diff --git "a/trainer_state.json" "b/trainer_state.json" --- "a/trainer_state.json" +++ "b/trainer_state.json" @@ -1,6 +1,6 @@ { - "best_metric": 0.5539772727272727, - "best_model_checkpoint": "videomae-base-finetuned-ucf101-subset/checkpoint-1600", + "best_metric": 0.514367816091954, + "best_model_checkpoint": "videomae-base-finetuned-ucf101-subset/checkpoint-1824", "epoch": 119.00833333333334, "eval_steps": 500, "global_step": 1920, @@ -10,2453 +10,2453 @@ "log_history": [ { "epoch": 0.005208333333333333, - "grad_norm": 3.7305831909179688, + "grad_norm": 5.254432678222656, "learning_rate": 2.604166666666667e-06, - "loss": 4.2451, + "loss": 4.2685, "step": 10 }, { "epoch": 0.008333333333333333, - "eval_accuracy": 0.01065340909090909, - "eval_loss": 4.228378772735596, - "eval_runtime": 133.9626, - "eval_samples_per_second": 10.51, - "eval_steps_per_second": 0.164, + "eval_accuracy": 0.014367816091954023, + "eval_loss": 4.244288921356201, + "eval_runtime": 144.9444, + "eval_samples_per_second": 9.604, + "eval_steps_per_second": 0.152, "step": 16 }, { "epoch": 1.0020833333333334, - "grad_norm": 3.0969808101654053, + "grad_norm": 4.573583602905273, "learning_rate": 5.208333333333334e-06, - "loss": 4.231, + "loss": 4.2711, "step": 20 }, { "epoch": 1.0072916666666667, - "grad_norm": 4.606812000274658, + "grad_norm": 6.3157782554626465, "learning_rate": 7.8125e-06, - "loss": 4.2251, + "loss": 4.2462, "step": 30 }, { "epoch": 1.0083333333333333, - "eval_accuracy": 0.01065340909090909, - "eval_loss": 4.215222358703613, - "eval_runtime": 129.5768, - "eval_samples_per_second": 10.866, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.016522988505747127, + "eval_loss": 4.222896575927734, + "eval_runtime": 126.0729, + "eval_samples_per_second": 11.041, + "eval_steps_per_second": 0.175, "step": 32 }, { "epoch": 2.004166666666667, - "grad_norm": 3.574765682220459, + "grad_norm": 5.368467330932617, "learning_rate": 1.0416666666666668e-05, - "loss": 4.2276, + "loss": 4.2193, "step": 40 }, { "epoch": 2.0083333333333333, - "eval_accuracy": 0.012073863636363636, - "eval_loss": 4.209568500518799, - "eval_runtime": 130.2735, - "eval_samples_per_second": 10.808, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.01293103448275862, + "eval_loss": 4.211052417755127, + "eval_runtime": 156.4771, + "eval_samples_per_second": 8.896, + "eval_steps_per_second": 0.141, "step": 48 }, { "epoch": 3.0010416666666666, - "grad_norm": 4.630516052246094, + "grad_norm": 4.4701738357543945, "learning_rate": 1.3020833333333334e-05, - "loss": 4.2126, + "loss": 4.2291, "step": 50 }, { "epoch": 3.00625, - "grad_norm": 3.0715420246124268, + "grad_norm": 3.1355085372924805, "learning_rate": 1.5625e-05, - "loss": 4.2146, + "loss": 4.2251, "step": 60 }, { "epoch": 3.0083333333333333, - "eval_accuracy": 0.014914772727272728, - "eval_loss": 4.212440490722656, - "eval_runtime": 128.256, - "eval_samples_per_second": 10.978, - "eval_steps_per_second": 0.172, + "eval_accuracy": 0.013649425287356323, + "eval_loss": 4.212449550628662, + "eval_runtime": 164.2057, + "eval_samples_per_second": 8.477, + "eval_steps_per_second": 0.134, "step": 64 }, { "epoch": 4.003125, - "grad_norm": 2.934682846069336, + "grad_norm": 3.2147018909454346, "learning_rate": 1.8229166666666668e-05, - "loss": 4.2235, + "loss": 4.2197, "step": 70 }, { "epoch": 4.008333333333334, - "grad_norm": 8.70511245727539, + "grad_norm": 11.049864768981934, "learning_rate": 2.0833333333333336e-05, - "loss": 4.2217, + "loss": 4.231, "step": 80 }, { "epoch": 4.008333333333334, - "eval_accuracy": 0.01775568181818182, - "eval_loss": 4.2041802406311035, - "eval_runtime": 126.2921, - "eval_samples_per_second": 11.149, - "eval_steps_per_second": 0.174, + "eval_accuracy": 0.015804597701149427, + "eval_loss": 4.213141918182373, + "eval_runtime": 115.2514, + "eval_samples_per_second": 12.078, + "eval_steps_per_second": 0.191, "step": 80 }, { "epoch": 5.005208333333333, - "grad_norm": 2.5297322273254395, + "grad_norm": 2.9415507316589355, "learning_rate": 2.34375e-05, - "loss": 4.2091, + "loss": 4.2197, "step": 90 }, { "epoch": 5.008333333333334, - "eval_accuracy": 0.0234375, - "eval_loss": 4.205115795135498, - "eval_runtime": 127.2102, - "eval_samples_per_second": 11.068, - "eval_steps_per_second": 0.173, + "eval_accuracy": 0.015086206896551725, + "eval_loss": 4.20913553237915, + "eval_runtime": 132.3001, + "eval_samples_per_second": 10.522, + "eval_steps_per_second": 0.166, "step": 96 }, { "epoch": 6.002083333333333, - "grad_norm": 2.4549720287323, + "grad_norm": 2.928065776824951, "learning_rate": 2.604166666666667e-05, - "loss": 4.2256, + "loss": 4.2237, "step": 100 }, { "epoch": 6.007291666666666, - "grad_norm": 2.3255746364593506, + "grad_norm": 2.455629587173462, "learning_rate": 2.8645833333333333e-05, - "loss": 4.2085, + "loss": 4.2239, "step": 110 }, { "epoch": 6.008333333333334, - "eval_accuracy": 0.018465909090909092, - "eval_loss": 4.193885803222656, - "eval_runtime": 130.9427, - "eval_samples_per_second": 10.753, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.010775862068965518, + "eval_loss": 4.2099738121032715, + "eval_runtime": 137.9002, + "eval_samples_per_second": 10.094, + "eval_steps_per_second": 0.16, "step": 112 }, { "epoch": 7.004166666666666, - "grad_norm": 2.676177978515625, + "grad_norm": 2.583024024963379, "learning_rate": 3.125e-05, - "loss": 4.2044, + "loss": 4.2246, "step": 120 }, { "epoch": 7.008333333333334, - "eval_accuracy": 0.0390625, - "eval_loss": 4.1791582107543945, - "eval_runtime": 125.2796, - "eval_samples_per_second": 11.239, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.014367816091954023, + "eval_loss": 4.211906909942627, + "eval_runtime": 131.9857, + "eval_samples_per_second": 10.547, + "eval_steps_per_second": 0.167, "step": 128 }, { "epoch": 8.001041666666667, - "grad_norm": 2.616762161254883, + "grad_norm": 2.4764182567596436, "learning_rate": 3.385416666666667e-05, - "loss": 4.1911, + "loss": 4.2207, "step": 130 }, { "epoch": 8.00625, - "grad_norm": 2.983841896057129, + "grad_norm": 2.181648015975952, "learning_rate": 3.6458333333333336e-05, - "loss": 4.1624, + "loss": 4.2131, "step": 140 }, { "epoch": 8.008333333333333, - "eval_accuracy": 0.022017045454545456, - "eval_loss": 4.201480388641357, - "eval_runtime": 136.9987, - "eval_samples_per_second": 10.277, - "eval_steps_per_second": 0.161, + "eval_accuracy": 0.017241379310344827, + "eval_loss": 4.203038215637207, + "eval_runtime": 161.0561, + "eval_samples_per_second": 8.643, + "eval_steps_per_second": 0.137, "step": 144 }, { "epoch": 9.003125, - "grad_norm": 2.874126434326172, + "grad_norm": 2.136594533920288, "learning_rate": 3.90625e-05, - "loss": 4.1349, + "loss": 4.2064, "step": 150 }, { "epoch": 9.008333333333333, - "grad_norm": 10.345636367797852, + "grad_norm": 7.214341163635254, "learning_rate": 4.166666666666667e-05, - "loss": 4.1253, + "loss": 4.236, "step": 160 }, { "epoch": 9.008333333333333, - "eval_accuracy": 0.029829545454545456, - "eval_loss": 4.1215105056762695, - "eval_runtime": 132.9763, - "eval_samples_per_second": 10.588, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.01293103448275862, + "eval_loss": 4.197862148284912, + "eval_runtime": 121.1445, + "eval_samples_per_second": 11.49, + "eval_steps_per_second": 0.182, "step": 160 }, { "epoch": 10.005208333333334, - "grad_norm": 3.5383424758911133, + "grad_norm": 2.4436376094818115, "learning_rate": 4.4270833333333337e-05, - "loss": 4.0308, + "loss": 4.177, "step": 170 }, { "epoch": 10.008333333333333, - "eval_accuracy": 0.07102272727272728, - "eval_loss": 4.002456188201904, - "eval_runtime": 127.9472, - "eval_samples_per_second": 11.005, - "eval_steps_per_second": 0.172, + "eval_accuracy": 0.032327586206896554, + "eval_loss": 4.132750511169434, + "eval_runtime": 132.3308, + "eval_samples_per_second": 10.519, + "eval_steps_per_second": 0.166, "step": 176 }, { "epoch": 11.002083333333333, - "grad_norm": 4.1187543869018555, + "grad_norm": 3.184831142425537, "learning_rate": 4.6875e-05, - "loss": 3.9438, + "loss": 4.1437, "step": 180 }, { "epoch": 11.007291666666667, - "grad_norm": 4.370882034301758, + "grad_norm": 4.086268901824951, "learning_rate": 4.947916666666667e-05, - "loss": 3.8065, + "loss": 4.0656, "step": 190 }, { "epoch": 11.008333333333333, - "eval_accuracy": 0.08380681818181818, - "eval_loss": 3.872328758239746, - "eval_runtime": 142.2129, - "eval_samples_per_second": 9.901, - "eval_steps_per_second": 0.155, + "eval_accuracy": 0.031609195402298854, + "eval_loss": 4.0372843742370605, + "eval_runtime": 128.8196, + "eval_samples_per_second": 10.806, + "eval_steps_per_second": 0.171, "step": 192 }, { "epoch": 12.004166666666666, - "grad_norm": 3.4101152420043945, + "grad_norm": 4.314004421234131, "learning_rate": 4.976851851851852e-05, - "loss": 3.7614, + "loss": 3.9339, "step": 200 }, { "epoch": 12.008333333333333, - "eval_accuracy": 0.09943181818181818, - "eval_loss": 3.7138783931732178, - "eval_runtime": 125.568, - "eval_samples_per_second": 11.213, - "eval_steps_per_second": 0.175, + "eval_accuracy": 0.05028735632183908, + "eval_loss": 3.884765386581421, + "eval_runtime": 129.9723, + "eval_samples_per_second": 10.71, + "eval_steps_per_second": 0.169, "step": 208 }, { "epoch": 13.001041666666667, - "grad_norm": 4.585361480712891, + "grad_norm": 4.494894027709961, "learning_rate": 4.947916666666667e-05, - "loss": 3.6302, + "loss": 3.8696, "step": 210 }, { "epoch": 13.00625, - "grad_norm": 4.209935665130615, + "grad_norm": 3.506269931793213, "learning_rate": 4.9189814814814815e-05, - "loss": 3.4761, + "loss": 3.7197, "step": 220 }, { "epoch": 13.008333333333333, - "eval_accuracy": 0.1434659090909091, - "eval_loss": 3.6160459518432617, - "eval_runtime": 133.4296, - "eval_samples_per_second": 10.552, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.07830459770114942, + "eval_loss": 3.768052816390991, + "eval_runtime": 122.2061, + "eval_samples_per_second": 11.391, + "eval_steps_per_second": 0.18, "step": 224 }, { "epoch": 14.003125, - "grad_norm": 4.6234049797058105, + "grad_norm": 3.8743553161621094, "learning_rate": 4.8900462962962965e-05, - "loss": 3.3764, + "loss": 3.6495, "step": 230 }, { "epoch": 14.008333333333333, - "grad_norm": 13.464088439941406, + "grad_norm": 13.633111000061035, "learning_rate": 4.8611111111111115e-05, - "loss": 3.278, + "loss": 3.5657, "step": 240 }, { "epoch": 14.008333333333333, - "eval_accuracy": 0.1924715909090909, - "eval_loss": 3.3939576148986816, - "eval_runtime": 129.3255, - "eval_samples_per_second": 10.887, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.11853448275862069, + "eval_loss": 3.5790250301361084, + "eval_runtime": 131.0036, + "eval_samples_per_second": 10.626, + "eval_steps_per_second": 0.168, "step": 240 }, { "epoch": 15.005208333333334, - "grad_norm": 4.547211647033691, + "grad_norm": 3.9671919345855713, "learning_rate": 4.8321759259259265e-05, - "loss": 3.0999, + "loss": 3.3108, "step": 250 }, { "epoch": 15.008333333333333, - "eval_accuracy": 0.2080965909090909, - "eval_loss": 3.3182637691497803, - "eval_runtime": 123.805, - "eval_samples_per_second": 11.373, - "eval_steps_per_second": 0.178, + "eval_accuracy": 0.14655172413793102, + "eval_loss": 3.564199924468994, + "eval_runtime": 143.9145, + "eval_samples_per_second": 9.672, + "eval_steps_per_second": 0.153, "step": 256 }, { "epoch": 16.002083333333335, - "grad_norm": 4.825346946716309, + "grad_norm": 4.458219528198242, "learning_rate": 4.803240740740741e-05, - "loss": 3.0213, + "loss": 3.2608, "step": 260 }, { "epoch": 16.007291666666667, - "grad_norm": 5.93734884262085, + "grad_norm": 4.314128398895264, "learning_rate": 4.774305555555556e-05, - "loss": 2.9721, + "loss": 3.1687, "step": 270 }, { "epoch": 16.008333333333333, - "eval_accuracy": 0.24857954545454544, - "eval_loss": 3.159555673599243, - "eval_runtime": 124.115, - "eval_samples_per_second": 11.344, - "eval_steps_per_second": 0.177, + "eval_accuracy": 0.18318965517241378, + "eval_loss": 3.32100772857666, + "eval_runtime": 118.6776, + "eval_samples_per_second": 11.729, + "eval_steps_per_second": 0.185, "step": 272 }, { "epoch": 17.004166666666666, - "grad_norm": 4.84201717376709, + "grad_norm": 4.7279372215271, "learning_rate": 4.745370370370371e-05, - "loss": 2.8064, + "loss": 3.1376, "step": 280 }, { "epoch": 17.008333333333333, - "eval_accuracy": 0.26704545454545453, - "eval_loss": 3.023179054260254, - "eval_runtime": 131.2545, - "eval_samples_per_second": 10.727, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.22916666666666666, + "eval_loss": 3.194972276687622, + "eval_runtime": 131.4719, + "eval_samples_per_second": 10.588, + "eval_steps_per_second": 0.167, "step": 288 }, { "epoch": 18.001041666666666, - "grad_norm": 5.273012161254883, + "grad_norm": 6.386988162994385, "learning_rate": 4.716435185185186e-05, - "loss": 2.813, + "loss": 3.0321, "step": 290 }, { "epoch": 18.00625, - "grad_norm": 5.6201276779174805, + "grad_norm": 5.669564247131348, "learning_rate": 4.6875e-05, - "loss": 2.6554, + "loss": 2.8366, "step": 300 }, { "epoch": 18.008333333333333, - "eval_accuracy": 0.2911931818181818, - "eval_loss": 2.9448060989379883, - "eval_runtime": 123.4251, - "eval_samples_per_second": 11.408, - "eval_steps_per_second": 0.178, + "eval_accuracy": 0.2492816091954023, + "eval_loss": 3.1269867420196533, + "eval_runtime": 113.3706, + "eval_samples_per_second": 12.278, + "eval_steps_per_second": 0.194, "step": 304 }, { "epoch": 19.003125, - "grad_norm": 7.461442470550537, + "grad_norm": 6.121387004852295, "learning_rate": 4.658564814814815e-05, - "loss": 2.6668, + "loss": 2.8072, "step": 310 }, { "epoch": 19.008333333333333, - "grad_norm": 14.555140495300293, + "grad_norm": 13.984885215759277, "learning_rate": 4.62962962962963e-05, - "loss": 2.5052, + "loss": 2.6811, "step": 320 }, { "epoch": 19.008333333333333, - "eval_accuracy": 0.3309659090909091, - "eval_loss": 2.828484535217285, - "eval_runtime": 127.0838, - "eval_samples_per_second": 11.079, - "eval_steps_per_second": 0.173, + "eval_accuracy": 0.28951149425287354, + "eval_loss": 2.999732732772827, + "eval_runtime": 120.8139, + "eval_samples_per_second": 11.522, + "eval_steps_per_second": 0.182, "step": 320 }, { "epoch": 20.005208333333332, - "grad_norm": 6.753479957580566, + "grad_norm": 5.307132244110107, "learning_rate": 4.6006944444444444e-05, - "loss": 2.4322, + "loss": 2.6104, "step": 330 }, { "epoch": 20.008333333333333, - "eval_accuracy": 0.34019886363636365, - "eval_loss": 2.7478535175323486, - "eval_runtime": 125.7771, - "eval_samples_per_second": 11.194, - "eval_steps_per_second": 0.175, + "eval_accuracy": 0.27729885057471265, + "eval_loss": 2.9776358604431152, + "eval_runtime": 144.4295, + "eval_samples_per_second": 9.638, + "eval_steps_per_second": 0.152, "step": 336 }, { "epoch": 21.002083333333335, - "grad_norm": 5.665611743927002, + "grad_norm": 5.0186638832092285, "learning_rate": 4.5717592592592594e-05, - "loss": 2.4376, + "loss": 2.5321, "step": 340 }, { "epoch": 21.007291666666667, - "grad_norm": 5.6526570320129395, + "grad_norm": 5.094930648803711, "learning_rate": 4.5428240740740744e-05, - "loss": 2.3193, + "loss": 2.5156, "step": 350 }, { "epoch": 21.008333333333333, - "eval_accuracy": 0.3309659090909091, - "eval_loss": 2.794116735458374, - "eval_runtime": 126.5702, - "eval_samples_per_second": 11.124, - "eval_steps_per_second": 0.174, + "eval_accuracy": 0.3125, + "eval_loss": 2.862489700317383, + "eval_runtime": 144.5549, + "eval_samples_per_second": 9.63, + "eval_steps_per_second": 0.152, "step": 352 }, { "epoch": 22.004166666666666, - "grad_norm": 6.312890529632568, + "grad_norm": 5.007638931274414, "learning_rate": 4.5138888888888894e-05, - "loss": 2.2565, + "loss": 2.3804, "step": 360 }, { "epoch": 22.008333333333333, - "eval_accuracy": 0.3671875, - "eval_loss": 2.6383402347564697, - "eval_runtime": 123.0345, - "eval_samples_per_second": 11.444, - "eval_steps_per_second": 0.179, + "eval_accuracy": 0.2974137931034483, + "eval_loss": 2.822291135787964, + "eval_runtime": 161.0637, + "eval_samples_per_second": 8.643, + "eval_steps_per_second": 0.137, "step": 368 }, { "epoch": 23.001041666666666, - "grad_norm": 5.865190505981445, + "grad_norm": 6.657947540283203, "learning_rate": 4.484953703703704e-05, - "loss": 2.1257, + "loss": 2.418, "step": 370 }, { "epoch": 23.00625, - "grad_norm": 5.813269138336182, + "grad_norm": 5.21594762802124, "learning_rate": 4.456018518518519e-05, - "loss": 2.1405, + "loss": 2.2389, "step": 380 }, { "epoch": 23.008333333333333, - "eval_accuracy": 0.36079545454545453, - "eval_loss": 2.5905861854553223, - "eval_runtime": 126.2698, - "eval_samples_per_second": 11.151, - "eval_steps_per_second": 0.174, + "eval_accuracy": 0.34123563218390807, + "eval_loss": 2.7041523456573486, + "eval_runtime": 163.8868, + "eval_samples_per_second": 8.494, + "eval_steps_per_second": 0.134, "step": 384 }, { "epoch": 24.003125, - "grad_norm": 6.6013569831848145, + "grad_norm": 6.3385009765625, "learning_rate": 4.4270833333333337e-05, - "loss": 1.9715, + "loss": 2.195, "step": 390 }, { "epoch": 24.008333333333333, - "grad_norm": 14.786352157592773, + "grad_norm": 14.863781929016113, "learning_rate": 4.3981481481481486e-05, - "loss": 2.1049, + "loss": 2.1889, "step": 400 }, { "epoch": 24.008333333333333, - "eval_accuracy": 0.37855113636363635, - "eval_loss": 2.5515265464782715, - "eval_runtime": 134.242, - "eval_samples_per_second": 10.489, - "eval_steps_per_second": 0.164, + "eval_accuracy": 0.3225574712643678, + "eval_loss": 2.6846468448638916, + "eval_runtime": 162.1045, + "eval_samples_per_second": 8.587, + "eval_steps_per_second": 0.136, "step": 400 }, { "epoch": 25.005208333333332, - "grad_norm": 6.399519920349121, + "grad_norm": 7.218128204345703, "learning_rate": 4.369212962962963e-05, - "loss": 1.8424, + "loss": 1.9829, "step": 410 }, { "epoch": 25.008333333333333, - "eval_accuracy": 0.39204545454545453, - "eval_loss": 2.4692277908325195, - "eval_runtime": 134.5245, - "eval_samples_per_second": 10.466, - "eval_steps_per_second": 0.164, + "eval_accuracy": 0.35919540229885055, + "eval_loss": 2.6066086292266846, + "eval_runtime": 143.5991, + "eval_samples_per_second": 9.694, + "eval_steps_per_second": 0.153, "step": 416 }, { "epoch": 26.002083333333335, - "grad_norm": 5.429991245269775, + "grad_norm": 5.6082763671875, "learning_rate": 4.340277777777778e-05, - "loss": 2.0616, + "loss": 2.1112, "step": 420 }, { "epoch": 26.007291666666667, - "grad_norm": 6.501383304595947, + "grad_norm": 5.779495716094971, "learning_rate": 4.311342592592593e-05, - "loss": 1.8685, + "loss": 1.9466, "step": 430 }, { "epoch": 26.008333333333333, - "eval_accuracy": 0.4275568181818182, - "eval_loss": 2.4325406551361084, - "eval_runtime": 138.3084, - "eval_samples_per_second": 10.18, - "eval_steps_per_second": 0.159, + "eval_accuracy": 0.3642241379310345, + "eval_loss": 2.584472417831421, + "eval_runtime": 134.6836, + "eval_samples_per_second": 10.335, + "eval_steps_per_second": 0.163, "step": 432 }, { "epoch": 27.004166666666666, - "grad_norm": 6.909026622772217, + "grad_norm": 5.747377395629883, "learning_rate": 4.282407407407408e-05, - "loss": 1.7478, + "loss": 1.8991, "step": 440 }, { "epoch": 27.008333333333333, - "eval_accuracy": 0.4147727272727273, - "eval_loss": 2.416501522064209, - "eval_runtime": 128.9504, - "eval_samples_per_second": 10.919, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.3922413793103448, + "eval_loss": 2.5150339603424072, + "eval_runtime": 160.9218, + "eval_samples_per_second": 8.65, + "eval_steps_per_second": 0.137, "step": 448 }, { "epoch": 28.001041666666666, - "grad_norm": 6.359575271606445, + "grad_norm": 5.8224639892578125, "learning_rate": 4.253472222222222e-05, - "loss": 2.0103, + "loss": 1.9699, "step": 450 }, { "epoch": 28.00625, - "grad_norm": 7.018951416015625, + "grad_norm": 6.3729329109191895, "learning_rate": 4.224537037037037e-05, - "loss": 1.7072, + "loss": 1.8629, "step": 460 }, { "epoch": 28.008333333333333, - "eval_accuracy": 0.4268465909090909, - "eval_loss": 2.3617048263549805, - "eval_runtime": 126.6826, - "eval_samples_per_second": 11.114, - "eval_steps_per_second": 0.174, + "eval_accuracy": 0.3994252873563218, + "eval_loss": 2.4958128929138184, + "eval_runtime": 133.0775, + "eval_samples_per_second": 10.46, + "eval_steps_per_second": 0.165, "step": 464 }, { "epoch": 29.003125, - "grad_norm": 5.399332523345947, + "grad_norm": 7.140518665313721, "learning_rate": 4.195601851851852e-05, - "loss": 1.8236, + "loss": 1.8899, "step": 470 }, { "epoch": 29.008333333333333, - "grad_norm": 14.481021881103516, + "grad_norm": 27.132600784301758, "learning_rate": 4.166666666666667e-05, - "loss": 1.7206, + "loss": 1.8563, "step": 480 }, { "epoch": 29.008333333333333, - "eval_accuracy": 0.4303977272727273, - "eval_loss": 2.372326135635376, - "eval_runtime": 128.5457, - "eval_samples_per_second": 10.953, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.3994252873563218, + "eval_loss": 2.503603219985962, + "eval_runtime": 146.1023, + "eval_samples_per_second": 9.528, + "eval_steps_per_second": 0.151, "step": 480 }, { "epoch": 30.005208333333332, - "grad_norm": 7.11575984954834, + "grad_norm": 6.655496120452881, "learning_rate": 4.1377314814814815e-05, - "loss": 1.693, + "loss": 1.832, "step": 490 }, { "epoch": 30.008333333333333, - "eval_accuracy": 0.4424715909090909, - "eval_loss": 2.2890071868896484, - "eval_runtime": 128.9784, - "eval_samples_per_second": 10.917, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.40373563218390807, + "eval_loss": 2.4212334156036377, + "eval_runtime": 144.6194, + "eval_samples_per_second": 9.625, + "eval_steps_per_second": 0.152, "step": 496 }, { "epoch": 31.002083333333335, - "grad_norm": 7.223865985870361, + "grad_norm": 7.301593780517578, "learning_rate": 4.1087962962962965e-05, - "loss": 1.6722, + "loss": 1.6729, "step": 500 }, { "epoch": 31.007291666666667, - "grad_norm": 7.608266353607178, + "grad_norm": 6.040745735168457, "learning_rate": 4.0798611111111115e-05, - "loss": 1.6347, + "loss": 1.7148, "step": 510 }, { "epoch": 31.008333333333333, - "eval_accuracy": 0.44105113636363635, - "eval_loss": 2.244246482849121, - "eval_runtime": 129.1752, - "eval_samples_per_second": 10.9, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.42528735632183906, + "eval_loss": 2.389084815979004, + "eval_runtime": 141.2811, + "eval_samples_per_second": 9.853, + "eval_steps_per_second": 0.156, "step": 512 }, { "epoch": 32.00416666666667, - "grad_norm": 7.774643898010254, + "grad_norm": 6.346589088439941, "learning_rate": 4.0509259259259265e-05, - "loss": 1.5276, + "loss": 1.6525, "step": 520 }, { "epoch": 32.00833333333333, - "eval_accuracy": 0.46732954545454547, - "eval_loss": 2.210369110107422, - "eval_runtime": 120.0766, - "eval_samples_per_second": 11.726, - "eval_steps_per_second": 0.183, + "eval_accuracy": 0.4109195402298851, + "eval_loss": 2.3817079067230225, + "eval_runtime": 122.5427, + "eval_samples_per_second": 11.359, + "eval_steps_per_second": 0.18, "step": 528 }, { "epoch": 33.001041666666666, - "grad_norm": 6.766725063323975, + "grad_norm": 6.116299629211426, "learning_rate": 4.021990740740741e-05, - "loss": 1.5203, + "loss": 1.6292, "step": 530 }, { "epoch": 33.00625, - "grad_norm": 5.995940208435059, + "grad_norm": 6.249851703643799, "learning_rate": 3.993055555555556e-05, - "loss": 1.4576, + "loss": 1.6489, "step": 540 }, { "epoch": 33.00833333333333, - "eval_accuracy": 0.45454545454545453, - "eval_loss": 2.227916717529297, - "eval_runtime": 124.0445, - "eval_samples_per_second": 11.351, - "eval_steps_per_second": 0.177, + "eval_accuracy": 0.4274425287356322, + "eval_loss": 2.3350534439086914, + "eval_runtime": 118.7762, + "eval_samples_per_second": 11.72, + "eval_steps_per_second": 0.185, "step": 544 }, { "epoch": 34.003125, - "grad_norm": 6.610184669494629, + "grad_norm": 5.983939170837402, "learning_rate": 3.964120370370371e-05, - "loss": 1.5347, + "loss": 1.6495, "step": 550 }, { "epoch": 34.00833333333333, - "grad_norm": 24.694988250732422, + "grad_norm": 17.406435012817383, "learning_rate": 3.935185185185186e-05, - "loss": 1.5455, + "loss": 1.6928, "step": 560 }, { "epoch": 34.00833333333333, - "eval_accuracy": 0.4524147727272727, - "eval_loss": 2.204953908920288, - "eval_runtime": 131.2835, - "eval_samples_per_second": 10.725, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.4188218390804598, + "eval_loss": 2.3494503498077393, + "eval_runtime": 112.8904, + "eval_samples_per_second": 12.331, + "eval_steps_per_second": 0.195, "step": 560 }, { "epoch": 35.005208333333336, - "grad_norm": 6.918396472930908, + "grad_norm": 6.939193248748779, "learning_rate": 3.90625e-05, - "loss": 1.4485, + "loss": 1.503, "step": 570 }, { "epoch": 35.00833333333333, - "eval_accuracy": 0.4737215909090909, - "eval_loss": 2.1584527492523193, - "eval_runtime": 134.6721, - "eval_samples_per_second": 10.455, - "eval_steps_per_second": 0.163, + "eval_accuracy": 0.4274425287356322, + "eval_loss": 2.2961111068725586, + "eval_runtime": 122.4307, + "eval_samples_per_second": 11.37, + "eval_steps_per_second": 0.18, "step": 576 }, { "epoch": 36.00208333333333, - "grad_norm": 6.077505588531494, + "grad_norm": 8.047181129455566, "learning_rate": 3.877314814814815e-05, - "loss": 1.4391, + "loss": 1.4577, "step": 580 }, { "epoch": 36.00729166666667, - "grad_norm": 6.617316722869873, + "grad_norm": 5.338744640350342, "learning_rate": 3.84837962962963e-05, - "loss": 1.3896, + "loss": 1.5126, "step": 590 }, { "epoch": 36.00833333333333, - "eval_accuracy": 0.4446022727272727, - "eval_loss": 2.1850500106811523, - "eval_runtime": 129.9914, - "eval_samples_per_second": 10.831, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.4454022988505747, + "eval_loss": 2.2619619369506836, + "eval_runtime": 143.5341, + "eval_samples_per_second": 9.698, + "eval_steps_per_second": 0.153, "step": 592 }, { "epoch": 37.00416666666667, - "grad_norm": 7.900727272033691, + "grad_norm": 6.184549808502197, "learning_rate": 3.8194444444444444e-05, - "loss": 1.3766, + "loss": 1.4732, "step": 600 }, { "epoch": 37.00833333333333, - "eval_accuracy": 0.4872159090909091, - "eval_loss": 2.118501663208008, - "eval_runtime": 131.0307, - "eval_samples_per_second": 10.746, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.4367816091954023, + "eval_loss": 2.2635250091552734, + "eval_runtime": 145.8205, + "eval_samples_per_second": 9.546, + "eval_steps_per_second": 0.151, "step": 608 }, { "epoch": 38.001041666666666, - "grad_norm": 6.658583641052246, + "grad_norm": 6.274258136749268, "learning_rate": 3.7905092592592594e-05, - "loss": 1.287, + "loss": 1.52, "step": 610 }, { "epoch": 38.00625, - "grad_norm": 5.483121395111084, + "grad_norm": 6.8546576499938965, "learning_rate": 3.7615740740740744e-05, - "loss": 1.4035, + "loss": 1.5553, "step": 620 }, { "epoch": 38.00833333333333, - "eval_accuracy": 0.4794034090909091, - "eval_loss": 2.116427183151245, - "eval_runtime": 133.3027, - "eval_samples_per_second": 10.562, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.4489942528735632, + "eval_loss": 2.2326316833496094, + "eval_runtime": 121.0135, + "eval_samples_per_second": 11.503, + "eval_steps_per_second": 0.182, "step": 624 }, { "epoch": 39.003125, - "grad_norm": 6.027029991149902, + "grad_norm": 6.396101951599121, "learning_rate": 3.7326388888888893e-05, - "loss": 1.416, + "loss": 1.4549, "step": 630 }, { "epoch": 39.00833333333333, - "grad_norm": 17.550609588623047, + "grad_norm": 22.021570205688477, "learning_rate": 3.7037037037037037e-05, - "loss": 1.5892, + "loss": 1.7115, "step": 640 }, { "epoch": 39.00833333333333, - "eval_accuracy": 0.48011363636363635, - "eval_loss": 2.102943181991577, - "eval_runtime": 124.9558, - "eval_samples_per_second": 11.268, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.44037356321839083, + "eval_loss": 2.2265872955322266, + "eval_runtime": 132.2101, + "eval_samples_per_second": 10.529, + "eval_steps_per_second": 0.166, "step": 640 }, { "epoch": 40.005208333333336, - "grad_norm": 8.286770820617676, + "grad_norm": 5.9612627029418945, "learning_rate": 3.6747685185185186e-05, - "loss": 1.3647, + "loss": 1.4851, "step": 650 }, { "epoch": 40.00833333333333, - "eval_accuracy": 0.4928977272727273, - "eval_loss": 2.0912482738494873, - "eval_runtime": 125.7809, - "eval_samples_per_second": 11.194, - "eval_steps_per_second": 0.175, + "eval_accuracy": 0.4274425287356322, + "eval_loss": 2.2690393924713135, + "eval_runtime": 133.5177, + "eval_samples_per_second": 10.426, + "eval_steps_per_second": 0.165, "step": 656 }, { "epoch": 41.00208333333333, - "grad_norm": 6.9772138595581055, + "grad_norm": 5.4721550941467285, "learning_rate": 3.6458333333333336e-05, - "loss": 1.3769, + "loss": 1.4883, "step": 660 }, { "epoch": 41.00729166666667, - "grad_norm": 6.215446472167969, + "grad_norm": 8.272603034973145, "learning_rate": 3.6168981481481486e-05, - "loss": 1.388, + "loss": 1.455, "step": 670 }, { "epoch": 41.00833333333333, - "eval_accuracy": 0.47301136363636365, - "eval_loss": 2.1330864429473877, - "eval_runtime": 134.2267, - "eval_samples_per_second": 10.49, - "eval_steps_per_second": 0.164, + "eval_accuracy": 0.45689655172413796, + "eval_loss": 2.192133665084839, + "eval_runtime": 134.8188, + "eval_samples_per_second": 10.325, + "eval_steps_per_second": 0.163, "step": 672 }, { "epoch": 42.00416666666667, - "grad_norm": 6.233066082000732, + "grad_norm": 6.0662407875061035, "learning_rate": 3.587962962962963e-05, - "loss": 1.3425, + "loss": 1.4827, "step": 680 }, { "epoch": 42.00833333333333, - "eval_accuracy": 0.4794034090909091, - "eval_loss": 2.143657922744751, - "eval_runtime": 126.4352, - "eval_samples_per_second": 11.136, - "eval_steps_per_second": 0.174, + "eval_accuracy": 0.4504310344827586, + "eval_loss": 2.238664388656616, + "eval_runtime": 129.8736, + "eval_samples_per_second": 10.718, + "eval_steps_per_second": 0.169, "step": 688 }, { "epoch": 43.001041666666666, - "grad_norm": 7.223750591278076, + "grad_norm": 8.369169235229492, "learning_rate": 3.559027777777778e-05, - "loss": 1.1958, + "loss": 1.2639, "step": 690 }, { "epoch": 43.00625, - "grad_norm": 6.594886779785156, + "grad_norm": 9.68371868133545, "learning_rate": 3.530092592592593e-05, - "loss": 1.2909, + "loss": 1.4839, "step": 700 }, { "epoch": 43.00833333333333, - "eval_accuracy": 0.4715909090909091, - "eval_loss": 2.109005928039551, - "eval_runtime": 129.7594, - "eval_samples_per_second": 10.851, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.4432471264367816, + "eval_loss": 2.2020034790039062, + "eval_runtime": 146.2451, + "eval_samples_per_second": 9.518, + "eval_steps_per_second": 0.15, "step": 704 }, { "epoch": 44.003125, - "grad_norm": 5.431155681610107, + "grad_norm": 6.168234825134277, "learning_rate": 3.501157407407408e-05, - "loss": 1.2277, + "loss": 1.1762, "step": 710 }, { "epoch": 44.00833333333333, - "grad_norm": 23.15688133239746, + "grad_norm": 29.257152557373047, "learning_rate": 3.472222222222222e-05, - "loss": 1.2757, + "loss": 1.2879, "step": 720 }, { "epoch": 44.00833333333333, - "eval_accuracy": 0.4900568181818182, - "eval_loss": 2.0685956478118896, - "eval_runtime": 117.5767, - "eval_samples_per_second": 11.975, - "eval_steps_per_second": 0.187, + "eval_accuracy": 0.4396551724137931, + "eval_loss": 2.195861339569092, + "eval_runtime": 118.2366, + "eval_samples_per_second": 11.773, + "eval_steps_per_second": 0.186, "step": 720 }, { "epoch": 45.005208333333336, - "grad_norm": 6.826360702514648, + "grad_norm": 7.243413925170898, "learning_rate": 3.443287037037037e-05, - "loss": 1.181, + "loss": 1.2722, "step": 730 }, { "epoch": 45.00833333333333, - "eval_accuracy": 0.4893465909090909, - "eval_loss": 2.0484628677368164, - "eval_runtime": 122.183, - "eval_samples_per_second": 11.524, - "eval_steps_per_second": 0.18, + "eval_accuracy": 0.44396551724137934, + "eval_loss": 2.2157773971557617, + "eval_runtime": 131.0025, + "eval_samples_per_second": 10.626, + "eval_steps_per_second": 0.168, "step": 736 }, { "epoch": 46.00208333333333, - "grad_norm": 6.446974277496338, + "grad_norm": 6.851569652557373, "learning_rate": 3.414351851851852e-05, - "loss": 1.2034, + "loss": 1.4204, "step": 740 }, { "epoch": 46.00729166666667, - "grad_norm": 6.026095390319824, + "grad_norm": 5.8133463859558105, "learning_rate": 3.385416666666667e-05, - "loss": 1.1825, + "loss": 1.2225, "step": 750 }, { "epoch": 46.00833333333333, - "eval_accuracy": 0.484375, - "eval_loss": 2.0560503005981445, - "eval_runtime": 129.8384, - "eval_samples_per_second": 10.844, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.46623563218390807, + "eval_loss": 2.1567935943603516, + "eval_runtime": 151.0401, + "eval_samples_per_second": 9.216, + "eval_steps_per_second": 0.146, "step": 752 }, { "epoch": 47.00416666666667, - "grad_norm": 7.490734577178955, + "grad_norm": 5.665341377258301, "learning_rate": 3.3564814814814815e-05, - "loss": 1.1594, + "loss": 1.1821, "step": 760 }, { "epoch": 47.00833333333333, - "eval_accuracy": 0.49644886363636365, - "eval_loss": 2.0327041149139404, - "eval_runtime": 125.1147, - "eval_samples_per_second": 11.254, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.47629310344827586, + "eval_loss": 2.1312315464019775, + "eval_runtime": 113.8029, + "eval_samples_per_second": 12.232, + "eval_steps_per_second": 0.193, "step": 768 }, { "epoch": 48.001041666666666, - "grad_norm": 6.074591636657715, + "grad_norm": 6.758544445037842, "learning_rate": 3.3275462962962965e-05, - "loss": 1.0989, + "loss": 1.2554, "step": 770 }, { "epoch": 48.00625, - "grad_norm": 7.924504280090332, + "grad_norm": 7.730432510375977, "learning_rate": 3.2986111111111115e-05, - "loss": 1.1699, + "loss": 1.2406, "step": 780 }, { "epoch": 48.00833333333333, - "eval_accuracy": 0.4765625, - "eval_loss": 2.095003366470337, - "eval_runtime": 131.8398, - "eval_samples_per_second": 10.68, - "eval_steps_per_second": 0.167, + "eval_accuracy": 0.47844827586206895, + "eval_loss": 2.1162068843841553, + "eval_runtime": 138.6268, + "eval_samples_per_second": 10.041, + "eval_steps_per_second": 0.159, "step": 784 }, { "epoch": 49.003125, - "grad_norm": 7.409806728363037, + "grad_norm": 5.654488563537598, "learning_rate": 3.2696759259259265e-05, - "loss": 1.2027, + "loss": 1.2432, "step": 790 }, { "epoch": 49.00833333333333, - "grad_norm": 23.586345672607422, + "grad_norm": 21.777118682861328, "learning_rate": 3.240740740740741e-05, - "loss": 1.1908, + "loss": 1.1717, "step": 800 }, { "epoch": 49.00833333333333, - "eval_accuracy": 0.4850852272727273, - "eval_loss": 2.0465078353881836, - "eval_runtime": 129.3358, - "eval_samples_per_second": 10.886, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.4755747126436782, + "eval_loss": 2.1368021965026855, + "eval_runtime": 124.2033, + "eval_samples_per_second": 11.207, + "eval_steps_per_second": 0.177, "step": 800 }, { "epoch": 50.005208333333336, - "grad_norm": 7.079756736755371, + "grad_norm": 6.306783199310303, "learning_rate": 3.211805555555556e-05, - "loss": 1.1149, + "loss": 1.2366, "step": 810 }, { "epoch": 50.00833333333333, - "eval_accuracy": 0.48792613636363635, - "eval_loss": 2.0569851398468018, - "eval_runtime": 123.867, - "eval_samples_per_second": 11.367, - "eval_steps_per_second": 0.178, + "eval_accuracy": 0.4834770114942529, + "eval_loss": 2.1134183406829834, + "eval_runtime": 139.8423, + "eval_samples_per_second": 9.954, + "eval_steps_per_second": 0.157, "step": 816 }, { "epoch": 51.00208333333333, - "grad_norm": 5.848146438598633, + "grad_norm": 6.758057117462158, "learning_rate": 3.182870370370371e-05, - "loss": 1.0907, + "loss": 1.1925, "step": 820 }, { "epoch": 51.00729166666667, - "grad_norm": 7.725104808807373, + "grad_norm": 6.371450424194336, "learning_rate": 3.153935185185186e-05, - "loss": 1.1388, + "loss": 1.2534, "step": 830 }, { "epoch": 51.00833333333333, - "eval_accuracy": 0.4978693181818182, - "eval_loss": 2.0232162475585938, - "eval_runtime": 124.9726, - "eval_samples_per_second": 11.266, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.4734195402298851, + "eval_loss": 2.096381664276123, + "eval_runtime": 153.8561, + "eval_samples_per_second": 9.047, + "eval_steps_per_second": 0.143, "step": 832 }, { "epoch": 52.00416666666667, - "grad_norm": 6.411397457122803, + "grad_norm": 6.33159065246582, "learning_rate": 3.125e-05, - "loss": 1.0421, + "loss": 1.2322, "step": 840 }, { "epoch": 52.00833333333333, - "eval_accuracy": 0.49857954545454547, - "eval_loss": 2.0132648944854736, - "eval_runtime": 130.0486, - "eval_samples_per_second": 10.827, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.4640804597701149, + "eval_loss": 2.150590658187866, + "eval_runtime": 115.0189, + "eval_samples_per_second": 12.102, + "eval_steps_per_second": 0.191, "step": 848 }, { "epoch": 53.001041666666666, - "grad_norm": 6.417796611785889, + "grad_norm": 5.289525985717773, "learning_rate": 3.0960648148148144e-05, - "loss": 1.128, + "loss": 1.2281, "step": 850 }, { "epoch": 53.00625, - "grad_norm": 6.483155727386475, + "grad_norm": 6.158384799957275, "learning_rate": 3.06712962962963e-05, - "loss": 1.1243, + "loss": 1.2742, "step": 860 }, { "epoch": 53.00833333333333, - "eval_accuracy": 0.4900568181818182, - "eval_loss": 2.0420944690704346, - "eval_runtime": 131.0098, - "eval_samples_per_second": 10.747, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.45905172413793105, + "eval_loss": 2.171858310699463, + "eval_runtime": 139.4117, + "eval_samples_per_second": 9.985, + "eval_steps_per_second": 0.158, "step": 864 }, { "epoch": 54.003125, - "grad_norm": 7.885180950164795, + "grad_norm": 4.942337989807129, "learning_rate": 3.0381944444444444e-05, - "loss": 1.0331, + "loss": 1.1286, "step": 870 }, { "epoch": 54.00833333333333, - "grad_norm": 28.23676300048828, + "grad_norm": 20.070188522338867, "learning_rate": 3.0092592592592593e-05, - "loss": 1.1064, + "loss": 1.132, "step": 880 }, { "epoch": 54.00833333333333, - "eval_accuracy": 0.5042613636363636, - "eval_loss": 1.9613640308380127, - "eval_runtime": 129.6513, - "eval_samples_per_second": 10.86, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.46479885057471265, + "eval_loss": 2.1557226181030273, + "eval_runtime": 133.4523, + "eval_samples_per_second": 10.431, + "eval_steps_per_second": 0.165, "step": 880 }, { "epoch": 55.005208333333336, - "grad_norm": 7.0548319816589355, + "grad_norm": 7.356424808502197, "learning_rate": 2.980324074074074e-05, - "loss": 0.9778, + "loss": 1.1306, "step": 890 }, { "epoch": 55.00833333333333, - "eval_accuracy": 0.5071022727272727, - "eval_loss": 1.9938956499099731, - "eval_runtime": 125.6294, - "eval_samples_per_second": 11.208, - "eval_steps_per_second": 0.175, + "eval_accuracy": 0.48778735632183906, + "eval_loss": 2.101032018661499, + "eval_runtime": 120.4781, + "eval_samples_per_second": 11.554, + "eval_steps_per_second": 0.183, "step": 896 }, { "epoch": 56.00208333333333, - "grad_norm": 7.993034362792969, + "grad_norm": 6.045055389404297, "learning_rate": 2.951388888888889e-05, - "loss": 1.0192, + "loss": 1.1428, "step": 900 }, { "epoch": 56.00729166666667, - "grad_norm": 8.833446502685547, + "grad_norm": 8.3392915725708, "learning_rate": 2.9224537037037036e-05, - "loss": 1.1417, + "loss": 1.2719, "step": 910 }, { "epoch": 56.00833333333333, - "eval_accuracy": 0.5106534090909091, - "eval_loss": 1.977372646331787, - "eval_runtime": 141.0715, - "eval_samples_per_second": 9.981, - "eval_steps_per_second": 0.156, + "eval_accuracy": 0.4619252873563218, + "eval_loss": 2.148237466812134, + "eval_runtime": 113.6281, + "eval_samples_per_second": 12.25, + "eval_steps_per_second": 0.194, "step": 912 }, { "epoch": 57.00416666666667, - "grad_norm": 6.480681896209717, + "grad_norm": 5.768066883087158, "learning_rate": 2.8935185185185186e-05, - "loss": 1.0578, + "loss": 1.1549, "step": 920 }, { "epoch": 57.00833333333333, - "eval_accuracy": 0.5305397727272727, - "eval_loss": 1.9625455141067505, - "eval_runtime": 137.3354, - "eval_samples_per_second": 10.252, - "eval_steps_per_second": 0.16, + "eval_accuracy": 0.4813218390804598, + "eval_loss": 2.0960781574249268, + "eval_runtime": 151.6984, + "eval_samples_per_second": 9.176, + "eval_steps_per_second": 0.145, "step": 928 }, { "epoch": 58.001041666666666, - "grad_norm": 6.03174352645874, + "grad_norm": 7.034135818481445, "learning_rate": 2.8645833333333333e-05, - "loss": 0.9695, + "loss": 1.0771, "step": 930 }, { "epoch": 58.00625, - "grad_norm": 5.930421352386475, + "grad_norm": 7.099185466766357, "learning_rate": 2.8356481481481483e-05, - "loss": 1.0904, + "loss": 1.1495, "step": 940 }, { "epoch": 58.00833333333333, - "eval_accuracy": 0.5056818181818182, - "eval_loss": 1.971279501914978, - "eval_runtime": 124.5828, - "eval_samples_per_second": 11.302, - "eval_steps_per_second": 0.177, + "eval_accuracy": 0.47629310344827586, + "eval_loss": 2.124603509902954, + "eval_runtime": 127.3723, + "eval_samples_per_second": 10.929, + "eval_steps_per_second": 0.173, "step": 944 }, { "epoch": 59.003125, - "grad_norm": 7.177443027496338, + "grad_norm": 6.804399013519287, "learning_rate": 2.806712962962963e-05, - "loss": 1.0493, + "loss": 1.1043, "step": 950 }, { "epoch": 59.00833333333333, - "grad_norm": 32.896873474121094, + "grad_norm": 31.837963104248047, "learning_rate": 2.777777777777778e-05, - "loss": 1.2569, + "loss": 1.2539, "step": 960 }, { "epoch": 59.00833333333333, - "eval_accuracy": 0.5255681818181818, - "eval_loss": 1.9495558738708496, - "eval_runtime": 125.0434, - "eval_samples_per_second": 11.26, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.48060344827586204, + "eval_loss": 2.1117918491363525, + "eval_runtime": 142.6628, + "eval_samples_per_second": 9.757, + "eval_steps_per_second": 0.154, "step": 960 }, { "epoch": 60.005208333333336, - "grad_norm": 6.914553642272949, + "grad_norm": 7.992947578430176, "learning_rate": 2.7488425925925926e-05, - "loss": 1.076, + "loss": 1.1719, "step": 970 }, { "epoch": 60.00833333333333, - "eval_accuracy": 0.5369318181818182, - "eval_loss": 1.9237945079803467, - "eval_runtime": 130.5511, - "eval_samples_per_second": 10.785, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.49353448275862066, + "eval_loss": 2.0665571689605713, + "eval_runtime": 129.4806, + "eval_samples_per_second": 10.751, + "eval_steps_per_second": 0.17, "step": 976 }, { "epoch": 61.00208333333333, - "grad_norm": 6.131436347961426, + "grad_norm": 5.261826038360596, "learning_rate": 2.7199074074074076e-05, - "loss": 0.9972, + "loss": 1.055, "step": 980 }, { "epoch": 61.00729166666667, - "grad_norm": 8.375364303588867, + "grad_norm": 8.344293594360352, "learning_rate": 2.6909722222222222e-05, - "loss": 1.018, + "loss": 1.1108, "step": 990 }, { "epoch": 61.00833333333333, - "eval_accuracy": 0.515625, - "eval_loss": 1.9578070640563965, - "eval_runtime": 128.8964, - "eval_samples_per_second": 10.924, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.4834770114942529, + "eval_loss": 2.0630218982696533, + "eval_runtime": 111.5142, + "eval_samples_per_second": 12.483, + "eval_steps_per_second": 0.197, "step": 992 }, { "epoch": 62.00416666666667, - "grad_norm": 8.409364700317383, + "grad_norm": 7.500652313232422, "learning_rate": 2.6620370370370372e-05, - "loss": 0.8569, + "loss": 1.0417, "step": 1000 }, { "epoch": 62.00833333333333, - "eval_accuracy": 0.5184659090909091, - "eval_loss": 1.9410430192947388, - "eval_runtime": 126.0956, - "eval_samples_per_second": 11.166, + "eval_accuracy": 0.4899425287356322, + "eval_loss": 2.0634570121765137, + "eval_runtime": 126.2555, + "eval_samples_per_second": 11.025, "eval_steps_per_second": 0.174, "step": 1008 }, { "epoch": 63.001041666666666, - "grad_norm": 6.331228256225586, + "grad_norm": 6.283971786499023, "learning_rate": 2.633101851851852e-05, - "loss": 1.1689, + "loss": 1.2428, "step": 1010 }, { "epoch": 63.00625, - "grad_norm": 5.7633161544799805, + "grad_norm": 7.511735439300537, "learning_rate": 2.604166666666667e-05, - "loss": 0.9847, + "loss": 1.1755, "step": 1020 }, { "epoch": 63.00833333333333, - "eval_accuracy": 0.5134943181818182, - "eval_loss": 1.965501308441162, - "eval_runtime": 134.8984, - "eval_samples_per_second": 10.437, - "eval_steps_per_second": 0.163, + "eval_accuracy": 0.47198275862068967, + "eval_loss": 2.0862152576446533, + "eval_runtime": 124.1187, + "eval_samples_per_second": 11.215, + "eval_steps_per_second": 0.177, "step": 1024 }, { "epoch": 64.003125, - "grad_norm": 6.778160572052002, + "grad_norm": 6.025021076202393, "learning_rate": 2.5752314814814815e-05, - "loss": 1.0541, + "loss": 1.0898, "step": 1030 }, { "epoch": 64.00833333333334, - "grad_norm": 15.068023681640625, + "grad_norm": 23.160959243774414, "learning_rate": 2.5462962962962965e-05, - "loss": 0.8992, + "loss": 1.0512, "step": 1040 }, { "epoch": 64.00833333333334, - "eval_accuracy": 0.5184659090909091, - "eval_loss": 1.97407865524292, - "eval_runtime": 131.1607, - "eval_samples_per_second": 10.735, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.48778735632183906, + "eval_loss": 2.073040723800659, + "eval_runtime": 128.3013, + "eval_samples_per_second": 10.849, + "eval_steps_per_second": 0.171, "step": 1040 }, { "epoch": 65.00520833333333, - "grad_norm": 6.6322126388549805, + "grad_norm": 5.617940902709961, "learning_rate": 2.517361111111111e-05, - "loss": 0.9781, + "loss": 0.9824, "step": 1050 }, { "epoch": 65.00833333333334, - "eval_accuracy": 0.5248579545454546, - "eval_loss": 1.9591399431228638, - "eval_runtime": 130.8261, - "eval_samples_per_second": 10.762, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.4906609195402299, + "eval_loss": 2.070878744125366, + "eval_runtime": 122.5737, + "eval_samples_per_second": 11.356, + "eval_steps_per_second": 0.179, "step": 1056 }, { "epoch": 66.00208333333333, - "grad_norm": 6.270596504211426, + "grad_norm": 5.892145156860352, "learning_rate": 2.488425925925926e-05, - "loss": 1.1274, + "loss": 1.1867, "step": 1060 }, { "epoch": 66.00729166666666, - "grad_norm": 5.216790676116943, + "grad_norm": 7.249321937561035, "learning_rate": 2.4594907407407408e-05, - "loss": 0.9016, + "loss": 1.0924, "step": 1070 }, { "epoch": 66.00833333333334, - "eval_accuracy": 0.5134943181818182, - "eval_loss": 1.9802237749099731, - "eval_runtime": 138.409, - "eval_samples_per_second": 10.173, - "eval_steps_per_second": 0.159, + "eval_accuracy": 0.46120689655172414, + "eval_loss": 2.16375732421875, + "eval_runtime": 110.676, + "eval_samples_per_second": 12.577, + "eval_steps_per_second": 0.199, "step": 1072 }, { "epoch": 67.00416666666666, - "grad_norm": 4.889580249786377, + "grad_norm": 6.243124485015869, "learning_rate": 2.4305555555555558e-05, - "loss": 0.9443, + "loss": 1.1027, "step": 1080 }, { "epoch": 67.00833333333334, - "eval_accuracy": 0.5035511363636364, - "eval_loss": 1.9881755113601685, - "eval_runtime": 132.577, - "eval_samples_per_second": 10.62, - "eval_steps_per_second": 0.166, + "eval_accuracy": 0.47772988505747127, + "eval_loss": 2.0572071075439453, + "eval_runtime": 123.131, + "eval_samples_per_second": 11.305, + "eval_steps_per_second": 0.179, "step": 1088 }, { "epoch": 68.00104166666667, - "grad_norm": 8.338197708129883, + "grad_norm": 7.192434787750244, "learning_rate": 2.4016203703703704e-05, - "loss": 0.9719, + "loss": 0.942, "step": 1090 }, { "epoch": 68.00625, - "grad_norm": 7.402223110198975, + "grad_norm": 5.061285018920898, "learning_rate": 2.3726851851851854e-05, - "loss": 0.9359, + "loss": 1.0956, "step": 1100 }, { "epoch": 68.00833333333334, - "eval_accuracy": 0.5092329545454546, - "eval_loss": 2.0045785903930664, - "eval_runtime": 122.9735, - "eval_samples_per_second": 11.45, - "eval_steps_per_second": 0.179, + "eval_accuracy": 0.4892241379310345, + "eval_loss": 2.050154685974121, + "eval_runtime": 159.2133, + "eval_samples_per_second": 8.743, + "eval_steps_per_second": 0.138, "step": 1104 }, { "epoch": 69.003125, - "grad_norm": 5.299436092376709, + "grad_norm": 6.748392581939697, "learning_rate": 2.34375e-05, - "loss": 0.9554, + "loss": 1.0495, "step": 1110 }, { "epoch": 69.00833333333334, - "grad_norm": 22.75337028503418, + "grad_norm": 21.487567901611328, "learning_rate": 2.314814814814815e-05, - "loss": 0.7735, + "loss": 0.8823, "step": 1120 }, { "epoch": 69.00833333333334, - "eval_accuracy": 0.5063920454545454, - "eval_loss": 2.017183542251587, - "eval_runtime": 132.6432, - "eval_samples_per_second": 10.615, - "eval_steps_per_second": 0.166, + "eval_accuracy": 0.4755747126436782, + "eval_loss": 2.112837076187134, + "eval_runtime": 115.1946, + "eval_samples_per_second": 12.084, + "eval_steps_per_second": 0.191, "step": 1120 }, { "epoch": 70.00520833333333, - "grad_norm": 6.869340896606445, + "grad_norm": 6.899831295013428, "learning_rate": 2.2858796296296297e-05, - "loss": 0.9405, + "loss": 1.0344, "step": 1130 }, { "epoch": 70.00833333333334, - "eval_accuracy": 0.5269886363636364, - "eval_loss": 1.955207109451294, - "eval_runtime": 135.5316, - "eval_samples_per_second": 10.389, - "eval_steps_per_second": 0.162, + "eval_accuracy": 0.47270114942528735, + "eval_loss": 2.095017433166504, + "eval_runtime": 147.4741, + "eval_samples_per_second": 9.439, + "eval_steps_per_second": 0.149, "step": 1136 }, { "epoch": 71.00208333333333, - "grad_norm": 7.737886428833008, + "grad_norm": 6.141778469085693, "learning_rate": 2.2569444444444447e-05, - "loss": 0.9419, + "loss": 1.0074, "step": 1140 }, { "epoch": 71.00729166666666, - "grad_norm": 6.463024139404297, + "grad_norm": 5.4406352043151855, "learning_rate": 2.2280092592592593e-05, - "loss": 0.9709, + "loss": 1.0887, "step": 1150 }, { "epoch": 71.00833333333334, - "eval_accuracy": 0.5227272727272727, - "eval_loss": 1.9572663307189941, - "eval_runtime": 134.3176, - "eval_samples_per_second": 10.483, - "eval_steps_per_second": 0.164, + "eval_accuracy": 0.4942528735632184, + "eval_loss": 2.0542972087860107, + "eval_runtime": 125.512, + "eval_samples_per_second": 11.091, + "eval_steps_per_second": 0.175, "step": 1152 }, { "epoch": 72.00416666666666, - "grad_norm": 6.5754923820495605, + "grad_norm": 6.820794582366943, "learning_rate": 2.1990740740740743e-05, - "loss": 0.9914, + "loss": 1.0763, "step": 1160 }, { "epoch": 72.00833333333334, - "eval_accuracy": 0.5248579545454546, - "eval_loss": 1.9767735004425049, - "eval_runtime": 125.4669, - "eval_samples_per_second": 11.222, - "eval_steps_per_second": 0.175, + "eval_accuracy": 0.4906609195402299, + "eval_loss": 2.053474187850952, + "eval_runtime": 138.7554, + "eval_samples_per_second": 10.032, + "eval_steps_per_second": 0.159, "step": 1168 }, { "epoch": 73.00104166666667, - "grad_norm": 6.831247329711914, + "grad_norm": 6.077042102813721, "learning_rate": 2.170138888888889e-05, - "loss": 0.8804, + "loss": 0.9623, "step": 1170 }, { "epoch": 73.00625, - "grad_norm": 7.8077392578125, + "grad_norm": 8.70025634765625, "learning_rate": 2.141203703703704e-05, - "loss": 0.8487, + "loss": 0.9652, "step": 1180 }, { "epoch": 73.00833333333334, - "eval_accuracy": 0.5326704545454546, - "eval_loss": 1.9569602012634277, - "eval_runtime": 134.6123, - "eval_samples_per_second": 10.46, - "eval_steps_per_second": 0.163, + "eval_accuracy": 0.5, + "eval_loss": 2.028047561645508, + "eval_runtime": 126.5448, + "eval_samples_per_second": 11.0, + "eval_steps_per_second": 0.174, "step": 1184 }, { "epoch": 74.003125, - "grad_norm": 5.786107063293457, + "grad_norm": 7.789737701416016, "learning_rate": 2.1122685185185186e-05, - "loss": 0.9529, + "loss": 0.9284, "step": 1190 }, { "epoch": 74.00833333333334, - "grad_norm": 21.245386123657227, + "grad_norm": 15.859330177307129, "learning_rate": 2.0833333333333336e-05, - "loss": 0.835, + "loss": 1.0445, "step": 1200 }, { "epoch": 74.00833333333334, - "eval_accuracy": 0.5241477272727273, - "eval_loss": 1.9758590459823608, - "eval_runtime": 129.6528, - "eval_samples_per_second": 10.86, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.48204022988505746, + "eval_loss": 2.055063486099243, + "eval_runtime": 143.3562, + "eval_samples_per_second": 9.71, + "eval_steps_per_second": 0.153, "step": 1200 }, { "epoch": 75.00520833333333, - "grad_norm": 5.697115421295166, + "grad_norm": 6.045502662658691, "learning_rate": 2.0543981481481483e-05, - "loss": 0.8914, + "loss": 0.9844, "step": 1210 }, { "epoch": 75.00833333333334, - "eval_accuracy": 0.5298295454545454, - "eval_loss": 1.9309029579162598, - "eval_runtime": 130.377, - "eval_samples_per_second": 10.799, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.5043103448275862, + "eval_loss": 2.051361083984375, + "eval_runtime": 128.397, + "eval_samples_per_second": 10.841, + "eval_steps_per_second": 0.171, "step": 1216 }, { "epoch": 76.00208333333333, - "grad_norm": 4.525904655456543, + "grad_norm": 4.970855712890625, "learning_rate": 2.0254629629629632e-05, - "loss": 0.9268, + "loss": 1.1088, "step": 1220 }, { "epoch": 76.00729166666666, - "grad_norm": 6.657559871673584, + "grad_norm": 7.561370849609375, "learning_rate": 1.996527777777778e-05, - "loss": 0.9242, + "loss": 1.0809, "step": 1230 }, { "epoch": 76.00833333333334, - "eval_accuracy": 0.5241477272727273, - "eval_loss": 1.9594990015029907, - "eval_runtime": 130.7531, - "eval_samples_per_second": 10.768, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.5021551724137931, + "eval_loss": 2.05523419380188, + "eval_runtime": 129.1447, + "eval_samples_per_second": 10.779, + "eval_steps_per_second": 0.17, "step": 1232 }, { "epoch": 77.00416666666666, - "grad_norm": 6.962241172790527, + "grad_norm": 6.377460479736328, "learning_rate": 1.967592592592593e-05, - "loss": 0.8235, + "loss": 1.1158, "step": 1240 }, { "epoch": 77.00833333333334, - "eval_accuracy": 0.5276988636363636, - "eval_loss": 1.9556376934051514, - "eval_runtime": 124.6215, - "eval_samples_per_second": 11.298, - "eval_steps_per_second": 0.177, + "eval_accuracy": 0.48778735632183906, + "eval_loss": 2.04689621925354, + "eval_runtime": 145.6931, + "eval_samples_per_second": 9.554, + "eval_steps_per_second": 0.151, "step": 1248 }, { "epoch": 78.00104166666667, - "grad_norm": 6.555630207061768, + "grad_norm": 7.5645904541015625, "learning_rate": 1.9386574074074075e-05, - "loss": 1.1044, + "loss": 0.9078, "step": 1250 }, { "epoch": 78.00625, - "grad_norm": 7.0458455085754395, + "grad_norm": 6.4087419509887695, "learning_rate": 1.9097222222222222e-05, - "loss": 0.8664, + "loss": 0.9017, "step": 1260 }, { "epoch": 78.00833333333334, - "eval_accuracy": 0.5134943181818182, - "eval_loss": 1.978991985321045, - "eval_runtime": 130.4341, - "eval_samples_per_second": 10.795, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.4906609195402299, + "eval_loss": 2.0516154766082764, + "eval_runtime": 121.723, + "eval_samples_per_second": 11.436, + "eval_steps_per_second": 0.181, "step": 1264 }, { "epoch": 79.003125, - "grad_norm": 4.815821170806885, + "grad_norm": 5.661813735961914, "learning_rate": 1.8807870370370372e-05, - "loss": 0.7884, + "loss": 0.8761, "step": 1270 }, { "epoch": 79.00833333333334, - "grad_norm": 26.361722946166992, + "grad_norm": 33.69865798950195, "learning_rate": 1.8518518518518518e-05, - "loss": 0.8699, + "loss": 1.0449, "step": 1280 }, { "epoch": 79.00833333333334, - "eval_accuracy": 0.5227272727272727, - "eval_loss": 1.9835097789764404, - "eval_runtime": 119.0563, - "eval_samples_per_second": 11.826, - "eval_steps_per_second": 0.185, + "eval_accuracy": 0.4863505747126437, + "eval_loss": 2.0770106315612793, + "eval_runtime": 129.9823, + "eval_samples_per_second": 10.709, + "eval_steps_per_second": 0.169, "step": 1280 }, { "epoch": 80.00520833333333, - "grad_norm": 4.6080474853515625, + "grad_norm": 5.260113716125488, "learning_rate": 1.8229166666666668e-05, - "loss": 0.9112, + "loss": 1.0167, "step": 1290 }, { "epoch": 80.00833333333334, - "eval_accuracy": 0.5291193181818182, - "eval_loss": 1.9426430463790894, - "eval_runtime": 137.1174, - "eval_samples_per_second": 10.269, - "eval_steps_per_second": 0.16, + "eval_accuracy": 0.4942528735632184, + "eval_loss": 2.0373690128326416, + "eval_runtime": 120.0265, + "eval_samples_per_second": 11.597, + "eval_steps_per_second": 0.183, "step": 1296 }, { "epoch": 81.00208333333333, - "grad_norm": 5.045470237731934, + "grad_norm": 6.28135347366333, "learning_rate": 1.7939814814814815e-05, - "loss": 0.8392, + "loss": 1.0278, "step": 1300 }, { "epoch": 81.00729166666666, - "grad_norm": 6.203034400939941, + "grad_norm": 7.153203010559082, "learning_rate": 1.7650462962962965e-05, - "loss": 0.7901, + "loss": 0.975, "step": 1310 }, { "epoch": 81.00833333333334, - "eval_accuracy": 0.5255681818181818, - "eval_loss": 1.959786295890808, - "eval_runtime": 120.2871, - "eval_samples_per_second": 11.705, - "eval_steps_per_second": 0.183, + "eval_accuracy": 0.4942528735632184, + "eval_loss": 2.0630548000335693, + "eval_runtime": 127.397, + "eval_samples_per_second": 10.926, + "eval_steps_per_second": 0.173, "step": 1312 }, { "epoch": 82.00416666666666, - "grad_norm": 6.214341163635254, + "grad_norm": 6.697967052459717, "learning_rate": 1.736111111111111e-05, - "loss": 0.8186, + "loss": 0.9285, "step": 1320 }, { "epoch": 82.00833333333334, - "eval_accuracy": 0.5319602272727273, - "eval_loss": 1.9397464990615845, - "eval_runtime": 130.0941, - "eval_samples_per_second": 10.823, - "eval_steps_per_second": 0.169, + "eval_accuracy": 0.4870689655172414, + "eval_loss": 2.049909830093384, + "eval_runtime": 143.6575, + "eval_samples_per_second": 9.69, + "eval_steps_per_second": 0.153, "step": 1328 }, { "epoch": 83.00104166666667, - "grad_norm": 5.813119411468506, + "grad_norm": 7.098538875579834, "learning_rate": 1.707175925925926e-05, - "loss": 0.7438, + "loss": 0.8193, "step": 1330 }, { "epoch": 83.00625, - "grad_norm": 5.390239715576172, + "grad_norm": 5.9787468910217285, "learning_rate": 1.6782407407407408e-05, - "loss": 0.8229, + "loss": 0.9762, "step": 1340 }, { "epoch": 83.00833333333334, - "eval_accuracy": 0.5326704545454546, - "eval_loss": 1.938385009765625, - "eval_runtime": 124.8818, - "eval_samples_per_second": 11.275, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.4964080459770115, + "eval_loss": 2.061843156814575, + "eval_runtime": 107.3354, + "eval_samples_per_second": 12.969, + "eval_steps_per_second": 0.205, "step": 1344 }, { "epoch": 84.003125, - "grad_norm": 5.821295738220215, + "grad_norm": 6.269992351531982, "learning_rate": 1.6493055555555557e-05, - "loss": 1.0251, + "loss": 1.0743, "step": 1350 }, { "epoch": 84.00833333333334, - "grad_norm": 23.789480209350586, + "grad_norm": 30.140321731567383, "learning_rate": 1.6203703703703704e-05, - "loss": 0.9063, + "loss": 0.9454, "step": 1360 }, { "epoch": 84.00833333333334, - "eval_accuracy": 0.5291193181818182, - "eval_loss": 1.9323524236679077, - "eval_runtime": 122.6737, - "eval_samples_per_second": 11.478, - "eval_steps_per_second": 0.179, + "eval_accuracy": 0.4992816091954023, + "eval_loss": 2.046207904815674, + "eval_runtime": 143.1149, + "eval_samples_per_second": 9.726, + "eval_steps_per_second": 0.154, "step": 1360 }, { "epoch": 85.00520833333333, - "grad_norm": 7.094452381134033, + "grad_norm": 6.801133632659912, "learning_rate": 1.5914351851851854e-05, - "loss": 0.8843, + "loss": 0.8665, "step": 1370 }, { "epoch": 85.00833333333334, - "eval_accuracy": 0.5369318181818182, - "eval_loss": 1.9315829277038574, - "eval_runtime": 124.9222, - "eval_samples_per_second": 11.271, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.4892241379310345, + "eval_loss": 2.076526165008545, + "eval_runtime": 132.1822, + "eval_samples_per_second": 10.531, + "eval_steps_per_second": 0.166, "step": 1376 }, { "epoch": 86.00208333333333, - "grad_norm": 5.776772975921631, + "grad_norm": 6.922851085662842, "learning_rate": 1.5625e-05, - "loss": 0.7684, + "loss": 0.8844, "step": 1380 }, { "epoch": 86.00729166666666, - "grad_norm": 8.940707206726074, + "grad_norm": 6.79546594619751, "learning_rate": 1.533564814814815e-05, - "loss": 0.7904, + "loss": 0.9202, "step": 1390 }, { "epoch": 86.00833333333334, - "eval_accuracy": 0.5376420454545454, - "eval_loss": 1.926943302154541, - "eval_runtime": 126.2416, - "eval_samples_per_second": 11.153, - "eval_steps_per_second": 0.174, + "eval_accuracy": 0.4949712643678161, + "eval_loss": 2.0512619018554688, + "eval_runtime": 117.1467, + "eval_samples_per_second": 11.883, + "eval_steps_per_second": 0.188, "step": 1392 }, { "epoch": 87.00416666666666, - "grad_norm": 6.900567054748535, + "grad_norm": 5.933279514312744, "learning_rate": 1.5046296296296297e-05, - "loss": 0.7942, + "loss": 0.8186, "step": 1400 }, { "epoch": 87.00833333333334, - "eval_accuracy": 0.5291193181818182, - "eval_loss": 1.9505839347839355, - "eval_runtime": 129.4495, - "eval_samples_per_second": 10.877, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.5093390804597702, + "eval_loss": 2.0254082679748535, + "eval_runtime": 139.5548, + "eval_samples_per_second": 9.975, + "eval_steps_per_second": 0.158, "step": 1408 }, { "epoch": 88.00104166666667, - "grad_norm": 6.561887741088867, + "grad_norm": 7.079472541809082, "learning_rate": 1.4756944444444445e-05, - "loss": 0.7708, + "loss": 0.8471, "step": 1410 }, { "epoch": 88.00625, - "grad_norm": 6.302940368652344, + "grad_norm": 5.367373466491699, "learning_rate": 1.4467592592592593e-05, - "loss": 0.8798, + "loss": 0.8659, "step": 1420 }, { "epoch": 88.00833333333334, - "eval_accuracy": 0.5404829545454546, - "eval_loss": 1.9184972047805786, - "eval_runtime": 136.5919, - "eval_samples_per_second": 10.308, - "eval_steps_per_second": 0.161, + "eval_accuracy": 0.4791666666666667, + "eval_loss": 2.1059958934783936, + "eval_runtime": 109.122, + "eval_samples_per_second": 12.756, + "eval_steps_per_second": 0.202, "step": 1424 }, { "epoch": 89.003125, - "grad_norm": 6.015424728393555, + "grad_norm": 4.5717926025390625, "learning_rate": 1.4178240740740741e-05, - "loss": 0.7788, + "loss": 0.8132, "step": 1430 }, { "epoch": 89.00833333333334, - "grad_norm": 9.586054801940918, + "grad_norm": 11.70888900756836, "learning_rate": 1.388888888888889e-05, - "loss": 0.7678, + "loss": 0.8789, "step": 1440 }, { "epoch": 89.00833333333334, - "eval_accuracy": 0.5326704545454546, - "eval_loss": 1.9361698627471924, - "eval_runtime": 128.5409, - "eval_samples_per_second": 10.954, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.4964080459770115, + "eval_loss": 2.029611110687256, + "eval_runtime": 117.991, + "eval_samples_per_second": 11.798, + "eval_steps_per_second": 0.186, "step": 1440 }, { "epoch": 90.00520833333333, - "grad_norm": 6.406522750854492, + "grad_norm": 6.696922779083252, "learning_rate": 1.3599537037037038e-05, - "loss": 0.7589, + "loss": 0.8592, "step": 1450 }, { "epoch": 90.00833333333334, - "eval_accuracy": 0.5276988636363636, - "eval_loss": 1.9496175050735474, - "eval_runtime": 128.3627, - "eval_samples_per_second": 10.969, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.4849137931034483, + "eval_loss": 2.0757148265838623, + "eval_runtime": 125.7116, + "eval_samples_per_second": 11.073, + "eval_steps_per_second": 0.175, "step": 1456 }, { "epoch": 91.00208333333333, - "grad_norm": 5.885864734649658, + "grad_norm": 6.2329535484313965, "learning_rate": 1.3310185185185186e-05, - "loss": 0.7746, + "loss": 0.8472, "step": 1460 }, { "epoch": 91.00729166666666, - "grad_norm": 6.763460159301758, + "grad_norm": 7.030002117156982, "learning_rate": 1.3020833333333334e-05, - "loss": 0.6679, + "loss": 0.8093, "step": 1470 }, { "epoch": 91.00833333333334, - "eval_accuracy": 0.5298295454545454, - "eval_loss": 1.9507235288619995, - "eval_runtime": 133.1729, - "eval_samples_per_second": 10.573, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.4985632183908046, + "eval_loss": 2.0289297103881836, + "eval_runtime": 143.6382, + "eval_samples_per_second": 9.691, + "eval_steps_per_second": 0.153, "step": 1472 }, { "epoch": 92.00416666666666, - "grad_norm": 8.202223777770996, + "grad_norm": 5.606525897979736, "learning_rate": 1.2731481481481482e-05, - "loss": 0.8042, + "loss": 0.9074, "step": 1480 }, { "epoch": 92.00833333333334, - "eval_accuracy": 0.5369318181818182, - "eval_loss": 1.950987696647644, - "eval_runtime": 134.0073, - "eval_samples_per_second": 10.507, - "eval_steps_per_second": 0.164, + "eval_accuracy": 0.4920977011494253, + "eval_loss": 2.0539023876190186, + "eval_runtime": 145.5404, + "eval_samples_per_second": 9.564, + "eval_steps_per_second": 0.151, "step": 1488 }, { "epoch": 93.00104166666667, - "grad_norm": 6.163456439971924, + "grad_norm": 7.5830793380737305, "learning_rate": 1.244212962962963e-05, - "loss": 0.8758, + "loss": 0.82, "step": 1490 }, { "epoch": 93.00625, - "grad_norm": 8.015287399291992, + "grad_norm": 7.022946357727051, "learning_rate": 1.2152777777777779e-05, - "loss": 0.7722, + "loss": 0.82, "step": 1500 }, { "epoch": 93.00833333333334, - "eval_accuracy": 0.5333806818181818, - "eval_loss": 1.9502681493759155, - "eval_runtime": 125.0273, - "eval_samples_per_second": 11.262, - "eval_steps_per_second": 0.176, + "eval_accuracy": 0.5014367816091954, + "eval_loss": 2.0480799674987793, + "eval_runtime": 126.661, + "eval_samples_per_second": 10.99, + "eval_steps_per_second": 0.174, "step": 1504 }, { "epoch": 94.003125, - "grad_norm": 5.918884754180908, + "grad_norm": 8.179780960083008, "learning_rate": 1.1863425925925927e-05, - "loss": 0.7579, + "loss": 1.0252, "step": 1510 }, { "epoch": 94.00833333333334, - "grad_norm": 18.762502670288086, + "grad_norm": 18.23526954650879, "learning_rate": 1.1574074074074075e-05, - "loss": 0.6831, + "loss": 0.8318, "step": 1520 }, { "epoch": 94.00833333333334, - "eval_accuracy": 0.5348011363636364, - "eval_loss": 1.9531010389328003, - "eval_runtime": 120.9603, - "eval_samples_per_second": 11.64, - "eval_steps_per_second": 0.182, + "eval_accuracy": 0.5021551724137931, + "eval_loss": 2.030928134918213, + "eval_runtime": 144.5617, + "eval_samples_per_second": 9.629, + "eval_steps_per_second": 0.152, "step": 1520 }, { "epoch": 95.00520833333333, - "grad_norm": 7.426961421966553, + "grad_norm": 5.861065864562988, "learning_rate": 1.1284722222222223e-05, - "loss": 0.766, + "loss": 0.8337, "step": 1530 }, { "epoch": 95.00833333333334, - "eval_accuracy": 0.5383522727272727, - "eval_loss": 1.9345489740371704, - "eval_runtime": 129.3794, - "eval_samples_per_second": 10.883, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.5050287356321839, + "eval_loss": 2.0334632396698, + "eval_runtime": 153.2179, + "eval_samples_per_second": 9.085, + "eval_steps_per_second": 0.144, "step": 1536 }, { "epoch": 96.00208333333333, - "grad_norm": 5.175162315368652, + "grad_norm": 6.104484558105469, "learning_rate": 1.0995370370370372e-05, - "loss": 0.7168, + "loss": 0.8918, "step": 1540 }, { "epoch": 96.00729166666666, - "grad_norm": 7.063689708709717, + "grad_norm": 6.614218711853027, "learning_rate": 1.070601851851852e-05, - "loss": 0.8099, + "loss": 0.9089, "step": 1550 }, { "epoch": 96.00833333333334, - "eval_accuracy": 0.5376420454545454, - "eval_loss": 1.9349414110183716, - "eval_runtime": 123.5725, - "eval_samples_per_second": 11.394, - "eval_steps_per_second": 0.178, + "eval_accuracy": 0.5021551724137931, + "eval_loss": 2.045605421066284, + "eval_runtime": 123.198, + "eval_samples_per_second": 11.299, + "eval_steps_per_second": 0.179, "step": 1552 }, { "epoch": 97.00416666666666, - "grad_norm": 6.602786064147949, + "grad_norm": 6.799188137054443, "learning_rate": 1.0416666666666668e-05, - "loss": 0.7513, + "loss": 0.8189, "step": 1560 }, { "epoch": 97.00833333333334, - "eval_accuracy": 0.5461647727272727, - "eval_loss": 1.9238044023513794, - "eval_runtime": 139.032, - "eval_samples_per_second": 10.127, - "eval_steps_per_second": 0.158, + "eval_accuracy": 0.4985632183908046, + "eval_loss": 2.0107243061065674, + "eval_runtime": 132.087, + "eval_samples_per_second": 10.539, + "eval_steps_per_second": 0.167, "step": 1568 }, { "epoch": 98.00104166666667, - "grad_norm": 5.149336338043213, + "grad_norm": 6.34768533706665, "learning_rate": 1.0127314814814816e-05, - "loss": 0.8236, + "loss": 0.8, "step": 1570 }, { "epoch": 98.00625, - "grad_norm": 8.144251823425293, + "grad_norm": 6.501392364501953, "learning_rate": 9.837962962962964e-06, - "loss": 0.6561, + "loss": 0.7603, "step": 1580 }, { "epoch": 98.00833333333334, - "eval_accuracy": 0.5426136363636364, - "eval_loss": 1.9337557554244995, - "eval_runtime": 133.1537, - "eval_samples_per_second": 10.574, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.507183908045977, + "eval_loss": 2.014706611633301, + "eval_runtime": 107.2213, + "eval_samples_per_second": 12.982, + "eval_steps_per_second": 0.205, "step": 1584 }, { "epoch": 99.003125, - "grad_norm": 5.338318347930908, + "grad_norm": 4.626859664916992, "learning_rate": 9.548611111111111e-06, - "loss": 0.778, + "loss": 0.7984, "step": 1590 }, { "epoch": 99.00833333333334, - "grad_norm": 15.469060897827148, + "grad_norm": 19.75746726989746, "learning_rate": 9.259259259259259e-06, - "loss": 0.7423, + "loss": 0.9197, "step": 1600 }, { "epoch": 99.00833333333334, - "eval_accuracy": 0.5539772727272727, - "eval_loss": 1.901895523071289, - "eval_runtime": 131.3276, - "eval_samples_per_second": 10.721, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.5014367816091954, + "eval_loss": 2.043818712234497, + "eval_runtime": 144.3126, + "eval_samples_per_second": 9.646, + "eval_steps_per_second": 0.152, "step": 1600 }, { "epoch": 100.00520833333333, - "grad_norm": 4.922194957733154, + "grad_norm": 5.76575231552124, "learning_rate": 8.969907407407407e-06, - "loss": 0.7739, + "loss": 0.8021, "step": 1610 }, { "epoch": 100.00833333333334, - "eval_accuracy": 0.5504261363636364, - "eval_loss": 1.916474461555481, - "eval_runtime": 124.6426, - "eval_samples_per_second": 11.296, + "eval_accuracy": 0.5, + "eval_loss": 2.0310871601104736, + "eval_runtime": 124.2032, + "eval_samples_per_second": 11.207, "eval_steps_per_second": 0.177, "step": 1616 }, { "epoch": 101.00208333333333, - "grad_norm": 7.351630210876465, + "grad_norm": 7.446949005126953, "learning_rate": 8.680555555555556e-06, - "loss": 0.6743, + "loss": 0.8517, "step": 1620 }, { "epoch": 101.00729166666666, - "grad_norm": 5.356750965118408, + "grad_norm": 7.860858917236328, "learning_rate": 8.391203703703704e-06, - "loss": 0.6562, + "loss": 0.7474, "step": 1630 }, { "epoch": 101.00833333333334, - "eval_accuracy": 0.5433238636363636, - "eval_loss": 1.9271347522735596, - "eval_runtime": 121.1594, - "eval_samples_per_second": 11.621, - "eval_steps_per_second": 0.182, + "eval_accuracy": 0.5007183908045977, + "eval_loss": 2.0348401069641113, + "eval_runtime": 122.1603, + "eval_samples_per_second": 11.395, + "eval_steps_per_second": 0.18, "step": 1632 }, { "epoch": 102.00416666666666, - "grad_norm": 6.858729839324951, + "grad_norm": 6.173117160797119, "learning_rate": 8.101851851851852e-06, - "loss": 0.7182, + "loss": 0.9423, "step": 1640 }, { "epoch": 102.00833333333334, - "eval_accuracy": 0.5440340909090909, - "eval_loss": 1.9096194505691528, - "eval_runtime": 133.4895, - "eval_samples_per_second": 10.548, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.5007183908045977, + "eval_loss": 2.0177173614501953, + "eval_runtime": 127.9201, + "eval_samples_per_second": 10.882, + "eval_steps_per_second": 0.172, "step": 1648 }, { "epoch": 103.00104166666667, - "grad_norm": 6.165463924407959, + "grad_norm": 7.340484142303467, "learning_rate": 7.8125e-06, - "loss": 0.75, + "loss": 0.8699, "step": 1650 }, { "epoch": 103.00625, - "grad_norm": 7.791365623474121, + "grad_norm": 7.385336875915527, "learning_rate": 7.523148148148148e-06, - "loss": 0.6898, + "loss": 0.8135, "step": 1660 }, { "epoch": 103.00833333333334, - "eval_accuracy": 0.5482954545454546, - "eval_loss": 1.9213480949401855, - "eval_runtime": 121.4191, - "eval_samples_per_second": 11.596, - "eval_steps_per_second": 0.181, + "eval_accuracy": 0.5028735632183908, + "eval_loss": 2.0140013694763184, + "eval_runtime": 147.4404, + "eval_samples_per_second": 9.441, + "eval_steps_per_second": 0.149, "step": 1664 }, { "epoch": 104.003125, - "grad_norm": 7.106443405151367, + "grad_norm": 5.808743000030518, "learning_rate": 7.2337962962962966e-06, - "loss": 0.8279, + "loss": 0.8795, "step": 1670 }, { "epoch": 104.00833333333334, - "grad_norm": 11.49569320678711, + "grad_norm": 20.817060470581055, "learning_rate": 6.944444444444445e-06, - "loss": 0.6541, + "loss": 0.8244, "step": 1680 }, { "epoch": 104.00833333333334, - "eval_accuracy": 0.5433238636363636, - "eval_loss": 1.926275372505188, - "eval_runtime": 128.2374, - "eval_samples_per_second": 10.98, - "eval_steps_per_second": 0.172, + "eval_accuracy": 0.4985632183908046, + "eval_loss": 2.012415885925293, + "eval_runtime": 137.531, + "eval_samples_per_second": 10.121, + "eval_steps_per_second": 0.16, "step": 1680 }, { "epoch": 105.00520833333333, - "grad_norm": 5.9457316398620605, + "grad_norm": 7.121223449707031, "learning_rate": 6.655092592592593e-06, - "loss": 0.7131, + "loss": 0.8446, "step": 1690 }, { "epoch": 105.00833333333334, - "eval_accuracy": 0.546875, - "eval_loss": 1.9148098230361938, - "eval_runtime": 128.976, - "eval_samples_per_second": 10.917, + "eval_accuracy": 0.5064655172413793, + "eval_loss": 2.0021567344665527, + "eval_runtime": 128.7884, + "eval_samples_per_second": 10.808, "eval_steps_per_second": 0.171, "step": 1696 }, { "epoch": 106.00208333333333, - "grad_norm": 7.0640716552734375, + "grad_norm": 6.345254898071289, "learning_rate": 6.365740740740741e-06, - "loss": 0.7, + "loss": 0.7981, "step": 1700 }, { "epoch": 106.00729166666666, - "grad_norm": 5.8036699295043945, + "grad_norm": 8.024818420410156, "learning_rate": 6.076388888888889e-06, - "loss": 0.7076, + "loss": 0.7965, "step": 1710 }, { "epoch": 106.00833333333334, - "eval_accuracy": 0.5454545454545454, - "eval_loss": 1.9192754030227661, - "eval_runtime": 128.1384, - "eval_samples_per_second": 10.988, - "eval_steps_per_second": 0.172, + "eval_accuracy": 0.5107758620689655, + "eval_loss": 1.9956977367401123, + "eval_runtime": 119.9052, + "eval_samples_per_second": 11.609, + "eval_steps_per_second": 0.183, "step": 1712 }, { "epoch": 107.00416666666666, - "grad_norm": 6.425831317901611, + "grad_norm": 5.641129016876221, "learning_rate": 5.787037037037038e-06, - "loss": 0.7822, + "loss": 0.8256, "step": 1720 }, { "epoch": 107.00833333333334, - "eval_accuracy": 0.5440340909090909, - "eval_loss": 1.916624903678894, - "eval_runtime": 128.0714, - "eval_samples_per_second": 10.994, - "eval_steps_per_second": 0.172, + "eval_accuracy": 0.5107758620689655, + "eval_loss": 1.999496579170227, + "eval_runtime": 134.9022, + "eval_samples_per_second": 10.319, + "eval_steps_per_second": 0.163, "step": 1728 }, { "epoch": 108.00104166666667, - "grad_norm": 5.592070579528809, + "grad_norm": 6.308507919311523, "learning_rate": 5.497685185185186e-06, - "loss": 0.7629, + "loss": 0.7253, "step": 1730 }, { "epoch": 108.00625, - "grad_norm": 6.38240909576416, + "grad_norm": 6.363261699676514, "learning_rate": 5.208333333333334e-06, - "loss": 0.6955, + "loss": 0.8448, "step": 1740 }, { "epoch": 108.00833333333334, - "eval_accuracy": 0.5490056818181818, - "eval_loss": 1.9166675806045532, - "eval_runtime": 133.0053, - "eval_samples_per_second": 10.586, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.5093390804597702, + "eval_loss": 2.005643844604492, + "eval_runtime": 143.6016, + "eval_samples_per_second": 9.693, + "eval_steps_per_second": 0.153, "step": 1744 }, { "epoch": 109.003125, - "grad_norm": 4.68496036529541, + "grad_norm": 4.125205993652344, "learning_rate": 4.918981481481482e-06, - "loss": 0.8267, + "loss": 0.8608, "step": 1750 }, { "epoch": 109.00833333333334, - "grad_norm": 22.761470794677734, + "grad_norm": 1.9473612308502197, "learning_rate": 4.6296296296296296e-06, - "loss": 0.6939, + "loss": 0.7144, "step": 1760 }, { "epoch": 109.00833333333334, - "eval_accuracy": 0.5426136363636364, - "eval_loss": 1.9129440784454346, - "eval_runtime": 123.3598, - "eval_samples_per_second": 11.414, - "eval_steps_per_second": 0.178, + "eval_accuracy": 0.507183908045977, + "eval_loss": 2.0083651542663574, + "eval_runtime": 169.1186, + "eval_samples_per_second": 8.231, + "eval_steps_per_second": 0.13, "step": 1760 }, { "epoch": 110.00520833333333, - "grad_norm": 6.633167266845703, + "grad_norm": 7.592528343200684, "learning_rate": 4.340277777777778e-06, - "loss": 0.7149, + "loss": 0.7869, "step": 1770 }, { "epoch": 110.00833333333334, - "eval_accuracy": 0.5355113636363636, - "eval_loss": 1.9237289428710938, - "eval_runtime": 133.1436, - "eval_samples_per_second": 10.575, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.5114942528735632, + "eval_loss": 1.9966795444488525, + "eval_runtime": 122.5349, + "eval_samples_per_second": 11.36, + "eval_steps_per_second": 0.18, "step": 1776 }, { "epoch": 111.00208333333333, - "grad_norm": 6.027252674102783, + "grad_norm": 6.077642440795898, "learning_rate": 4.050925925925926e-06, - "loss": 0.7654, + "loss": 0.924, "step": 1780 }, { "epoch": 111.00729166666666, - "grad_norm": 7.40338134765625, + "grad_norm": 7.033000946044922, "learning_rate": 3.761574074074074e-06, - "loss": 0.7341, + "loss": 0.8149, "step": 1790 }, { "epoch": 111.00833333333334, - "eval_accuracy": 0.5433238636363636, - "eval_loss": 1.9047017097473145, - "eval_runtime": 133.6961, - "eval_samples_per_second": 10.531, - "eval_steps_per_second": 0.165, + "eval_accuracy": 0.5114942528735632, + "eval_loss": 1.997259259223938, + "eval_runtime": 131.0162, + "eval_samples_per_second": 10.625, + "eval_steps_per_second": 0.168, "step": 1792 }, { "epoch": 112.00416666666666, - "grad_norm": 6.874715328216553, + "grad_norm": 6.99604606628418, "learning_rate": 3.4722222222222224e-06, - "loss": 0.7101, + "loss": 0.7896, "step": 1800 }, { "epoch": 112.00833333333334, - "eval_accuracy": 0.5433238636363636, - "eval_loss": 1.9010353088378906, - "eval_runtime": 130.9035, - "eval_samples_per_second": 10.756, - "eval_steps_per_second": 0.168, + "eval_accuracy": 0.5122126436781609, + "eval_loss": 2.0014069080352783, + "eval_runtime": 135.8894, + "eval_samples_per_second": 10.244, + "eval_steps_per_second": 0.162, "step": 1808 }, { "epoch": 113.00104166666667, - "grad_norm": 6.984443664550781, + "grad_norm": 5.9290690422058105, "learning_rate": 3.1828703703703706e-06, - "loss": 0.6031, + "loss": 0.6788, "step": 1810 }, { "epoch": 113.00625, - "grad_norm": 7.135616302490234, + "grad_norm": 5.406201362609863, "learning_rate": 2.893518518518519e-06, - "loss": 0.764, + "loss": 0.8189, "step": 1820 }, { "epoch": 113.00833333333334, - "eval_accuracy": 0.5454545454545454, - "eval_loss": 1.9023563861846924, - "eval_runtime": 128.9627, - "eval_samples_per_second": 10.918, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.514367816091954, + "eval_loss": 1.998863697052002, + "eval_runtime": 120.7041, + "eval_samples_per_second": 11.532, + "eval_steps_per_second": 0.182, "step": 1824 }, { "epoch": 114.003125, - "grad_norm": 5.495314121246338, + "grad_norm": 7.0976881980896, "learning_rate": 2.604166666666667e-06, - "loss": 0.7183, + "loss": 0.7602, "step": 1830 }, { "epoch": 114.00833333333334, - "grad_norm": 25.88104248046875, + "grad_norm": 16.11831283569336, "learning_rate": 2.3148148148148148e-06, - "loss": 0.667, + "loss": 0.6775, "step": 1840 }, { "epoch": 114.00833333333334, - "eval_accuracy": 0.5475852272727273, - "eval_loss": 1.9040805101394653, - "eval_runtime": 132.319, - "eval_samples_per_second": 10.641, - "eval_steps_per_second": 0.166, + "eval_accuracy": 0.5122126436781609, + "eval_loss": 1.9956809282302856, + "eval_runtime": 143.0208, + "eval_samples_per_second": 9.733, + "eval_steps_per_second": 0.154, "step": 1840 }, { "epoch": 115.00520833333333, - "grad_norm": 7.5589280128479, + "grad_norm": 6.661690711975098, "learning_rate": 2.025462962962963e-06, - "loss": 0.7465, + "loss": 0.8642, "step": 1850 }, { "epoch": 115.00833333333334, - "eval_accuracy": 0.5482954545454546, - "eval_loss": 1.9005664587020874, - "eval_runtime": 129.3051, - "eval_samples_per_second": 10.889, - "eval_steps_per_second": 0.17, + "eval_accuracy": 0.5100574712643678, + "eval_loss": 2.000101089477539, + "eval_runtime": 116.9876, + "eval_samples_per_second": 11.899, + "eval_steps_per_second": 0.188, "step": 1856 }, { "epoch": 116.00208333333333, - "grad_norm": 8.857230186462402, + "grad_norm": 5.262021064758301, "learning_rate": 1.7361111111111112e-06, - "loss": 0.5929, + "loss": 0.7406, "step": 1860 }, { "epoch": 116.00729166666666, - "grad_norm": 5.749704837799072, + "grad_norm": 4.934671401977539, "learning_rate": 1.4467592592592594e-06, - "loss": 0.6935, + "loss": 0.7308, "step": 1870 }, { "epoch": 116.00833333333334, - "eval_accuracy": 0.5461647727272727, - "eval_loss": 1.901587724685669, - "eval_runtime": 135.6846, - "eval_samples_per_second": 10.377, + "eval_accuracy": 0.5086206896551724, + "eval_loss": 1.9895482063293457, + "eval_runtime": 135.8417, + "eval_samples_per_second": 10.247, "eval_steps_per_second": 0.162, "step": 1872 }, { "epoch": 117.00416666666666, - "grad_norm": 6.5324883460998535, + "grad_norm": 5.465080261230469, "learning_rate": 1.1574074074074074e-06, - "loss": 0.7306, + "loss": 0.8616, "step": 1880 }, { "epoch": 117.00833333333334, - "eval_accuracy": 0.5482954545454546, - "eval_loss": 1.9008712768554688, - "eval_runtime": 123.3184, - "eval_samples_per_second": 11.418, - "eval_steps_per_second": 0.178, + "eval_accuracy": 0.5079022988505747, + "eval_loss": 1.985381007194519, + "eval_runtime": 117.3279, + "eval_samples_per_second": 11.864, + "eval_steps_per_second": 0.188, "step": 1888 }, { "epoch": 118.00104166666667, - "grad_norm": 6.570457458496094, + "grad_norm": 6.937384128570557, "learning_rate": 8.680555555555556e-07, - "loss": 0.6663, + "loss": 0.7361, "step": 1890 }, { "epoch": 118.00625, - "grad_norm": 4.956265449523926, + "grad_norm": 5.723898887634277, "learning_rate": 5.787037037037037e-07, - "loss": 0.6578, + "loss": 0.7763, "step": 1900 }, { "epoch": 118.00833333333334, - "eval_accuracy": 0.5482954545454546, - "eval_loss": 1.9007748365402222, - "eval_runtime": 136.6271, - "eval_samples_per_second": 10.305, - "eval_steps_per_second": 0.161, + "eval_accuracy": 0.507183908045977, + "eval_loss": 1.9895660877227783, + "eval_runtime": 147.9074, + "eval_samples_per_second": 9.411, + "eval_steps_per_second": 0.149, "step": 1904 }, { "epoch": 119.003125, - "grad_norm": 7.002548694610596, + "grad_norm": 6.339710712432861, "learning_rate": 2.8935185185185185e-07, - "loss": 0.6235, + "loss": 0.7345, "step": 1910 }, { "epoch": 119.00833333333334, - "grad_norm": 2.7440524101257324, + "grad_norm": 20.4032039642334, "learning_rate": 0.0, - "loss": 0.6427, + "loss": 0.8009, "step": 1920 }, { "epoch": 119.00833333333334, - "eval_accuracy": 0.5504261363636364, - "eval_loss": 1.9013855457305908, - "eval_runtime": 128.5584, - "eval_samples_per_second": 10.952, - "eval_steps_per_second": 0.171, + "eval_accuracy": 0.507183908045977, + "eval_loss": 1.9902666807174683, + "eval_runtime": 138.8377, + "eval_samples_per_second": 10.026, + "eval_steps_per_second": 0.158, "step": 1920 }, { "epoch": 119.00833333333334, "step": 1920, "total_flos": 1.4452753827235627e+20, - "train_loss": 1.5160956154266994, - "train_runtime": 26874.3406, - "train_samples_per_second": 4.572, - "train_steps_per_second": 0.071 + "train_loss": 1.618524906039238, + "train_runtime": 28327.5233, + "train_samples_per_second": 4.338, + "train_steps_per_second": 0.068 }, { "epoch": 119.00833333333334, - "eval_accuracy": 0.5539772727272727, - "eval_loss": 1.901895523071289, - "eval_runtime": 137.4484, - "eval_samples_per_second": 10.244, - "eval_steps_per_second": 0.16, + "eval_accuracy": 0.514367816091954, + "eval_loss": 1.998863697052002, + "eval_runtime": 145.2117, + "eval_samples_per_second": 9.586, + "eval_steps_per_second": 0.152, "step": 1920 }, { "epoch": 119.00833333333334, - "eval_accuracy": 0.5539772727272727, - "eval_loss": 1.9018956422805786, - "eval_runtime": 133.7632, - "eval_samples_per_second": 10.526, - "eval_steps_per_second": 0.164, + "eval_accuracy": 0.514367816091954, + "eval_loss": 1.998863697052002, + "eval_runtime": 148.8403, + "eval_samples_per_second": 9.352, + "eval_steps_per_second": 0.148, "step": 1920 } ],