| { | |
| "best_metric": 0.33774590492248535, | |
| "best_model_checkpoint": "deepfake_vs_real_image_detection/checkpoint-1392", | |
| "epoch": 87.0, | |
| "eval_steps": 500, | |
| "global_step": 1392, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6765578635014837, | |
| "eval_loss": 1.4959243535995483, | |
| "eval_runtime": 4.5936, | |
| "eval_samples_per_second": 73.362, | |
| "eval_steps_per_second": 9.361, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.6913946587537092, | |
| "eval_loss": 1.4148070812225342, | |
| "eval_runtime": 5.1723, | |
| "eval_samples_per_second": 65.154, | |
| "eval_steps_per_second": 8.313, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7091988130563798, | |
| "eval_loss": 1.309860348701477, | |
| "eval_runtime": 4.8265, | |
| "eval_samples_per_second": 69.823, | |
| "eval_steps_per_second": 8.909, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.7240356083086054, | |
| "eval_loss": 1.2197798490524292, | |
| "eval_runtime": 4.7427, | |
| "eval_samples_per_second": 71.057, | |
| "eval_steps_per_second": 9.067, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.7477744807121661, | |
| "eval_loss": 1.1465003490447998, | |
| "eval_runtime": 4.8571, | |
| "eval_samples_per_second": 69.383, | |
| "eval_steps_per_second": 8.853, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.7655786350148368, | |
| "eval_loss": 1.0904771089553833, | |
| "eval_runtime": 4.7162, | |
| "eval_samples_per_second": 71.456, | |
| "eval_steps_per_second": 9.118, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.7626112759643917, | |
| "eval_loss": 1.0423768758773804, | |
| "eval_runtime": 4.7414, | |
| "eval_samples_per_second": 71.077, | |
| "eval_steps_per_second": 9.069, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.7626112759643917, | |
| "eval_loss": 0.9944449067115784, | |
| "eval_runtime": 4.7288, | |
| "eval_samples_per_second": 71.265, | |
| "eval_steps_per_second": 9.093, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.7626112759643917, | |
| "eval_loss": 0.9456853270530701, | |
| "eval_runtime": 4.6848, | |
| "eval_samples_per_second": 71.935, | |
| "eval_steps_per_second": 9.179, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.7626112759643917, | |
| "eval_loss": 0.8957338333129883, | |
| "eval_runtime": 4.7194, | |
| "eval_samples_per_second": 71.407, | |
| "eval_steps_per_second": 9.111, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.7626112759643917, | |
| "eval_loss": 0.852699875831604, | |
| "eval_runtime": 4.7114, | |
| "eval_samples_per_second": 71.529, | |
| "eval_steps_per_second": 9.127, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.7626112759643917, | |
| "eval_loss": 0.815643310546875, | |
| "eval_runtime": 4.7006, | |
| "eval_samples_per_second": 71.694, | |
| "eval_steps_per_second": 9.148, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.771513353115727, | |
| "eval_loss": 0.7831457853317261, | |
| "eval_runtime": 4.7983, | |
| "eval_samples_per_second": 70.233, | |
| "eval_steps_per_second": 8.961, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.7744807121661721, | |
| "eval_loss": 0.7558016180992126, | |
| "eval_runtime": 4.7359, | |
| "eval_samples_per_second": 71.159, | |
| "eval_steps_per_second": 9.08, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_accuracy": 0.7744807121661721, | |
| "eval_loss": 0.7284858226776123, | |
| "eval_runtime": 4.6713, | |
| "eval_samples_per_second": 72.143, | |
| "eval_steps_per_second": 9.205, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_accuracy": 0.7744807121661721, | |
| "eval_loss": 0.6914460062980652, | |
| "eval_runtime": 4.7809, | |
| "eval_samples_per_second": 70.489, | |
| "eval_steps_per_second": 8.994, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_accuracy": 0.7774480712166172, | |
| "eval_loss": 0.6581370830535889, | |
| "eval_runtime": 4.7253, | |
| "eval_samples_per_second": 71.318, | |
| "eval_steps_per_second": 9.1, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_accuracy": 0.7804154302670623, | |
| "eval_loss": 0.6323521733283997, | |
| "eval_runtime": 4.7945, | |
| "eval_samples_per_second": 70.289, | |
| "eval_steps_per_second": 8.969, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_accuracy": 0.7774480712166172, | |
| "eval_loss": 0.6072251796722412, | |
| "eval_runtime": 4.8798, | |
| "eval_samples_per_second": 69.06, | |
| "eval_steps_per_second": 8.812, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_accuracy": 0.7804154302670623, | |
| "eval_loss": 0.5835662484169006, | |
| "eval_runtime": 4.6039, | |
| "eval_samples_per_second": 73.199, | |
| "eval_steps_per_second": 9.34, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_accuracy": 0.7922848664688428, | |
| "eval_loss": 0.5593723654747009, | |
| "eval_runtime": 4.7955, | |
| "eval_samples_per_second": 70.274, | |
| "eval_steps_per_second": 8.967, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_accuracy": 0.7922848664688428, | |
| "eval_loss": 0.5364292860031128, | |
| "eval_runtime": 4.8089, | |
| "eval_samples_per_second": 70.078, | |
| "eval_steps_per_second": 8.942, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_accuracy": 0.7952522255192879, | |
| "eval_loss": 0.5182026028633118, | |
| "eval_runtime": 4.6149, | |
| "eval_samples_per_second": 73.024, | |
| "eval_steps_per_second": 9.318, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_accuracy": 0.8041543026706232, | |
| "eval_loss": 0.504047691822052, | |
| "eval_runtime": 4.5605, | |
| "eval_samples_per_second": 73.895, | |
| "eval_steps_per_second": 9.429, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_accuracy": 0.8041543026706232, | |
| "eval_loss": 0.4920850694179535, | |
| "eval_runtime": 4.6821, | |
| "eval_samples_per_second": 71.976, | |
| "eval_steps_per_second": 9.184, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_accuracy": 0.8160237388724035, | |
| "eval_loss": 0.48037412762641907, | |
| "eval_runtime": 4.7178, | |
| "eval_samples_per_second": 71.432, | |
| "eval_steps_per_second": 9.114, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_accuracy": 0.8189910979228486, | |
| "eval_loss": 0.4688655138015747, | |
| "eval_runtime": 4.6086, | |
| "eval_samples_per_second": 73.124, | |
| "eval_steps_per_second": 9.33, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_accuracy": 0.8249258160237388, | |
| "eval_loss": 0.4574899971485138, | |
| "eval_runtime": 4.6369, | |
| "eval_samples_per_second": 72.678, | |
| "eval_steps_per_second": 9.273, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_accuracy": 0.827893175074184, | |
| "eval_loss": 0.448953777551651, | |
| "eval_runtime": 4.7138, | |
| "eval_samples_per_second": 71.491, | |
| "eval_steps_per_second": 9.122, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_accuracy": 0.827893175074184, | |
| "eval_loss": 0.4414035379886627, | |
| "eval_runtime": 4.7192, | |
| "eval_samples_per_second": 71.41, | |
| "eval_steps_per_second": 9.112, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_accuracy": 0.827893175074184, | |
| "eval_loss": 0.4335722029209137, | |
| "eval_runtime": 4.6675, | |
| "eval_samples_per_second": 72.202, | |
| "eval_steps_per_second": 9.213, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 31.25, | |
| "grad_norm": 2.672912120819092, | |
| "learning_rate": 7.096774193548387e-07, | |
| "loss": 0.6719, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_accuracy": 0.8219584569732937, | |
| "eval_loss": 0.43041178584098816, | |
| "eval_runtime": 4.7762, | |
| "eval_samples_per_second": 70.559, | |
| "eval_steps_per_second": 9.003, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_accuracy": 0.8249258160237388, | |
| "eval_loss": 0.4242512881755829, | |
| "eval_runtime": 4.6367, | |
| "eval_samples_per_second": 72.681, | |
| "eval_steps_per_second": 9.274, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_accuracy": 0.827893175074184, | |
| "eval_loss": 0.421082079410553, | |
| "eval_runtime": 4.733, | |
| "eval_samples_per_second": 71.201, | |
| "eval_steps_per_second": 9.085, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_accuracy": 0.8338278931750742, | |
| "eval_loss": 0.41326144337654114, | |
| "eval_runtime": 4.6282, | |
| "eval_samples_per_second": 72.815, | |
| "eval_steps_per_second": 9.291, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_accuracy": 0.8338278931750742, | |
| "eval_loss": 0.41096124053001404, | |
| "eval_runtime": 4.6107, | |
| "eval_samples_per_second": 73.092, | |
| "eval_steps_per_second": 9.326, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_accuracy": 0.8367952522255193, | |
| "eval_loss": 0.404769629240036, | |
| "eval_runtime": 4.6205, | |
| "eval_samples_per_second": 72.936, | |
| "eval_steps_per_second": 9.306, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "eval_accuracy": 0.8427299703264095, | |
| "eval_loss": 0.40141698718070984, | |
| "eval_runtime": 4.6612, | |
| "eval_samples_per_second": 72.298, | |
| "eval_steps_per_second": 9.225, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.3952626883983612, | |
| "eval_runtime": 4.6328, | |
| "eval_samples_per_second": 72.742, | |
| "eval_steps_per_second": 9.282, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.390024334192276, | |
| "eval_runtime": 4.5638, | |
| "eval_samples_per_second": 73.843, | |
| "eval_steps_per_second": 9.422, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.3834027349948883, | |
| "eval_runtime": 4.6261, | |
| "eval_samples_per_second": 72.848, | |
| "eval_steps_per_second": 9.295, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.3803389072418213, | |
| "eval_runtime": 4.6158, | |
| "eval_samples_per_second": 73.009, | |
| "eval_steps_per_second": 9.316, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.37990784645080566, | |
| "eval_runtime": 4.6288, | |
| "eval_samples_per_second": 72.805, | |
| "eval_steps_per_second": 9.29, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.3741393983364105, | |
| "eval_runtime": 4.7575, | |
| "eval_samples_per_second": 70.836, | |
| "eval_steps_per_second": 9.038, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 45.0, | |
| "eval_accuracy": 0.8545994065281899, | |
| "eval_loss": 0.36607253551483154, | |
| "eval_runtime": 4.6476, | |
| "eval_samples_per_second": 72.511, | |
| "eval_steps_per_second": 9.252, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.3655509054660797, | |
| "eval_runtime": 4.6664, | |
| "eval_samples_per_second": 72.218, | |
| "eval_steps_per_second": 9.215, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 47.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.36614400148391724, | |
| "eval_runtime": 4.6792, | |
| "eval_samples_per_second": 72.021, | |
| "eval_steps_per_second": 9.19, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.3684944808483124, | |
| "eval_runtime": 4.6175, | |
| "eval_samples_per_second": 72.983, | |
| "eval_steps_per_second": 9.312, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.36265119910240173, | |
| "eval_runtime": 4.6573, | |
| "eval_samples_per_second": 72.36, | |
| "eval_steps_per_second": 9.233, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 50.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.359169065952301, | |
| "eval_runtime": 4.7664, | |
| "eval_samples_per_second": 70.703, | |
| "eval_steps_per_second": 9.021, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.35570138692855835, | |
| "eval_runtime": 4.7355, | |
| "eval_samples_per_second": 71.165, | |
| "eval_steps_per_second": 9.08, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.35541436076164246, | |
| "eval_runtime": 4.7442, | |
| "eval_samples_per_second": 71.035, | |
| "eval_steps_per_second": 9.064, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.3521236479282379, | |
| "eval_runtime": 4.6755, | |
| "eval_samples_per_second": 72.078, | |
| "eval_steps_per_second": 9.197, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.3523922264575958, | |
| "eval_runtime": 4.6657, | |
| "eval_samples_per_second": 72.23, | |
| "eval_steps_per_second": 9.216, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 55.0, | |
| "eval_accuracy": 0.8516320474777448, | |
| "eval_loss": 0.34878119826316833, | |
| "eval_runtime": 4.6488, | |
| "eval_samples_per_second": 72.491, | |
| "eval_steps_per_second": 9.25, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "eval_accuracy": 0.8516320474777448, | |
| "eval_loss": 0.3482997715473175, | |
| "eval_runtime": 4.683, | |
| "eval_samples_per_second": 71.963, | |
| "eval_steps_per_second": 9.182, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.348295122385025, | |
| "eval_runtime": 4.8638, | |
| "eval_samples_per_second": 69.287, | |
| "eval_steps_per_second": 8.841, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "eval_accuracy": 0.8456973293768546, | |
| "eval_loss": 0.3468620479106903, | |
| "eval_runtime": 4.8133, | |
| "eval_samples_per_second": 70.014, | |
| "eval_steps_per_second": 8.934, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "eval_accuracy": 0.8486646884272997, | |
| "eval_loss": 0.3429933488368988, | |
| "eval_runtime": 4.7367, | |
| "eval_samples_per_second": 71.146, | |
| "eval_steps_per_second": 9.078, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "eval_accuracy": 0.8516320474777448, | |
| "eval_loss": 0.347034752368927, | |
| "eval_runtime": 4.7735, | |
| "eval_samples_per_second": 70.598, | |
| "eval_steps_per_second": 9.008, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 61.0, | |
| "eval_accuracy": 0.8545994065281899, | |
| "eval_loss": 0.34952038526535034, | |
| "eval_runtime": 4.6627, | |
| "eval_samples_per_second": 72.276, | |
| "eval_steps_per_second": 9.222, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 62.0, | |
| "eval_accuracy": 0.857566765578635, | |
| "eval_loss": 0.34523066878318787, | |
| "eval_runtime": 4.761, | |
| "eval_samples_per_second": 70.783, | |
| "eval_steps_per_second": 9.032, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 62.5, | |
| "grad_norm": 4.032630920410156, | |
| "learning_rate": 3.8709677419354837e-07, | |
| "loss": 0.2134, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 63.0, | |
| "eval_accuracy": 0.8605341246290801, | |
| "eval_loss": 0.3458469808101654, | |
| "eval_runtime": 4.7825, | |
| "eval_samples_per_second": 70.465, | |
| "eval_steps_per_second": 8.991, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 64.0, | |
| "eval_accuracy": 0.8605341246290801, | |
| "eval_loss": 0.3463592529296875, | |
| "eval_runtime": 4.7675, | |
| "eval_samples_per_second": 70.687, | |
| "eval_steps_per_second": 9.019, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 65.0, | |
| "eval_accuracy": 0.8605341246290801, | |
| "eval_loss": 0.3484276831150055, | |
| "eval_runtime": 4.7586, | |
| "eval_samples_per_second": 70.82, | |
| "eval_steps_per_second": 9.036, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 66.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.34886136651039124, | |
| "eval_runtime": 4.7061, | |
| "eval_samples_per_second": 71.609, | |
| "eval_steps_per_second": 9.137, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 67.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.3485889434814453, | |
| "eval_runtime": 4.6029, | |
| "eval_samples_per_second": 73.215, | |
| "eval_steps_per_second": 9.342, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 68.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.3503687381744385, | |
| "eval_runtime": 4.7013, | |
| "eval_samples_per_second": 71.682, | |
| "eval_steps_per_second": 9.146, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 69.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.3495609164237976, | |
| "eval_runtime": 4.7225, | |
| "eval_samples_per_second": 71.36, | |
| "eval_steps_per_second": 9.105, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 70.0, | |
| "eval_accuracy": 0.8605341246290801, | |
| "eval_loss": 0.3446400761604309, | |
| "eval_runtime": 4.7445, | |
| "eval_samples_per_second": 71.029, | |
| "eval_steps_per_second": 9.063, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 71.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.3452640175819397, | |
| "eval_runtime": 4.6434, | |
| "eval_samples_per_second": 72.577, | |
| "eval_steps_per_second": 9.261, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 72.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.34035128355026245, | |
| "eval_runtime": 4.7053, | |
| "eval_samples_per_second": 71.622, | |
| "eval_steps_per_second": 9.139, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 73.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.33860069513320923, | |
| "eval_runtime": 4.6174, | |
| "eval_samples_per_second": 72.985, | |
| "eval_steps_per_second": 9.313, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 74.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.3384004831314087, | |
| "eval_runtime": 4.663, | |
| "eval_samples_per_second": 72.271, | |
| "eval_steps_per_second": 9.221, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 75.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.33985474705696106, | |
| "eval_runtime": 4.7211, | |
| "eval_samples_per_second": 71.382, | |
| "eval_steps_per_second": 9.108, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 76.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.3414258062839508, | |
| "eval_runtime": 4.7314, | |
| "eval_samples_per_second": 71.226, | |
| "eval_steps_per_second": 9.088, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 77.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.3427448272705078, | |
| "eval_runtime": 4.7038, | |
| "eval_samples_per_second": 71.644, | |
| "eval_steps_per_second": 9.141, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 78.0, | |
| "eval_accuracy": 0.8635014836795252, | |
| "eval_loss": 0.34361016750335693, | |
| "eval_runtime": 4.8723, | |
| "eval_samples_per_second": 69.167, | |
| "eval_steps_per_second": 8.825, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 79.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.34205198287963867, | |
| "eval_runtime": 4.7564, | |
| "eval_samples_per_second": 70.851, | |
| "eval_steps_per_second": 9.04, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.3414119780063629, | |
| "eval_runtime": 4.7639, | |
| "eval_samples_per_second": 70.74, | |
| "eval_steps_per_second": 9.026, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 81.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.3397296071052551, | |
| "eval_runtime": 4.7085, | |
| "eval_samples_per_second": 71.572, | |
| "eval_steps_per_second": 9.132, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 82.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.3387517035007477, | |
| "eval_runtime": 4.764, | |
| "eval_samples_per_second": 70.738, | |
| "eval_steps_per_second": 9.026, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 83.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.33925285935401917, | |
| "eval_runtime": 4.8225, | |
| "eval_samples_per_second": 69.88, | |
| "eval_steps_per_second": 8.916, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 84.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.3406101167201996, | |
| "eval_runtime": 4.6734, | |
| "eval_samples_per_second": 72.11, | |
| "eval_steps_per_second": 9.201, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 85.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.33851388096809387, | |
| "eval_runtime": 4.7413, | |
| "eval_samples_per_second": 71.077, | |
| "eval_steps_per_second": 9.069, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 86.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.3385758697986603, | |
| "eval_runtime": 4.707, | |
| "eval_samples_per_second": 71.596, | |
| "eval_steps_per_second": 9.135, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 87.0, | |
| "eval_accuracy": 0.8664688427299704, | |
| "eval_loss": 0.33774590492248535, | |
| "eval_runtime": 4.6381, | |
| "eval_samples_per_second": 72.658, | |
| "eval_steps_per_second": 9.271, | |
| "step": 1392 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 1600, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 100, | |
| "save_steps": 500, | |
| "total_flos": 3.4046105637170995e+18, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |