| { | |
| "best_metric": 0.3302349377015186, | |
| "best_model_checkpoint": "esm2_t12_35M_lora_binding_sites_2023-09-23_03-04-43/checkpoint-102604", | |
| "epoch": 1.0, | |
| "eval_steps": 500, | |
| "global_step": 102604, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00037015336998624066, | |
| "loss": 0.511, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0003701429941921846, | |
| "loss": 0.3377, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0003701256784994761, | |
| "loss": 0.3081, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00037010142355745155, | |
| "loss": 0.2888, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0003700702302756681, | |
| "loss": 0.272, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00037003209982386994, | |
| "loss": 0.2463, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0003699870336319441, | |
| "loss": 0.2251, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00036993503338986704, | |
| "loss": 0.2418, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0003698761010476413, | |
| "loss": 0.2225, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00036981023881522214, | |
| "loss": 0.2178, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00036973783033824974, | |
| "loss": 0.2234, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0003696581506109316, | |
| "loss": 0.1971, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00036957154916653366, | |
| "loss": 0.2054, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0003694780292525993, | |
| "loss": 0.2007, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00036937759437611316, | |
| "loss": 0.2011, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0003692702483033705, | |
| "loss": 0.199, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00036915599505983534, | |
| "loss": 0.1818, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0003690348389299898, | |
| "loss": 0.1932, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0003689074418813296, | |
| "loss": 0.2048, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0003687725283229007, | |
| "loss": 0.1888, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0003686307262581153, | |
| "loss": 0.1797, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0003684820410045328, | |
| "loss": 0.1881, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0003683264781378313, | |
| "loss": 0.1763, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00036816404349159854, | |
| "loss": 0.1855, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003679947431571131, | |
| "loss": 0.1742, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003678185834831161, | |
| "loss": 0.1748, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003676355710755728, | |
| "loss": 0.1668, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00036744571279742524, | |
| "loss": 0.1692, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003672490157683347, | |
| "loss": 0.1648, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00036704548736441486, | |
| "loss": 0.1657, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003668351352179549, | |
| "loss": 0.1606, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00036661796721713366, | |
| "loss": 0.1572, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003663939915057234, | |
| "loss": 0.1688, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000366163216482785, | |
| "loss": 0.163, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00036592565080235246, | |
| "loss": 0.1646, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003656813033731085, | |
| "loss": 0.1616, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003654301833580507, | |
| "loss": 0.164, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003651723001741478, | |
| "loss": 0.1687, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00036490900345841326, | |
| "loss": 0.1527, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003646376568946032, | |
| "loss": 0.1566, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003643595768815895, | |
| "loss": 0.1517, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00036407621456846186, | |
| "loss": 0.1483, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00036378473272777155, | |
| "loss": 0.1549, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00036348654942242547, | |
| "loss": 0.1531, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003631816758342598, | |
| "loss": 0.1522, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003628701233959952, | |
| "loss": 0.1649, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003625519037908086, | |
| "loss": 0.1421, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00036222702895189466, | |
| "loss": 0.1503, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003618955110620186, | |
| "loss": 0.1541, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003615573625530591, | |
| "loss": 0.159, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00036121259610554234, | |
| "loss": 0.1606, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00036086122464816625, | |
| "loss": 0.1547, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00036050326135731575, | |
| "loss": 0.1493, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00036013871965656874, | |
| "loss": 0.1463, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003597676132161928, | |
| "loss": 0.1599, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00035939186051079785, | |
| "loss": 0.1577, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003590076992338234, | |
| "loss": 0.1472, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00035861701563034195, | |
| "loss": 0.1449, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003582198243509391, | |
| "loss": 0.1362, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003578161402902372, | |
| "loss": 0.1365, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00035740597858633733, | |
| "loss": 0.1621, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00035698935462025107, | |
| "loss": 0.1458, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003565662840153244, | |
| "loss": 0.1536, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003561367826366511, | |
| "loss": 0.1459, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003557008665904783, | |
| "loss": 0.1464, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00035525855222360236, | |
| "loss": 0.1403, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00035480985612275573, | |
| "loss": 0.1448, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003543547951139852, | |
| "loss": 0.1433, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003538933862620208, | |
| "loss": 0.1488, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00035342564686963564, | |
| "loss": 0.1417, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00035295159447699774, | |
| "loss": 0.1327, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003524712468610113, | |
| "loss": 0.1454, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003519846220346511, | |
| "loss": 0.143, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00035149173824628615, | |
| "loss": 0.1376, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00035099261397899585, | |
| "loss": 0.1446, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00035048981012540415, | |
| "loss": 0.1521, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000349978292251401, | |
| "loss": 0.1337, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000349460590652505, | |
| "loss": 0.1307, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003489367247424602, | |
| "loss": 0.1368, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00034840671416617173, | |
| "loss": 0.1271, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003478705787989687, | |
| "loss": 0.1396, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003473283387458593, | |
| "loss": 0.1421, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003467800143407766, | |
| "loss": 0.151, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003462256261458161, | |
| "loss": 0.1343, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00034566519495046456, | |
| "loss": 0.1352, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003450987417708206, | |
| "loss": 0.1194, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00034452916500984617, | |
| "loss": 0.1401, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.000343950761655018, | |
| "loss": 0.1438, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003433664006069281, | |
| "loss": 0.1406, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003427761037790415, | |
| "loss": 0.1493, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003421798933074141, | |
| "loss": 0.1341, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00034157779154986285, | |
| "loss": 0.1321, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00034096982108512724, | |
| "loss": 0.1342, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00034035600471202216, | |
| "loss": 0.1304, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003397363654485838, | |
| "loss": 0.1437, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00033911092653120565, | |
| "loss": 0.1249, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003384797114137677, | |
| "loss": 0.1226, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00033784274376675655, | |
| "loss": 0.136, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.000337200047476378, | |
| "loss": 0.1373, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003365516466436611, | |
| "loss": 0.1386, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003359008500779356, | |
| "loss": 0.1671, | |
| "step": 20200 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00033524114153552924, | |
| "loss": 0.1352, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003345758019095085, | |
| "loss": 0.1252, | |
| "step": 20600 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00033390485615002473, | |
| "loss": 0.1447, | |
| "step": 20800 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00033322832941745856, | |
| "loss": 0.1242, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003325462470814765, | |
| "loss": 0.133, | |
| "step": 21200 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00033185863472007953, | |
| "loss": 0.1408, | |
| "step": 21400 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00033116551811864366, | |
| "loss": 0.1309, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003304669232689534, | |
| "loss": 0.1205, | |
| "step": 21800 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003297628763682268, | |
| "loss": 0.1218, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003290569646332609, | |
| "loss": 0.1538, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00032834211996763616, | |
| "loss": 0.1305, | |
| "step": 22400 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003276219029308289, | |
| "loss": 0.1412, | |
| "step": 22600 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003268963405308863, | |
| "loss": 0.1465, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00032616545997630613, | |
| "loss": 0.1264, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00032542928867501625, | |
| "loss": 0.1423, | |
| "step": 23200 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000324687854233347, | |
| "loss": 0.1356, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003239411844549956, | |
| "loss": 0.1384, | |
| "step": 23600 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00032318930733998385, | |
| "loss": 0.1441, | |
| "step": 23800 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000322432251083608, | |
| "loss": 0.1316, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003216700440753813, | |
| "loss": 0.1202, | |
| "step": 24200 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00032090271489796984, | |
| "loss": 0.1345, | |
| "step": 24400 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003201302923261202, | |
| "loss": 0.14, | |
| "step": 24600 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00031935280532558054, | |
| "loss": 0.1234, | |
| "step": 24800 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00031857028305201467, | |
| "loss": 0.1239, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00031778275484990823, | |
| "loss": 0.1352, | |
| "step": 25200 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00031699025025146887, | |
| "loss": 0.1283, | |
| "step": 25400 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003161927989755181, | |
| "loss": 0.1228, | |
| "step": 25600 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00031539043092637745, | |
| "loss": 0.1318, | |
| "step": 25800 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00031458317619274687, | |
| "loss": 0.1195, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003137710650465762, | |
| "loss": 0.1199, | |
| "step": 26200 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00031295822458142394, | |
| "loss": 0.14, | |
| "step": 26400 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003121365160534547, | |
| "loss": 0.1252, | |
| "step": 26600 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003113100428623896, | |
| "loss": 0.1377, | |
| "step": 26800 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00031047883600086813, | |
| "loss": 0.1306, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00030964292663904203, | |
| "loss": 0.1279, | |
| "step": 27200 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003088023461234062, | |
| "loss": 0.1364, | |
| "step": 27400 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00030795712597562287, | |
