| { | |
| "best_metric": 0.3088998794555664, | |
| "best_model_checkpoint": "bert-base-uncased-Regression-Edmunds_Car_Reviews/checkpoint-34290", | |
| "epoch": 3.0, | |
| "global_step": 34290, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.999941673957422e-05, | |
| "loss": 16.5207, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.988334791484398e-05, | |
| "loss": 1.4237, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9766695829687957e-05, | |
| "loss": 0.3652, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9650043744531936e-05, | |
| "loss": 0.3355, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9533391659375912e-05, | |
| "loss": 0.2945, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.941673957421989e-05, | |
| "loss": 0.2863, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9300087489063867e-05, | |
| "loss": 0.3029, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9183435403907846e-05, | |
| "loss": 0.3111, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9066783318751826e-05, | |
| "loss": 0.2634, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.89501312335958e-05, | |
| "loss": 0.278, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.883347914843978e-05, | |
| "loss": 0.2729, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8716827063283757e-05, | |
| "loss": 0.2829, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8600174978127736e-05, | |
| "loss": 0.2461, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8483522892971712e-05, | |
| "loss": 0.2497, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.836687080781569e-05, | |
| "loss": 0.2592, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.825021872265967e-05, | |
| "loss": 0.2701, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.813356663750365e-05, | |
| "loss": 0.2699, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.8016914552347625e-05, | |
| "loss": 0.2656, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.79002624671916e-05, | |
| "loss": 0.2521, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.778361038203558e-05, | |
| "loss": 0.2776, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.7666958296879556e-05, | |
| "loss": 0.2753, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.7550306211723536e-05, | |
| "loss": 0.2268, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.7433654126567515e-05, | |
| "loss": 0.2617, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7317002041411494e-05, | |
| "loss": 0.2425, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.720034995625547e-05, | |
| "loss": 0.2474, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.7083697871099446e-05, | |
| "loss": 0.2589, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.6967045785943425e-05, | |
| "loss": 0.2389, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.68503937007874e-05, | |
| "loss": 0.2494, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.673374161563138e-05, | |
| "loss": 0.2558, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.661708953047536e-05, | |
| "loss": 0.2524, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.650043744531934e-05, | |
| "loss": 0.2509, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.6383785360163315e-05, | |
| "loss": 0.2605, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.626713327500729e-05, | |
| "loss": 0.2354, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.615048118985127e-05, | |
| "loss": 0.27, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.603382910469525e-05, | |
| "loss": 0.2453, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.5917177019539225e-05, | |
| "loss": 0.2589, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5800524934383204e-05, | |
| "loss": 0.2513, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.5683872849227184e-05, | |
| "loss": 0.2583, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.556722076407116e-05, | |
| "loss": 0.257, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.5450568678915135e-05, | |
| "loss": 0.2363, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.5333916593759115e-05, | |
| "loss": 0.2378, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.5217264508603094e-05, | |
| "loss": 0.2438, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.510061242344707e-05, | |
| "loss": 0.2489, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.4983960338291049e-05, | |
| "loss": 0.2618, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.4867308253135026e-05, | |
| "loss": 0.2264, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.4750656167979002e-05, | |
| "loss": 0.2595, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.4634004082822982e-05, | |
| "loss": 0.2246, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.451735199766696e-05, | |
| "loss": 0.2281, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4400699912510938e-05, | |
| "loss": 0.217, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4284047827354914e-05, | |
| "loss": 0.2359, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4167395742198894e-05, | |
| "loss": 0.2322, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4050743657042871e-05, | |
| "loss": 0.2055, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.393409157188685e-05, | |
| "loss": 0.2374, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3817439486730826e-05, | |
| "loss": 0.2386, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3700787401574804e-05, | |
| "loss": 0.2323, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.3584135316418783e-05, | |
| "loss": 0.2362, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.3467483231262759e-05, | |
| "loss": 0.2211, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.3350831146106738e-05, | |
| "loss": 0.2142, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.24207346141338348, | |
| "eval_mae": 0.3126428425312042, | |
| "eval_mse": 0.24207346141338348, | |
| "eval_rmse": 0.49200960993766785, | |
| "eval_runtime": 434.9139, | |
| "eval_samples_per_second": 113.204, | |
| "eval_steps_per_second": 14.152, | |
| "step": 11430 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.3234179060950716e-05, | |
| "loss": 0.194, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.3117526975794695e-05, | |
| "loss": 0.1782, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.3000874890638671e-05, | |
| "loss": 0.1992, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.2884222805482649e-05, | |
| "loss": 0.1657, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.2767570720326628e-05, | |
| "loss": 0.2156, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.2650918635170604e-05, | |
| "loss": 0.1742, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.2534266550014583e-05, | |
| "loss": 0.156, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.241761446485856e-05, | |
| "loss": 0.2079, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.230096237970254e-05, | |
| "loss": 0.1635, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.2184310294546516e-05, | |
| "loss": 0.1987, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.2067658209390493e-05, | |
| "loss": 0.1683, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.1951006124234472e-05, | |
| "loss": 0.1735, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.183435403907845e-05, | |
| "loss": 0.1971, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.1717701953922428e-05, | |
| "loss": 0.2014, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.1601049868766405e-05, | |
| "loss": 0.2114, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.1484397783610384e-05, | |
| "loss": 0.1707, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.136774569845436e-05, | |
| "loss": 0.194, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.1251093613298338e-05, | |
| "loss": 0.1714, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.1134441528142317e-05, | |
| "loss": 0.1973, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.1017789442986295e-05, | |
| "loss": 0.1705, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.0901137357830272e-05, | |
| "loss": 0.1842, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.078448527267425e-05, | |
| "loss": 0.1578, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.0667833187518229e-05, | |
| "loss": 0.1598, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.0551181102362205e-05, | |
| "loss": 0.1895, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.0434529017206183e-05, | |
| "loss": 0.1846, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.0317876932050162e-05, | |
| "loss": 0.1875, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.020122484689414e-05, | |
| "loss": 0.1748, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.0084572761738117e-05, | |
| "loss": 0.1837, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 9.967920676582095e-06, | |
| "loss": 0.1888, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 9.851268591426074e-06, | |
| "loss": 0.2038, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 9.73461650627005e-06, | |
| "loss": 0.1736, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 9.617964421114027e-06, | |
| "loss": 0.1853, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 9.501312335958006e-06, | |
| "loss": 0.1762, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 9.384660250801984e-06, | |
| "loss": 0.1979, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 9.268008165645962e-06, | |
| "loss": 0.1747, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 9.15135608048994e-06, | |
| "loss": 0.1793, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.034703995333918e-06, | |
| "loss": 0.1788, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 8.918051910177894e-06, | |
| "loss": 0.1901, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 8.801399825021874e-06, | |
| "loss": 0.1755, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 8.684747739865851e-06, | |
| "loss": 0.1782, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 8.568095654709829e-06, | |
| "loss": 0.1928, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 8.451443569553806e-06, | |
| "loss": 0.1821, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 8.334791484397784e-06, | |
| "loss": 0.1836, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 8.218139399241763e-06, | |
| "loss": 0.1803, | |
| "step": 20200 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 8.101487314085739e-06, | |
| "loss": 0.176, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 7.984835228929718e-06, | |
| "loss": 0.2068, | |
| "step": 20600 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 7.868183143773696e-06, | |
| "loss": 0.2005, | |
| "step": 20800 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 7.751531058617673e-06, | |
| "loss": 0.1706, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 7.634878973461651e-06, | |
| "loss": 0.1821, | |
| "step": 21200 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 7.5182268883056285e-06, | |
| "loss": 0.2003, | |
| "step": 21400 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 7.401574803149607e-06, | |
| "loss": 0.1839, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 7.2849227179935845e-06, | |
| "loss": 0.191, | |
| "step": 21800 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 7.168270632837563e-06, | |
| "loss": 0.2031, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 7.0516185476815405e-06, | |
| "loss": 0.2054, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 6.934966462525519e-06, | |
| "loss": 0.1687, | |
| "step": 22400 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 6.818314377369496e-06, | |
| "loss": 0.176, | |
| "step": 22600 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 6.701662292213474e-06, | |
| "loss": 0.1931, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.2529958486557007, | |
| "eval_mae": 0.3336344063282013, | |
| "eval_mse": 0.2529958486557007, | |
| "eval_rmse": 0.5029869079589844, | |
| "eval_runtime": 433.6838, | |
| "eval_samples_per_second": 113.525, | |
| "eval_steps_per_second": 14.192, | |
| "step": 22860 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 6.585010207057452e-06, | |
| "loss": 0.1951, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 6.468358121901429e-06, | |
| "loss": 0.127, | |
| "step": 23200 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.351706036745408e-06, | |
| "loss": 0.1387, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 6.235053951589385e-06, | |
| "loss": 0.1436, | |
| "step": 23600 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 6.118401866433364e-06, | |
| "loss": 0.1446, | |
| "step": 23800 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 6.00174978127734e-06, | |
| "loss": 0.1315, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 5.885097696121319e-06, | |
| "loss": 0.1489, | |
