| { | |
| "best_metric": 0.96, | |
| "best_model_checkpoint": "timesformer-base-finetuned-k400-kinetic400-subset-epoch6-num_frame_10_myViT2window4_more_data_b4/checkpoint-26", | |
| "epoch": 5.097222222222222, | |
| "eval_steps": 500, | |
| "global_step": 72, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "accuracy": 0.25, | |
| "epoch": 0, | |
| "step": 0 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 1 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.03, | |
| "step": 2 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 0.04, | |
| "step": 3 | |
| }, | |
| { | |
| "accuracy": 0.5, | |
| "epoch": 0.06, | |
| "step": 4 | |
| }, | |
| { | |
| "accuracy": 0.25, | |
| "epoch": 0.07, | |
| "step": 5 | |
| }, | |
| { | |
| "accuracy": 0.5, | |
| "epoch": 0.08, | |
| "step": 6 | |
| }, | |
| { | |
| "accuracy": 0.5, | |
| "epoch": 0.1, | |
| "step": 7 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 0.11, | |
| "step": 8 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 0.12, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "grad_norm": 12.026504516601562, | |
| "learning_rate": 4.8437500000000005e-05, | |
| "loss": 1.407, | |
| "step": 10 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 10 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 11 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 12 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_accuracy": 0.92, | |
| "eval_loss": 0.6302356719970703, | |
| "eval_runtime": 26.1162, | |
| "eval_samples_per_second": 0.957, | |
| "eval_steps_per_second": 0.268, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.18, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 14 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 15 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 16 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 17 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 18 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "grad_norm": 1.0873479843139648, | |
| "learning_rate": 4.0625000000000005e-05, | |
| "loss": 0.2401, | |
| "step": 20 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 20 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 21 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 22 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 23 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 1.15, | |
| "step": 24 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 25 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "eval_accuracy": 0.96, | |
| "eval_loss": 0.23198732733726501, | |
| "eval_runtime": 25.4625, | |
| "eval_samples_per_second": 0.982, | |
| "eval_steps_per_second": 0.275, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.18, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 27 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 28 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "grad_norm": 0.09468521922826767, | |
| "learning_rate": 3.2812500000000005e-05, | |
| "loss": 0.0568, | |
| "step": 30 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 30 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 31 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 32 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 33 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 34 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 35 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 36 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 37 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 38 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "eval_accuracy": 0.96, | |
| "eval_loss": 0.1464664489030838, | |
| "eval_runtime": 25.5239, | |
| "eval_samples_per_second": 0.979, | |
| "eval_steps_per_second": 0.274, | |
| "step": 39 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.18, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "grad_norm": 0.08358396589756012, | |
| "learning_rate": 2.5e-05, | |
| "loss": 0.0104, | |
| "step": 40 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 40 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 41 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 42 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 43 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 44 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 45 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 46 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 47 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 48 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "grad_norm": 0.045649539679288864, | |
| "learning_rate": 1.71875e-05, | |
| "loss": 0.0045, | |
| "step": 50 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 50 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 51 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "eval_accuracy": 0.96, | |
| "eval_loss": 0.11306390166282654, | |
| "eval_runtime": 25.8653, | |
| "eval_samples_per_second": 0.967, | |
| "eval_steps_per_second": 0.271, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.18, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 53 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 54 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 55 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 56 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 57 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 58 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "grad_norm": 0.09641703963279724, | |
| "learning_rate": 9.375000000000001e-06, | |
| "loss": 0.0027, | |
| "step": 60 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 60 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 61 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 62 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 63 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 64 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "eval_accuracy": 0.92, | |
| "eval_loss": 0.12509094178676605, | |
| "eval_runtime": 25.9193, | |
| "eval_samples_per_second": 0.965, | |
| "eval_steps_per_second": 0.27, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.18, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 66 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 67 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 68 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "grad_norm": 0.029298102483153343, | |
| "learning_rate": 1.5625e-06, | |
| "loss": 0.0022, | |
| "step": 70 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 70 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 71 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "eval_accuracy": 0.92, | |
| "eval_loss": 0.1256319135427475, | |
| "eval_runtime": 26.5176, | |
| "eval_samples_per_second": 0.943, | |
| "eval_steps_per_second": 0.264, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "step": 72, | |
| "total_flos": 3.044708388358963e+17, | |
| "train_loss": 0.23946668859571218, | |
| "train_runtime": 463.5436, | |
| "train_samples_per_second": 0.621, | |
| "train_steps_per_second": 0.155 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "eval_accuracy": 0.96, | |
| "eval_loss": 0.18762756884098053, | |
| "eval_runtime": 18.2837, | |
| "eval_samples_per_second": 1.367, | |
| "eval_steps_per_second": 0.383, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 0.75, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "eval_accuracy": 0.96, | |
| "eval_loss": 0.18762756884098053, | |
| "eval_runtime": 17.5715, | |
| "eval_samples_per_second": 1.423, | |
| "eval_steps_per_second": 0.398, | |
| "step": 72 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 72, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 9223372036854775807, | |
| "save_steps": 500, | |
| "total_flos": 3.044708388358963e+17, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |