Add details for 12.5B checkpoint

README.md

@@ -38,6 +38,8 @@ in 'float16' rather than 'bfloat16'. Learning rate ramped up 6e-5 to 4e-4
 over the first 3000 iterations (786M tokens) and stayed there for the
 rest of the training process.
 
+![train_val](train_val.png)
+
 ---
 
 Evaluations

ckpt.pt.31000.val_2.6607

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b9f96f574207d59feca0595b7694d2d2d5ba61a6306652a6d7bb0f0d5c875a70
-size 4595163769

train_val.txt

@@ -0,0 +1,483 @@
+Overriding: eval_iters = 50Overriding: eval_iters = 50
+Overriding: eval_interval = 100Overriding: eval_interval = 100
+step 0: train loss 11.0252, val loss 11.0342
+step 100: train loss 8.3994, val loss 8.2066
+step 200: train loss 7.3136, val loss 7.1235
+step 300: train loss 6.5888, val loss 6.7433
+step 400: train loss 6.5067, val loss 6.4013
+step 500: train loss 6.1970, val loss 6.1153
+step 600: train loss 5.9715, val loss 6.0343
+step 700: train loss 5.7357, val loss 5.7946
+step 800: train loss 5.6244, val loss 5.7100
+step 900: train loss 5.4724, val loss 5.5178
+step 1000: train loss 5.4297, val loss 5.3089
+step 1100: train loss 5.0414, val loss 5.2748
+step 1200: train loss 4.9450, val loss 4.9600
+step 1300: train loss 4.6848, val loss 4.8181
+step 1400: train loss 4.5482, val loss 4.4525
+step 1500: train loss 4.4756, val loss 4.3209
+step 1600: train loss 4.2531, val loss 4.2776
+step 1700: train loss 4.2488, val loss 4.2306
+step 1800: train loss 4.0376, val loss 4.1076
+step 1900: train loss 4.0463, val loss 4.0019
+step 2000: train loss 3.9624, val loss 3.8664
+step 2100: train loss 3.9590, val loss 3.7839
+step 2200: train loss 3.9238, val loss 3.8385
+step 2300: train loss 3.6838, val loss 3.7538
+step 2400: train loss 3.7332, val loss 3.6593
+step 2500: train loss 3.7454, val loss 3.5440
+step 2600: train loss 3.5528, val loss 3.6207
+step 2700: train loss 3.5916, val loss 3.6545
+step 2800: train loss 3.7254, val loss 3.6136
+step 2900: train loss 3.5898, val loss 3.3846
+step 3000: train loss 3.5164, val loss 3.4608
+step 3100: train loss 3.6373, val loss 3.5505
+step 3200: train loss 3.5100, val loss 3.6281
+step 3300: train loss 3.5623, val loss 3.5894
+step 3400: train loss 3.4841, val loss 3.4290
+step 3500: train loss 3.5908, val loss 3.4267
+step 3600: train loss 3.4661, val loss 3.5482
+step 3700: train loss 3.4633, val loss 3.4274
+step 3800: train loss 3.4503, val loss 3.5384
+step 3900: train loss 3.3948, val loss 3.3274
+step 4000: train loss 3.4388, val loss 3.3746
+step 4100: train loss 3.3921, val loss 3.2486
+step 4200: train loss 3.4422, val loss 3.3624
+step 4300: train loss 3.3533, val loss 3.2563
+step 4400: train loss 3.3215, val loss 3.3935
+step 4500: train loss 3.4373, val loss 3.2724
+step 4600: train loss 3.2562, val loss 3.2819
+step 4700: train loss 3.3209, val loss 3.2646
+step 4800: train loss 3.1498, val loss 3.3252
+step 4900: train loss 3.3318, val loss 3.3322
+step 5000: train loss 3.1285, val loss 3.2495
+step 5100: train loss 3.3448, val loss 3.1907
+step 5200: train loss 3.3123, val loss 3.1915
+step 5300: train loss 3.2482, val loss 3.3080
+step 5400: train loss 3.0714, val loss 3.1940
+step 5500: train loss 3.1294, val loss 3.2508
+step 5600: train loss 3.2360, val loss 3.0566
+step 5700: train loss 3.2703, val loss 3.1624
+step 5800: train loss 3.3135, val loss 3.1183
+step 5900: train loss 3.2142, val loss 3.1934
+step 6000: train loss 3.2289, val loss 3.1825
+step 6100: train loss 3.0920, val loss 3.1858
+step 6200: train loss 3.2835, val loss 3.1578
+step 6300: train loss 3.1277, val loss 3.1348
+step 6400: train loss 3.0799, val loss 3.2929
+step 6500: train loss 3.0791, val loss 3.2397
+step 6600: train loss 3.2201, val loss 3.2587
+step 6700: train loss 3.0092, val loss 3.2005
+step 6800: train loss 3.0824, val loss 3.0970
+step 6900: train loss 3.2339, val loss 3.1762
+step 7000: train loss 3.1754, val loss 3.1966
+step 7100: train loss 3.1720, val loss 3.1533
+step 7200: train loss 3.1673, val loss 3.1003
+step 7300: train loss 3.1047, val loss 3.1397
+step 7400: train loss 3.1211, val loss 3.1447
+step 7500: train loss 3.1564, val loss 3.0936
+step 7600: train loss 3.0931, val loss 3.1315
+step 7700: train loss 2.9800, val loss 3.2394
+step 7800: train loss 3.1775, val loss 3.2620
+step 7900: train loss 3.0847, val loss 3.0954
+step 8000: train loss 3.0581, val loss 3.0713
+step 8100: train loss 3.1880, val loss 3.0542
+step 8200: train loss 3.1568, val loss 3.0514
+step 8300: train loss 3.0128, val loss 3.1295
+step 8400: train loss 3.2077, val loss 3.0505
+step 8500: train loss 3.0058, val loss 3.1052
+step 8600: train loss 3.0915, val loss 2.8884
+step 8700: train loss 3.1190, val loss 3.0491
+step 8800: train loss 2.9319, val loss 2.9831
+step 8900: train loss 2.9605, val loss 3.0030
+step 9000: train loss 3.0953, val loss 2.9161
+step 9100: train loss 3.1344, val loss 3.0248
+step 9200: train loss 2.9525, val loss 3.0419
+step 9300: train loss 2.9842, val loss 2.9508
+step 9400: train loss 3.1642, val loss 3.0025
+step 9500: train loss 2.9276, val loss 3.0674
+step 9600: train loss 3.0968, val loss 3.0211
+step 9700: train loss 3.1166, val loss 3.0580
+step 9800: train loss 2.9912, val loss 2.9596
+step 9900: train loss 2.9809, val loss 2.9561
+step 10000: train loss 3.0402, val loss 2.9424
+step 10100: train loss 2.9393, val loss 2.9690
+step 10200: train loss 3.0273, val loss 3.0578
+step 10300: train loss 2.9466, val loss 3.1119
+step 10400: train loss 2.9821, val loss 2.9871
+step 10500: train loss 3.0022, val loss 3.0068
+step 10600: train loss 2.9527, val loss 3.0174
+step 10700: train loss 3.0224, val loss 3.0772
+step 10800: train loss 2.9642, val loss 3.0270
+step 10900: train loss 2.9446, val loss 2.9751
+step 11000: train loss 2.9466, val loss 2.9945
+step 11100: train loss 2.9304, val loss 2.9444
+step 11200: train loss 2.9619, val loss 3.0315
+step 11300: train loss 2.9358, val loss 2.9847
+step 11400: train loss 3.0165, val loss 2.7416
+step 11500: train loss 2.8405, val loss 3.0835
+step 11600: train loss 3.0746, val loss 3.0534
+step 11700: train loss 2.9898, val loss 2.9221
+step 11800: train loss 2.8608, val loss 3.0250
+step 11900: train loss 2.9855, val loss 2.9443
+step 12000: train loss 2.9834, val loss 2.9962
+step 12100: train loss 2.8355, val loss 3.0118
+step 12200: train loss 2.9886, val loss 2.9714
+step 12300: train loss 2.9457, val loss 2.9599
+step 12400: train loss 2.8276, val loss 3.0673
+step 12500: train loss 2.9246, val loss 2.9800
+step 12600: train loss 3.0029, val loss 2.8929
+step 12700: train loss 2.9373, val loss 2.9386
+step 12800: train loss 2.9504, val loss 3.0079
+step 12900: train loss 2.9921, val loss 2.9243
+step 13000: train loss 2.9724, val loss 3.0502
+step 13100: train loss 2.9558, val loss 2.8818
+step 13200: train loss 2.9938, val loss 2.9503
+step 13300: train loss 2.9165, val loss 3.0683
+step 13400: train loss 2.9777, val loss 2.9374
+step 13500: train loss 3.0141, val loss 2.9254
+step 13600: train loss 2.8655, val loss 2.9531
+step 13700: train loss 2.8848, val loss 3.0087
+step 13800: train loss 2.9226, val loss 2.8738
+step 13900: train loss 2.8910, val loss 2.9250
+step 14000: train loss 2.9752, val loss 2.9531
+step 14100: train loss 2.9497, val loss 2.9894
+step 14200: train loss 2.9779, val loss 2.9911
+step 14300: train loss 2.9423, val loss 2.7897
+step 14400: train loss 3.0338, val loss 3.0204
+step 14500: train loss 2.8680, val loss 2.8760
+step 14600: train loss 3.0093, val loss 2.8034
+step 14700: train loss 2.9222, val loss 2.8466
+step 14800: train loss 2.7877, val loss 2.8845
+step 14900: train loss 2.9715, val loss 3.0005
+step 15000: train loss 3.0018, val loss 3.0310
+step 15100: train loss 2.9654, val loss 2.9176
+step 15200: train loss 2.9580, val loss 2.9125
+step 15300: train loss 3.0046, val loss 2.8712
+step 15400: train loss 3.0046, val loss 2.9361
+step 15500: train loss 2.7949, val loss 2.8170
+step 15600: train loss 2.9127, val loss 2.9011
+step 15700: train loss 2.9440, val loss 2.9167
+step 15800: train loss 2.8596, val loss 2.8605
+step 15900: train loss 2.8704, val loss 2.8725
+step 16000: train loss 2.8634, val loss 3.0975
+step 16100: train loss 2.9963, val loss 2.7633
+step 16200: train loss 2.9618, val loss 2.9352
+step 16300: train loss 2.7306, val loss 2.9384
+step 16400: train loss 2.9731, val loss 2.9716
+step 16500: train loss 2.8599, val loss 3.0492
+step 16600: train loss 2.8712, val loss 2.9475
+step 16700: train loss 2.9567, val loss 2.8846
+step 16800: train loss 2.8565, val loss 3.0182
+step 16900: train loss 2.8318, val loss 3.0222
+step 17000: train loss 3.0119, val loss 2.8964
+step 17100: train loss 2.8578, val loss 2.7679
+step 17200: train loss 2.8943, val loss 2.9294
+step 17300: train loss 2.8835, val loss 2.8658
+step 17400: train loss 2.9415, val loss 2.9057
+step 17500: train loss 2.8730, val loss 2.7631
+step 17600: train loss 2.7918, val loss 2.7859
+step 17700: train loss 2.9455, val loss 2.9624
+step 17800: train loss 2.7874, val loss 2.8241
+step 17900: train loss 2.9045, val loss 2.8924
+step 18000: train loss 2.6872, val loss 2.9278
+step 18100: train loss 2.9407, val loss 2.9969
+step 18200: train loss 3.0288, val loss 2.9354
+step 18300: train loss 2.8862, val loss 2.8489
+step 18400: train loss 2.8283, val loss 2.8086
+step 18500: train loss 2.8491, val loss 2.8545
+step 18600: train loss 2.8140, val loss 2.9770
+step 18700: train loss 2.9287, val loss 2.8787
+step 18800: train loss 3.0498, val loss 2.7461
+step 18900: train loss 2.9223, val loss 2.8665
+step 19000: train loss 2.9418, val loss 2.9149
+step 19100: train loss 2.6789, val loss 2.9049
+step 19200: train loss 2.8974, val loss 2.8892
+step 19300: train loss 2.8448, val loss 2.9557
+step 19400: train loss 2.8466, val loss 2.9635
+step 19500: train loss 2.8872, val loss 2.8272
+step 19600: train loss 2.7967, val loss 3.0509
+step 19700: train loss 2.8516, val loss 2.7520
+step 19800: train loss 3.0064, val loss 2.8897
+step 19900: train loss 2.8801, val loss 2.9297
+step 20000: train loss 2.8270, val loss 2.9379
+step 20100: train loss 2.8988, val loss 2.8314
+step 20200: train loss 2.6983, val loss 2.9195
+step 20300: train loss 2.8345, val loss 2.8455
+step 20400: train loss 2.7777, val loss 2.9164
+step 20500: train loss 2.9010, val loss 2.8442
+step 20600: train loss 2.8983, val loss 2.8687
+step 20700: train loss 2.7852, val loss 2.8359
+step 20800: train loss 2.6776, val loss 2.8802
+step 20900: train loss 2.7957, val loss 2.9362
+step 21000: train loss 2.8322, val loss 2.8738
+step 21100: train loss 2.8448, val loss 2.8849
+step 21200: train loss 2.9563, val loss 3.0302
+step 21300: train loss 2.9416, val loss 2.7907
+step 21400: train loss 2.7988, val loss 2.8956
+step 21500: train loss 2.8556, val loss 2.8462
+step 21600: train loss 2.8326, val loss 2.8084
+step 21700: train loss 2.8916, val loss 2.9479
+step 21800: train loss 2.6759, val loss 2.8316
+step 21900: train loss 2.7605, val loss 2.8726
+step 22000: train loss 2.8973, val loss 2.7646
+step 22100: train loss 2.7950, val loss 2.8894
+step 22200: train loss 2.8879, val loss 2.8456
+step 22300: train loss 2.8610, val loss 2.7752
+step 22400: train loss 2.8503, val loss 2.7268
+step 22500: train loss 2.7624, val loss 2.8039
+step 22600: train loss 2.7896, val loss 2.9268
+step 22700: train loss 2.9371, val loss 2.8718
+step 22800: train loss 2.9747, val loss 2.7481
+step 22900: train loss 2.8736, val loss 2.8353
+step 23000: train loss 2.8346, val loss 2.7387
+step 23100: train loss 2.8266, val loss 2.9682
+step 23200: train loss 2.8811, val loss 2.8276
+step 23300: train loss 2.8492, val loss 2.7715
+step 23400: train loss 2.9512, val loss 2.8733
+step 23500: train loss 2.8948, val loss 2.8610
+step 23600: train loss 2.9883, val loss 2.8248
+step 23700: train loss 2.7142, val loss 2.9138
+step 23800: train loss 2.7128, val loss 2.8417
+step 23900: train loss 3.0065, val loss 2.8004
+step 24000: train loss 2.8458, val loss 2.7381
+step 24100: train loss 2.7890, val loss 2.8468
+step 24200: train loss 2.9545, val loss 2.7933
+step 24300: train loss 2.8738, val loss 2.9072
+step 24400: train loss 2.8440, val loss 2.7552
+step 24500: train loss 2.8107, val loss 2.7479
+step 24600: train loss 2.8175, val loss 2.8063
+step 24700: train loss 2.9319, val loss 2.8145
+step 24800: train loss 2.8535, val loss 2.8273
+step 24900: train loss 2.7535, val loss 2.9339
+step 25000: train loss 2.7998, val loss 2.8346
+step 25100: train loss 2.8028, val loss 2.7334
+step 25200: train loss 3.0190, val loss 2.7507
+step 25300: train loss 2.9597, val loss 2.7477
+step 25400: train loss 3.0206, val loss 2.8678
+step 25500: train loss 2.8184, val loss 2.8603
+step 25600: train loss 2.8984, val loss 2.7563
+step 25700: train loss 2.7563, val loss 2.8466
+step 25800: train loss 2.8035, val loss 2.8461
+step 25900: train loss 2.8879, val loss 3.0032
+step 26000: train loss 2.8628, val loss 2.8316
+step 26100: train loss 2.8199, val loss 2.8175
+step 26200: train loss 2.8381, val loss 2.7543
+step 26300: train loss 2.7932, val loss 2.7437
+step 26400: train loss 2.7451, val loss 2.8037
+step 26500: train loss 2.8398, val loss 2.7688
+step 26600: train loss 2.8197, val loss 2.6988
+step 26700: train loss 2.8181, val loss 2.8315
+step 26800: train loss 2.7584, val loss 2.6994
+step 26900: train loss 2.7917, val loss 2.7537