| "step": 24200 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 5.768445610965296e-06, | |
| "loss": 0.1527, | |
| "step": 24400 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 5.651793525809274e-06, | |
| "loss": 0.1516, | |
| "step": 24600 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 5.535141440653252e-06, | |
| "loss": 0.1203, | |
| "step": 24800 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 5.41848935549723e-06, | |
| "loss": 0.1204, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 5.301837270341208e-06, | |
| "loss": 0.15, | |
| "step": 25200 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 5.185185185185185e-06, | |
| "loss": 0.1453, | |
| "step": 25400 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 5.068533100029163e-06, | |
| "loss": 0.1406, | |
| "step": 25600 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 4.951881014873141e-06, | |
| "loss": 0.1263, | |
| "step": 25800 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 4.835228929717119e-06, | |
| "loss": 0.1187, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 4.718576844561097e-06, | |
| "loss": 0.1363, | |
| "step": 26200 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 4.6019247594050745e-06, | |
| "loss": 0.1284, | |
| "step": 26400 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 4.485272674249053e-06, | |
| "loss": 0.1371, | |
| "step": 26600 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.3686205890930305e-06, | |
| "loss": 0.128, | |
| "step": 26800 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 4.251968503937008e-06, | |
| "loss": 0.1246, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 4.135316418780986e-06, | |
| "loss": 0.1283, | |
| "step": 27200 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 4.018664333624964e-06, | |
| "loss": 0.1382, | |
| "step": 27400 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 3.902012248468942e-06, | |
| "loss": 0.1324, | |
| "step": 27600 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 3.7853601633129196e-06, | |
| "loss": 0.1305, | |
| "step": 27800 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 3.668708078156897e-06, | |
| "loss": 0.1319, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 3.552055993000875e-06, | |
| "loss": 0.1322, | |
| "step": 28200 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 3.435403907844853e-06, | |
| "loss": 0.1424, | |
| "step": 28400 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 3.318751822688831e-06, | |
| "loss": 0.1473, | |
| "step": 28600 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 3.2020997375328087e-06, | |
| "loss": 0.1361, | |
| "step": 28800 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 3.0854476523767863e-06, | |
| "loss": 0.128, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 2.9687955672207643e-06, | |
| "loss": 0.1212, | |
| "step": 29200 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 2.852143482064742e-06, | |
| "loss": 0.13, | |
| "step": 29400 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 2.73549139690872e-06, | |
| "loss": 0.1401, | |
| "step": 29600 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 2.618839311752698e-06, | |
| "loss": 0.1412, | |
| "step": 29800 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 2.502187226596676e-06, | |
| "loss": 0.1262, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 2.3855351414406534e-06, | |
| "loss": 0.1412, | |
| "step": 30200 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 2.2688830562846314e-06, | |
| "loss": 0.1441, | |
| "step": 30400 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 2.152230971128609e-06, | |
| "loss": 0.1443, | |
| "step": 30600 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 2.035578885972587e-06, | |
| "loss": 0.1541, | |
| "step": 30800 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 1.918926800816565e-06, | |
| "loss": 0.1187, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 1.8022747156605425e-06, | |
| "loss": 0.1254, | |
| "step": 31200 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 1.6856226305045203e-06, | |
| "loss": 0.1403, | |
| "step": 31400 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 1.5689705453484983e-06, | |
| "loss": 0.1272, | |
| "step": 31600 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 1.452318460192476e-06, | |
| "loss": 0.1223, | |
| "step": 31800 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 1.335666375036454e-06, | |
| "loss": 0.1448, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 1.2190142898804318e-06, | |
| "loss": 0.1376, | |
| "step": 32200 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 1.1023622047244096e-06, | |
| "loss": 0.1246, | |
| "step": 32400 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 9.857101195683874e-07, | |
| "loss": 0.1548, | |
| "step": 32600 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 8.690580344123652e-07, | |
| "loss": 0.1408, | |
| "step": 32800 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 7.524059492563431e-07, | |
| "loss": 0.1282, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 6.357538641003208e-07, | |
| "loss": 0.1552, | |
| "step": 33200 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 5.191017789442987e-07, | |
| "loss": 0.1272, | |
| "step": 33400 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 4.024496937882765e-07, | |
| "loss": 0.1166, | |
| "step": 33600 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 2.8579760863225433e-07, | |
| "loss": 0.1245, | |
| "step": 33800 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 1.6914552347623217e-07, | |
| "loss": 0.1233, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 5.249343832020998e-08, | |
| "loss": 0.1192, | |
| "step": 34200 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 0.23235057294368744, | |
| "eval_mae": 0.3088998794555664, | |
| "eval_mse": 0.23235058784484863, | |
| "eval_rmse": 0.4820275902748108, | |
| "eval_runtime": 434.9142, | |
| "eval_samples_per_second": 113.204, | |
| "eval_steps_per_second": 14.152, | |
| "step": 34290 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 34290, | |
| "total_flos": 1.1316755048115336e+16, | |
| "train_loss": 0.19905866315323814, | |
| "train_runtime": 6850.4631, | |
| "train_samples_per_second": 40.041, | |
| "train_steps_per_second": 5.006 | |
| } | |
| ], | |
| "max_steps": 34290, | |
| "num_train_epochs": 3, | |
| "total_flos": 1.1316755048115336e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |