02:31] INFO    hypersteer.data.axbench:90 → Limited to 5 concepts with 360 examples
+[22:02:31] INFO    hypersteer.data.axbench:90 → Processed dataset with 360 examples ready for training
+[22:02:31] INFO    __main__:90 → Loaded and processed dataset with 360 examples
+[22:02:31] INFO    __main__:90 → Training HyperSteer with 5 concepts
+[22:02:31] INFO    __main__:90 → Training on 324 examples, validating on 36 examples
+[22:02:35] INFO    hypersteer.training.trainer:90 → Training with step limit: 120 steps
+[22:02:35] INFO    hypersteer.training.trainer:90 → Training configuration:
+[22:02:35] INFO    hypersteer.training.trainer:90 →   - Batch size: 4
+[22:02:35] INFO    hypersteer.training.trainer:90 →   - Gradient accumulation steps: 1
+[22:02:35] INFO    hypersteer.training.trainer:90 →   - Steps per epoch: 81
+[22:02:35] INFO    hypersteer.training.trainer:90 →   - Total training steps: 120
+[22:02:35] INFO    hypersteer.training.trainer:90 →   - Estimated epochs to complete: 1.48
+Starting from v4.46, the `logits` model output will have the same type as the model (except at train time, where it will always be FP32)
+[22:02:36] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.136397361755371, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 30.625, 'grad_norms_train/total': 10.0625, 'lr_train/main': 8e-05, 'counters/global_step': 0.0}
+[22:02:37] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.8061232566833496, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 8.125, 'grad_norms_train/total': 8.125, 'lr_train/main': 7.933333333333334e-05, 'counters/global_step': 1.0}
+[22:02:37] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.4270992279052734, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 6.28125, 'grad_norms_train/total': 6.28125, 'lr_train/main': 7.866666666666666e-05, 'counters/global_step': 2.0}
+[22:02:37] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2590880393981934, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 7.03125, 'grad_norms_train/total': 7.03125, 'lr_train/main': 7.8e-05, 'counters/global_step': 3.0}
+[22:02:38] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.6617071628570557, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 7.25, 'grad_norms_train/total': 7.25, 'lr_train/main': 7.733333333333333e-05, 'counters/global_step': 4.0}
+[22:02:38] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7490148544311523, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 5.34375, 'grad_norms_train/total': 5.34375, 'lr_train/main': 7.666666666666668e-05, 'counters/global_step': 5.0}
+[22:02:39] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.21524977684021, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 7.53125, 'grad_norms_train/total': 7.53125, 'lr_train/main': 7.6e-05, 'counters/global_step': 6.0}
+[22:02:39] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.209523916244507, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 5.03125, 'grad_norms_train/total': 5.03125, 'lr_train/main': 7.533333333333334e-05, 'counters/global_step': 7.0}
+[22:02:40] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.861177682876587, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 6.0, 'grad_norms_train/total': 6.0, 'lr_train/main': 7.466666666666667e-05, 'counters/global_step': 8.0}
+[22:02:40] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.204949378967285, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.53125, 'grad_norms_train/total': 4.53125, 'lr_train/main': 7.400000000000001e-05, 'counters/global_step': 9.0}
+[22:02:40] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.76420259475708, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.265625, 'grad_norms_train/total': 3.265625, 'lr_train/main': 7.333333333333333e-05, 'counters/global_step': 10.0}
+[22:02:41] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.657315969467163, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.765625, 'grad_norms_train/total': 3.765625, 'lr_train/main': 7.266666666666667e-05, 'counters/global_step': 11.0}
+[22:02:41] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.8396332263946533, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.640625, 'grad_norms_train/total': 3.640625, 'lr_train/main': 7.2e-05, 'counters/global_step': 12.0}
+[22:02:42] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.36782169342041, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.15625, 'grad_norms_train/total': 4.15625, 'lr_train/main': 7.133333333333334e-05, 'counters/global_step': 13.0}
+[22:02:42] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 3.0285732746124268, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.8125, 'grad_norms_train/total': 4.8125, 'lr_train/main': 7.066666666666667e-05, 'counters/global_step': 14.0}
+[22:02:42] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2328054904937744, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.71875, 'grad_norms_train/total': 4.71875, 'lr_train/main': 7.000000000000001e-05, 'counters/global_step': 15.0}
+[22:02:43] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2460739612579346, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.546875, 'grad_norms_train/total': 3.546875, 'lr_train/main': 6.933333333333334e-05, 'counters/global_step': 16.0}
+[22:02:43] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.866987705230713, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.0, 'grad_norms_train/total': 4.0, 'lr_train/main': 6.866666666666666e-05, 'counters/global_step': 17.0}
+[22:02:44] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2487246990203857, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.3125, 'grad_norms_train/total': 4.3125, 'lr_train/main': 6.8e-05, 'counters/global_step': 18.0}
+[22:02:44] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.218796968460083, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.015625, 'grad_norms_train/total': 3.015625, 'lr_train/main': 6.733333333333333e-05, 'counters/global_step': 19.0}
+[22:02:45] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.760968565940857, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.609375, 'grad_norms_train/total': 3.609375, 'lr_train/main': 6.666666666666667e-05, 'counters/global_step': 20.0}
+[22:02:45] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.409729242324829, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.5, 'grad_norms_train/total': 4.5, 'lr_train/main': 6.6e-05, 'counters/global_step': 21.0}
+[22:02:45] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.9436427354812622, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.953125, 'grad_norms_train/total': 2.953125, 'lr_train/main': 6.533333333333334e-05, 'counters/global_step': 22.0}
+[22:02:46] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.0295557975769043, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.640625, 'grad_norms_train/total': 3.640625, 'lr_train/main': 6.466666666666668e-05, 'counters/global_step': 23.0}
+[22:02:46] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.124727487564087, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.078125, 'grad_norms_train/total': 3.078125, 'lr_train/main': 6.400000000000001e-05, 'counters/global_step': 24.0}
+[22:02:46] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.5254278182983398, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 4.0625, 'grad_norms_train/total': 4.0625, 'lr_train/main': 6.333333333333333e-05, 'counters/global_step': 25.0}
+[22:02:47] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7669650316238403, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.8125, 'grad_norms_train/total': 2.8125, 'lr_train/main': 6.266666666666667e-05, 'counters/global_step': 26.0}
+[22:02:47] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.063863515853882, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.546875, 'grad_norms_train/total': 3.546875, 'lr_train/main': 6.2e-05, 'counters/global_step': 27.0}
+[22:02:48] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.555615782737732, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.140625, 'grad_norms_train/total': 2.140625, 'lr_train/main': 6.133333333333334e-05, 'counters/global_step': 28.0}
+[22:02:48] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.590324640274048, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.328125, 'grad_norms_train/total': 3.328125, 'lr_train/main': 6.0666666666666666e-05, 'counters/global_step': 29.0}
+[22:02:49] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.061677932739258, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.53125, 'grad_norms_train/total': 2.53125, 'lr_train/main': 6.000000000000001e-05, 'counters/global_step': 30.0}
+[22:02:49] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.887484312057495, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.9375, 'grad_norms_train/total': 3.9375, 'lr_train/main': 5.9333333333333343e-05, 'counters/global_step': 31.0}
+[22:02:49] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.7663204669952393, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.984375, 'grad_norms_train/total': 2.984375, 'lr_train/main': 5.8666666666666665e-05, 'counters/global_step': 32.0}
+[22:02:50] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2939393520355225, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.640625, 'grad_norms_train/total': 2.640625, 'lr_train/main': 5.8e-05, 'counters/global_step': 33.0}
+[22:02:50] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.126936197280884, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.703125, 'grad_norms_train/total': 2.703125, 'lr_train/main': 5.7333333333333336e-05, 'counters/global_step': 34.0}
+[22:02:51] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.6013834476470947, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.125, 'grad_norms_train/total': 3.125, 'lr_train/main': 5.666666666666668e-05, 'counters/global_step': 35.0}
+[22:02:51] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.131223201751709, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.375, 'grad_norms_train/total': 3.375, 'lr_train/main': 5.6e-05, 'counters/global_step': 36.0}
+[22:02:51] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.4475510120391846, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.609375, 'grad_norms_train/total': 2.609375, 'lr_train/main': 5.5333333333333334e-05, 'counters/global_step': 37.0}
+[22:02:52] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.787796974182129, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.765625, 'grad_norms_train/total': 2.765625, 'lr_train/main': 5.466666666666667e-05, 'counters/global_step': 38.0}
+[22:02:52] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8406022787094116, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.265625, 'grad_norms_train/total': 3.265625, 'lr_train/main': 5.4000000000000005e-05, 'counters/global_step': 39.0}
+[22:02:53] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2866830825805664, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.453125, 'grad_norms_train/total': 3.453125, 'lr_train/main': 5.333333333333333e-05, 'counters/global_step': 40.0}
+[22:02:53] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8580210208892822, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.328125, 'grad_norms_train/total': 2.328125, 'lr_train/main': 5.266666666666667e-05, 'counters/global_step': 41.0}
+[22:02:54] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.3744038343429565, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.078125, 'grad_norms_train/total': 2.078125, 'lr_train/main': 5.2000000000000004e-05, 'counters/global_step': 42.0}
+[22:02:54] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.3904608488082886, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.0, 'grad_norms_train/total': 2.0, 'lr_train/main': 5.133333333333334e-05, 'counters/global_step': 43.0}
+[22:02:54] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.303591251373291, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.640625, 'grad_norms_train/total': 2.640625, 'lr_train/main': 5.066666666666667e-05, 'counters/global_step': 44.0}
+[22:02:55] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.9559040069580078, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.609375, 'grad_norms_train/total': 2.609375, 'lr_train/main': 5e-05, 'counters/global_step': 45.0}
+[22:02:55] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.631472110748291, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.375, 'grad_norms_train/total': 2.375, 'lr_train/main': 4.933333333333334e-05, 'counters/global_step': 46.0}
+[22:02:56] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7817332744598389, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.46875, 'grad_norms_train/total': 2.46875, 'lr_train/main': 4.8666666666666666e-05, 'counters/global_step': 47.0}
+[22:02:56] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 0.9990031719207764, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.90625, 'grad_norms_train/total': 1.90625, 'lr_train/main': 4.8e-05, 'counters/global_step': 48.0}
+[22:02:56] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.6810382604599, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.890625, 'grad_norms_train/total': 2.890625, 'lr_train/main': 4.7333333333333336e-05, 'counters/global_step': 49.0}
+[22:03:05] INFO    hypersteer.training.trainer:90 → Saved checkpoint to assets/checkpoints/train_20261004_220214357693/train/checkpoints/step_50
+[22:03:08] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.1283950805664062, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.28125, 'grad_norms_train/total': 3.28125, 'lr_train/main': 4.666666666666667e-05, 'counters/global_step': 50.0}
+[22:03:09] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8295527696609497, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.671875, 'grad_norms_train/total': 2.671875, 'lr_train/main': 4.6e-05, 'counters/global_step': 51.0}
+[22:03:09] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.5988845825195312, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.28125, 'grad_norms_train/total': 3.28125, 'lr_train/main': 4.5333333333333335e-05, 'counters/global_step': 52.0}
+[22:03:10] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2205307483673096, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.09375, 'grad_norms_train/total': 3.09375, 'lr_train/main': 4.466666666666667e-05, 'counters/global_step': 53.0}
+[22:03:10] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.4861518144607544, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.046875, 'grad_norms_train/total': 2.046875, 'lr_train/main': 4.4000000000000006e-05, 'counters/global_step': 54.0}
+[22:03:10] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.4796069860458374, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.71875, 'grad_norms_train/total': 2.71875, 'lr_train/main': 4.3333333333333334e-05, 'counters/global_step': 55.0}
+[22:03:11] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.127621650695801, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.671875, 'grad_norms_train/total': 2.671875, 'lr_train/main': 4.266666666666667e-05, 'counters/global_step': 56.0}
+[22:03:11] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.0393054485321045, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.609375, 'grad_norms_train/total': 2.609375, 'lr_train/main': 4.2000000000000004e-05, 'counters/global_step': 57.0}
+[22:03:12] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.238032817840576, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.859375, 'grad_norms_train/total': 2.859375, 'lr_train/main': 4.133333333333334e-05, 'counters/global_step': 58.0}
+[22:03:12] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8819373846054077, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.796875, 'grad_norms_train/total': 2.796875, 'lr_train/main': 4.066666666666667e-05, 'counters/global_step': 59.0}
+[22:03:13] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.6185877323150635, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.859375, 'grad_norms_train/total': 2.859375, 'lr_train/main': 4e-05, 'counters/global_step': 60.0}
+[22:03:13] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.1532063484191895, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.796875, 'grad_norms_train/total': 2.796875, 'lr_train/main': 3.933333333333333e-05, 'counters/global_step': 61.0}
+[22:03:14] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7693978548049927, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.375, 'grad_norms_train/total': 2.375, 'lr_train/main': 3.866666666666667e-05, 'counters/global_step': 62.0}
+[22:03:14] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.883247971534729, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.078125, 'grad_norms_train/total': 3.078125, 'lr_train/main': 3.8e-05, 'counters/global_step': 63.0}
+[22:03:15] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.352245569229126, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.59375, 'grad_norms_train/total': 2.59375, 'lr_train/main': 3.733333333333334e-05, 'counters/global_step': 64.0}
+[22:03:15] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8544378280639648, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.8125, 'grad_norms_train/total': 2.8125, 'lr_train/main': 3.6666666666666666e-05, 'counters/global_step': 65.0}
+[22:03:15] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.9862251281738281, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.203125, 'grad_norms_train/total': 2.203125, 'lr_train/main': 3.6e-05, 'counters/global_step': 66.0}
+[22:03:16] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.6000604629516602, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.9921875, 'grad_norms_train/total': 1.9921875, 'lr_train/main': 3.5333333333333336e-05, 'counters/global_step': 67.0}
+[22:03:16] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.3180644512176514, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.765625, 'grad_norms_train/total': 2.765625, 'lr_train/main': 3.466666666666667e-05, 'counters/global_step': 68.0}
+[22:03:16] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.421393871307373, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.609375, 'grad_norms_train/total': 2.609375, 'lr_train/main': 3.4e-05, 'counters/global_step': 69.0}
+[22:03:17] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.087498426437378, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.578125, 'grad_norms_train/total': 2.578125, 'lr_train/main': 3.3333333333333335e-05, 'counters/global_step': 70.0}
+[22:03:17] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7978473901748657, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.1875, 'grad_norms_train/total': 2.1875, 'lr_train/main': 3.266666666666667e-05, 'counters/global_step': 71.0}
+[22:03:18] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.303210496902466, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.828125, 'grad_norms_train/total': 2.828125, 'lr_train/main': 3.2000000000000005e-05, 'counters/global_step': 72.0}
+[22:03:18] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.5675880908966064, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.765625, 'grad_norms_train/total': 2.765625, 'lr_train/main': 3.1333333333333334e-05, 'counters/global_step': 73.0}
+[22:03:18] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.473565101623535, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.640625, 'grad_norms_train/total': 2.640625, 'lr_train/main': 3.066666666666667e-05, 'counters/global_step': 74.0}
+[22:03:19] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.171020984649658, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.4375, 'grad_norms_train/total': 2.4375, 'lr_train/main': 3.0000000000000004e-05, 'counters/global_step': 75.0}
+[22:03:19] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7160491943359375, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.453125, 'grad_norms_train/total': 2.453125, 'lr_train/main': 2.9333333333333333e-05, 'counters/global_step': 76.0}
+[22:03:20] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.6938750743865967, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.625, 'grad_norms_train/total': 3.625, 'lr_train/main': 2.8666666666666668e-05, 'counters/global_step': 77.0}
+[22:03:20] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.9226447343826294, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.546875, 'grad_norms_train/total': 2.546875, 'lr_train/main': 2.8e-05, 'counters/global_step': 78.0}
+[22:03:20] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.1258268356323242, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.8828125, 'grad_norms_train/total': 1.8828125, 'lr_train/main': 2.7333333333333335e-05, 'counters/global_step': 79.0}
+[22:03:21] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8315863609313965, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.15625, 'grad_norms_train/total': 2.15625, 'lr_train/main': 2.6666666666666667e-05, 'counters/global_step': 80.0}
+[22:03:21] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 0.7933372259140015, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.671875, 'grad_norms_train/total': 1.671875, 'lr_train/main': 2.6000000000000002e-05, 'counters/global_step': 81.0}
+[22:03:22] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.2071311473846436, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.578125, 'grad_norms_train/total': 2.578125, 'lr_train/main': 2.5333333333333334e-05, 'counters/global_step': 82.0}
+[22:03:22] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.9848285913467407, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.765625, 'grad_norms_train/total': 2.765625, 'lr_train/main': 2.466666666666667e-05, 'counters/global_step': 83.0}
+[22:03:23] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.3570990562438965, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.78125, 'grad_norms_train/total': 2.78125, 'lr_train/main': 2.4e-05, 'counters/global_step': 84.0}
+[22:03:23] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.1348540782928467, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.53125, 'grad_norms_train/total': 3.53125, 'lr_train/main': 2.3333333333333336e-05, 'counters/global_step': 85.0}
+[22:03:23] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.5858500003814697, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.953125, 'grad_norms_train/total': 2.953125, 'lr_train/main': 2.2666666666666668e-05, 'counters/global_step': 86.0}
+[22:03:24] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.2749254703521729, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.7265625, 'grad_norms_train/total': 1.7265625, 'lr_train/main': 2.2000000000000003e-05, 'counters/global_step': 87.0}
+[22:03:24] INFO    hypersteer.training.trainer:90 → Uploaded checkpoint to RSD002/hypersteer-gemma2-2b-l20/train_20261004_220214357693/train/checkpoints/step_50
+[22:03:24] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.131559133529663, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.328125, 'grad_norms_train/total': 2.328125, 'lr_train/main': 2.1333333333333335e-05, 'counters/global_step': 88.0}
+[22:03:25] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.0927770137786865, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.4375, 'grad_norms_train/total': 2.4375, 'lr_train/main': 2.066666666666667e-05, 'counters/global_step': 89.0}
+[22:03:25] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.857898235321045, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.421875, 'grad_norms_train/total': 2.421875, 'lr_train/main': 2e-05, 'counters/global_step': 90.0}
+[22:03:25] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.227248430252075, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.140625, 'grad_norms_train/total': 2.140625, 'lr_train/main': 1.9333333333333333e-05, 'counters/global_step': 91.0}
+[22:03:26] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.3254125118255615, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.796875, 'grad_norms_train/total': 2.796875, 'lr_train/main': 1.866666666666667e-05, 'counters/global_step': 92.0}
+[22:03:26] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.9005273580551147, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.75, 'grad_norms_train/total': 2.75, 'lr_train/main': 1.8e-05, 'counters/global_step': 93.0}
+[22:03:27] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.290607452392578, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.671875, 'grad_norms_train/total': 2.671875, 'lr_train/main': 1.7333333333333336e-05, 'counters/global_step': 94.0}
+[22:03:27] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.4246745109558105, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.59375, 'grad_norms_train/total': 2.59375, 'lr_train/main': 1.6666666666666667e-05, 'counters/global_step': 95.0}
+[22:03:27] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8835093975067139, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.3125, 'grad_norms_train/total': 2.3125, 'lr_train/main': 1.6000000000000003e-05, 'counters/global_step': 96.0}
+[22:03:28] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7305606603622437, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.0, 'grad_norms_train/total': 3.0, 'lr_train/main': 1.5333333333333334e-05, 'counters/global_step': 97.0}
+[22:03:28] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.428774833679199, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.09375, 'grad_norms_train/total': 3.09375, 'lr_train/main': 1.4666666666666666e-05, 'counters/global_step': 98.0}
+[22:03:29] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.6375865936279297, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.03125, 'grad_norms_train/total': 2.03125, 'lr_train/main': 1.4e-05, 'counters/global_step': 99.0}
+[22:03:29] INFO    hypersteer.training.trainer:90 → Running validation at step 100
+[22:03:30] INFO    hypersteer.training.trainer:90 → {'counters/val_step': 100.0, 'counters/global_val_step': 0.0}
+[22:03:39] INFO    hypersteer.training.trainer:90 → Saved checkpoint to assets/checkpoints/train_20261004_220214357693/train/checkpoints/step_100
+[22:03:40] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.4711313247680664, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.09375, 'grad_norms_train/total': 2.09375, 'lr_train/main': 1.3333333333333333e-05, 'counters/global_step': 100.0}
+[22:03:40] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8098468780517578, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.515625, 'grad_norms_train/total': 2.515625, 'lr_train/main': 1.2666666666666667e-05, 'counters/global_step': 101.0}
+[22:03:41] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.3524158000946045, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.46875, 'grad_norms_train/total': 2.46875, 'lr_train/main': 1.2e-05, 'counters/global_step': 102.0}
+[22:03:41] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.3072856664657593, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.96875, 'grad_norms_train/total': 1.96875, 'lr_train/main': 1.1333333333333334e-05, 'counters/global_step': 103.0}
+[22:03:42] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.132373332977295, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.578125, 'grad_norms_train/total': 2.578125, 'lr_train/main': 1.0666666666666667e-05, 'counters/global_step': 104.0}
+[22:03:42] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8288068771362305, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.078125, 'grad_norms_train/total': 2.078125, 'lr_train/main': 1e-05, 'counters/global_step': 105.0}
+[22:03:42] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8742825984954834, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.5625, 'grad_norms_train/total': 2.5625, 'lr_train/main': 9.333333333333334e-06, 'counters/global_step': 106.0}
+[22:03:43] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8266394138336182, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.21875, 'grad_norms_train/total': 2.21875, 'lr_train/main': 8.666666666666668e-06, 'counters/global_step': 107.0}
+[22:03:43] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 2.207374334335327, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.609375, 'grad_norms_train/total': 2.609375, 'lr_train/main': 8.000000000000001e-06, 'counters/global_step': 108.0}
+[22:03:43] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.995695948600769, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.53125, 'grad_norms_train/total': 2.53125, 'lr_train/main': 7.333333333333333e-06, 'counters/global_step': 109.0}
+[22:03:44] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.018134593963623, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.6796875, 'grad_norms_train/total': 1.6796875, 'lr_train/main': 6.666666666666667e-06, 'counters/global_step': 110.0}
+[22:03:44] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.698227047920227, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 3.0, 'grad_norms_train/total': 3.0, 'lr_train/main': 6e-06, 'counters/global_step': 111.0}
+[22:03:45] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.2632185220718384, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 1.9453125, 'grad_norms_train/total': 1.9453125, 'lr_train/main': 5.333333333333334e-06, 'counters/global_step': 112.0}
+[22:03:45] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.6968705654144287, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.109375, 'grad_norms_train/total': 2.109375, 'lr_train/main': 4.666666666666667e-06, 'counters/global_step': 113.0}
+[22:03:45] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.9926882982254028, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.90625, 'grad_norms_train/total': 2.90625, 'lr_train/main': 4.000000000000001e-06, 'counters/global_step': 114.0}
+[22:03:46] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.5217572450637817, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.4375, 'grad_norms_train/total': 2.4375, 'lr_train/main': 3.3333333333333333e-06, 'counters/global_step': 115.0}
+[22:03:46] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.7003213167190552, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.453125, 'grad_norms_train/total': 2.453125, 'lr_train/main': 2.666666666666667e-06, 'counters/global_step': 116.0}
+[22:03:47] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.8324843645095825, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.53125, 'grad_norms_train/total': 2.53125, 'lr_train/main': 2.0000000000000003e-06, 'counters/global_step': 117.0}
+[22:03:47] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.451796293258667, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.171875, 'grad_norms_train/total': 2.171875, 'lr_train/main': 1.3333333333333334e-06, 'counters/global_step': 118.0}
+[22:03:48] INFO    hypersteer.training.trainer:90 → {'loss_train/main': 1.5711824893951416, 'grad_norms_train/ax': 0.0, 'grad_norms_train/concept_embedding': 2.078125, 'grad_norms_train/total': 2.078125, 'lr_train/main': 6.666666666666667e-07, 'counters/global_step': 119.0}
+[22:03:48] INFO    hypersteer.training.trainer:90 → Reached step limit of 120 steps. Terminating training.
+[22:04:14] INFO    hypersteer.training.trainer:90 → Uploaded checkpoint to RSD002/hypersteer-gemma2-2b-l20/train_20261004_220214357693/train/checkpoints/step_100