+step 27000: train loss 2.6462, val loss 2.7579
+step 27100: train loss 2.8499, val loss 2.7959
+step 27200: train loss 2.8724, val loss 2.8232
+step 27300: train loss 2.7593, val loss 2.8665
+step 27400: train loss 2.8588, val loss 2.9407
+step 27500: train loss 2.7949, val loss 2.6853
+step 27600: train loss 2.7752, val loss 2.8110
+step 27700: train loss 2.9131, val loss 2.9227
+step 27800: train loss 2.7813, val loss 2.7983
+step 27900: train loss 2.7238, val loss 2.9116
+step 28000: train loss 2.6029, val loss 2.6874
+step 28100: train loss 2.7992, val loss 2.8840
+step 28200: train loss 2.8726, val loss 2.7155
+step 28300: train loss 2.8896, val loss 2.7741
+step 28400: train loss 2.8420, val loss 2.7712
+step 28500: train loss 2.7476, val loss 2.8297
+step 28600: train loss 2.8152, val loss 2.8123
+step 28700: train loss 2.8929, val loss 2.8723
+step 28800: train loss 2.8116, val loss 2.8850
+step 28900: train loss 2.8026, val loss 2.8580
+step 29000: train loss 2.6830, val loss 2.7671
+step 29100: train loss 2.7769, val loss 2.8252
+step 29200: train loss 2.8928, val loss 2.7823
+step 29300: train loss 2.7859, val loss 2.8006
+step 29400: train loss 2.8484, val loss 2.8032
+step 29500: train loss 2.8194, val loss 2.7389
+step 29600: train loss 2.8775, val loss 2.8360
+step 29700: train loss 2.7912, val loss 2.7585
+step 29800: train loss 2.8499, val loss 2.8210
+step 29900: train loss 2.9061, val loss 2.6846
+step 30000: train loss 2.7540, val loss 2.8391
+step 30100: train loss 2.8292, val loss 2.8358
+step 30200: train loss 2.5902, val loss 2.8730
+step 30300: train loss 2.8947, val loss 2.8475
+step 30400: train loss 2.8898, val loss 2.7538
+step 30500: train loss 2.8530, val loss 2.8979
+step 30600: train loss 2.8079, val loss 2.8202
+step 30700: train loss 2.6925, val loss 2.7329
+step 30800: train loss 2.7408, val loss 2.7117
+step 30900: train loss 2.7052, val loss 2.8759
+step 31000: train loss 2.7108, val loss 2.6607
+step 31100: train loss 2.8145, val loss 2.7848
+step 31200: train loss 2.8752, val loss 2.8979
+step 31300: train loss 2.6798, val loss 2.8022
+step 31400: train loss 2.9750, val loss 2.6888
+step 31500: train loss 2.6494, val loss 2.8619
+step 31600: train loss 2.8156, val loss 2.8232
+step 31700: train loss 2.7252, val loss 2.7410
+step 31800: train loss 2.6924, val loss 2.7541
+step 31900: train loss 2.8176, val loss 2.9296
+step 32000: train loss 2.8469, val loss 2.8549
+step 32100: train loss 2.8750, val loss 2.9075
+step 32200: train loss 2.8387, val loss 2.7277
+step 32300: train loss 2.7656, val loss 2.7939
+step 32400: train loss 2.6632, val loss 2.7976
+step 32500: train loss 2.7674, val loss 2.7517
+step 32600: train loss 2.8411, val loss 2.7297
+step 32700: train loss 2.8641, val loss 2.7247
+step 32800: train loss 2.6665, val loss 2.7943
+step 32900: train loss 2.8883, val loss 2.7321
+step 33000: train loss 2.8978, val loss 2.7700
+step 33100: train loss 2.7607, val loss 2.6791
+step 33200: train loss 2.7516, val loss 2.8169
+step 33300: train loss 2.8498, val loss 2.6707
+step 33400: train loss 2.8504, val loss 2.9119
+step 33500: train loss 2.7596, val loss 2.9151
+step 33600: train loss 2.9359, val loss 2.9191
+step 33700: train loss 2.7263, val loss 2.8193
+step 33800: train loss 2.8230, val loss 2.8280
+step 33900: train loss 2.8378, val loss 2.7144
+step 34000: train loss 2.7823, val loss 2.8035
+step 34100: train loss 2.7779, val loss 2.8396
+step 34200: train loss 2.8372, val loss 2.8954
+step 34300: train loss 2.8226, val loss 2.6627
+step 34400: train loss 2.8642, val loss 2.8739
+step 34500: train loss 2.7282, val loss 2.6650
+step 34600: train loss 2.7650, val loss 2.7226
+step 34700: train loss 2.7236, val loss 2.6892
+step 34800: train loss 2.7721, val loss 2.9387
+step 34900: train loss 2.7465, val loss 2.7535
+step 35000: train loss 2.7129, val loss 2.7230
+step 35100: train loss 2.7448, val loss 2.7261
+step 35200: train loss 2.9534, val loss 2.7127
+step 35300: train loss 2.6951, val loss 2.8034
+step 35400: train loss 2.8718, val loss 2.7998
+step 35500: train loss 2.7152, val loss 2.7406
+step 35600: train loss 2.8066, val loss 2.7981
+step 35700: train loss 2.8076, val loss 2.7320
+step 35800: train loss 2.9054, val loss 2.7541
+step 35900: train loss 2.8348, val loss 2.6628
+step 36000: train loss 2.7294, val loss 2.7758
+step 36100: train loss 2.8457, val loss 2.8148
+step 36200: train loss 2.8626, val loss 2.8337
+step 36300: train loss 2.7538, val loss 2.8294
+step 36400: train loss 2.5631, val loss 2.7590
+step 36500: train loss 2.8542, val loss 2.7585
+step 36600: train loss 2.7567, val loss 2.8492
+step 36700: train loss 2.8481, val loss 2.7103
+step 36800: train loss 2.8135, val loss 2.7256
+step 36900: train loss 2.6976, val loss 2.6366
+step 37000: train loss 2.8643, val loss 2.7390
+step 37100: train loss 2.7979, val loss 2.6219
+step 37200: train loss 2.7855, val loss 2.8387
+step 37300: train loss 2.8332, val loss 2.8489
+step 37400: train loss 2.6962, val loss 2.9051
+step 37500: train loss 2.7735, val loss 2.8329
+step 37600: train loss 2.8305, val loss 2.7830
+step 37700: train loss 2.7930, val loss 2.7070
+step 37800: train loss 2.7834, val loss 2.7718
+step 37900: train loss 2.9645, val loss 2.7499
+step 38000: train loss 2.6900, val loss 2.8002
+step 38100: train loss 2.7324, val loss 2.8638
+step 38200: train loss 2.6724, val loss 2.7601
+step 38300: train loss 2.8456, val loss 2.7571
+step 38400: train loss 2.7720, val loss 2.8515
+step 38500: train loss 2.7960, val loss 2.8611
+step 38600: train loss 2.7673, val loss 2.8128
+step 38700: train loss 2.8076, val loss 2.8023
+step 38800: train loss 2.8252, val loss 2.7761
+step 38900: train loss 2.6206, val loss 2.8931
+step 39000: train loss 2.7810, val loss 2.6949
+step 39100: train loss 2.8880, val loss 2.6300
+step 39200: train loss 2.7765, val loss 2.8009
+step 39300: train loss 2.8100, val loss 2.9730
+step 39400: train loss 2.6373, val loss 2.7640
+step 39500: train loss 2.7533, val loss 2.7617
+step 39600: train loss 2.8452, val loss 2.8122
+step 39700: train loss 2.7849, val loss 2.8067
+step 39800: train loss 2.7890, val loss 2.7672
+step 39900: train loss 2.7164, val loss 2.6389
+step 40000: train loss 2.8189, val loss 2.7924
+step 40100: train loss 2.9345, val loss 2.9801
+step 40200: train loss 2.9074, val loss 2.7438
+step 40300: train loss 2.8472, val loss 2.7186
+step 40400: train loss 2.5992, val loss 2.7979
+step 40500: train loss 2.8513, val loss 2.7371
+step 40600: train loss 2.6937, val loss 2.7330
+step 40700: train loss 2.7758, val loss 2.7263
+step 40800: train loss 2.7242, val loss 2.8467
+step 40900: train loss 2.7578, val loss 2.9498
+step 41000: train loss 2.7946, val loss 2.7555
+step 41100: train loss 2.8186, val loss 2.7127
+step 41200: train loss 2.7768, val loss 2.7014
+step 41300: train loss 2.8141, val loss 2.7691
+step 41400: train loss 2.7520, val loss 2.6608
+step 41500: train loss 2.7952, val loss 2.8809
+step 41600: train loss 2.7405, val loss 2.8320
+step 41700: train loss 2.7319, val loss 2.6906
+step 41800: train loss 2.7042, val loss 2.8355
+step 41900: train loss 2.6836, val loss 2.7683
+step 42000: train loss 2.8002, val loss 2.7833
+step 42100: train loss 2.9250, val loss 2.7595
+step 42200: train loss 2.6998, val loss 2.8130
+step 42300: train loss 2.6696, val loss 2.7072
+step 42400: train loss 2.6971, val loss 2.7896
+step 42500: train loss 2.7793, val loss 2.8207
+step 42600: train loss 2.7416, val loss 2.6938
+step 42700: train loss 2.5605, val loss 2.8192
+step 42800: train loss 2.8029, val loss 2.6802
+step 42900: train loss 2.8314, val loss 2.7868
+step 43000: train loss 2.7065, val loss 2.5963
+step 43100: train loss 2.8072, val loss 2.7424
+step 43200: train loss 2.6797, val loss 2.7166
+step 43300: train loss 2.6579, val loss 2.7534
+step 43400: train loss 2.8590, val loss 2.8177
+step 43500: train loss 2.7240, val loss 2.8758
+step 43600: train loss 2.8024, val loss 2.7224
+step 43700: train loss 2.8347, val loss 2.7132
+step 43800: train loss 2.8055, val loss 2.6904
+step 43900: train loss 2.7516, val loss 2.7553
+step 44000: train loss 2.7896, val loss 2.7832
+step 44100: train loss 2.8472, val loss 2.7570
+step 44200: train loss 2.6282, val loss 2.6458
+step 44300: train loss 2.7891, val loss 2.6897
+step 44400: train loss 2.8262, val loss 2.7445
+step 44500: train loss 2.7764, val loss 2.7653
+step 44600: train loss 2.8129, val loss 2.7805
+step 44700: train loss 2.8649, val loss 2.8448
+step 44800: train loss 2.6760, val loss 2.7656
+step 44900: train loss 2.7011, val loss 2.7474
+step 45000: train loss 2.7879, val loss 2.6947
+step 45100: train loss 2.9080, val loss 2.7905
+step 45200: train loss 2.7495, val loss 2.7055
+step 45300: train loss 2.6580, val loss 2.8663
+step 45400: train loss 2.8094, val loss 2.8226
+step 45500: train loss 2.7298, val loss 2.8190
+step 45600: train loss 2.7434, val loss 2.6559
+step 45700: train loss 2.8474, val loss 2.7221
+step 45800: train loss 2.8787, val loss 2.8628
+step 45900: train loss 2.7202, val loss 2.6398
+step 46000: train loss 2.8298, val loss 2.8447
+step 46100: train loss 2.6955, val loss 2.8386
+step 46200: train loss 2.7849, val loss 2.6825
+step 46300: train loss 2.8191, val loss 2.7793
+step 46400: train loss 2.7815, val loss 2.7403
+step 46500: train loss 2.8007, val loss 2.7719
+step 46600: train loss 2.6661, val loss 2.8360
+step 46700: train loss 2.8279, val loss 2.7529
+step 46800: train loss 2.8326, val loss 2.7180
+step 46900: train loss 2.7323, val loss 2.8723
+step 47000: train loss 2.7846, val loss 2.7797
+step 47100: train loss 2.7533, val loss 2.7694
+step 47200: train loss 2.8556, val loss 2.7418
+step 47300: train loss 2.7036, val loss 2.7377
+step 47400: train loss 2.7860, val loss 2.8879
+step 47500: train loss 2.7223, val loss 2.8096
+step 47600: train loss 2.7241, val loss 2.9097
+step 47700: train loss 2.6891, val loss 2.8653
+step 47800: train loss 2.7393, val loss 2.7434
+step 47900: train loss 2.8288, val loss 2.6818
+step 48000: train loss 2.7092, val loss 2.7769