| "loss": 0.1277, | |
| "step": 27600 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003071072978913404, | |
| "loss": 0.1279, | |
| "step": 27800 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00030625289373900367, | |
| "loss": 0.1232, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00030539394555866, | |
| "loss": 0.1126, | |
| "step": 28200 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003045304855607569, | |
| "loss": 0.1346, | |
| "step": 28400 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00030366254612493445, | |
| "loss": 0.1329, | |
| "step": 28600 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003027901597988113, | |
| "loss": 0.119, | |
| "step": 28800 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003019177542252154, | |
| "loss": 0.1151, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003010365942515697, | |
| "loss": 0.1275, | |
| "step": 29200 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00030015108586048465, | |
| "loss": 0.129, | |
| "step": 29400 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00029926126225841235, | |
| "loss": 0.1256, | |
| "step": 29600 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0002983671568136248, | |
| "loss": 0.1373, | |
| "step": 29800 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0002974688030549625, | |
| "loss": 0.1343, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0002965662346705772, | |
| "loss": 0.1312, | |
| "step": 30200 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0002956594855066686, | |
| "loss": 0.1269, | |
| "step": 30400 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.000294748589566215, | |
| "loss": 0.1261, | |
| "step": 30600 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0002938335810076985, | |
| "loss": 0.128, | |
| "step": 30800 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00029291449414382365, | |
| "loss": 0.1204, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00029199136344023105, | |
| "loss": 0.1347, | |
| "step": 31200 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0002910642235142046, | |
| "loss": 0.1247, | |
| "step": 31400 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0002901331091333739, | |
| "loss": 0.1315, | |
| "step": 31600 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00028919805521440976, | |
| "loss": 0.1419, | |
| "step": 31800 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0002882638012678724, | |
| "loss": 0.1324, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00028732099287075494, | |
| "loss": 0.1295, | |
| "step": 32200 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00028637435038950646, | |
| "loss": 0.1195, | |
| "step": 32400 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0002854239093230996, | |
| "loss": 0.1232, | |
| "step": 32600 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00028446970531295363, | |
| "loss": 0.1208, | |
| "step": 32800 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0002835117741415974, | |
| "loss": 0.1496, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00028255015173132836, | |
| "loss": 0.1539, | |
| "step": 33200 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00028158487414286446, | |
| "loss": 0.1211, | |
| "step": 33400 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0002806159775739926, | |
| "loss": 0.1334, | |
| "step": 33600 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.000279643498358211, | |
| "loss": 0.1264, | |
| "step": 33800 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00027866747296336657, | |
| "loss": 0.1242, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0002776879379902876, | |
| "loss": 0.1311, | |
| "step": 34200 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000276704930171411, | |
| "loss": 0.1147, | |
| "step": 34400 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00027571848636940503, | |
| "loss": 0.1208, | |
| "step": 34600 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00027473360118339754, | |
| "loss": 0.1295, | |
| "step": 34800 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00027374041323396676, | |
| "loss": 0.1369, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00027274390047041496, | |
| "loss": 0.1372, | |
| "step": 35200 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00027174410026184416, | |
| "loss": 0.1289, | |
| "step": 35400 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0002707410501006349, | |
| "loss": 0.1395, | |
| "step": 35600 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00026973478760104085, | |
| "loss": 0.1263, | |
| "step": 35800 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0002687253504977778, | |
| "loss": 0.1176, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00026771277664460884, | |
| "loss": 0.127, | |
| "step": 36200 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00026669710401292467, | |
| "loss": 0.115, | |
| "step": 36400 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00026567837069031995, | |
| "loss": 0.1284, | |
| "step": 36600 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0002646566148791647, | |
| "loss": 0.1221, | |
| "step": 36800 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00026363700595470157, | |
| "loss": 0.1231, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00026260933485844465, | |
| "loss": 0.1225, | |
| "step": 37200 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00026157875636209074, | |
| "loss": 0.1123, | |
| "step": 37400 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00026054530911220257, | |
| "loss": 0.1161, | |
| "step": 37600 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00025950903186292056, | |
| "loss": 0.1133, | |
| "step": 37800 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002584699634745098, | |
| "loss": 0.1193, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002574281429119029, | |
| "loss": 0.1345, | |
| "step": 38200 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002563836092432384, | |
| "loss": 0.1171, | |
| "step": 38400 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00025533640163839615, | |
| "loss": 0.1251, | |
| "step": 38600 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00025428655936752815, | |
| "loss": 0.1217, | |
| "step": 38800 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00025323412179958617, | |
| "loss": 0.1208, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002521791284008452, | |
| "loss": 0.121, | |
| "step": 39200 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002511269124754187, | |
| "loss": 0.1364, | |
| "step": 39400 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00025006693848005583, | |
| "loss": 0.1225, | |
| "step": 39600 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00024900452742286573, | |
| "loss": 0.1217, | |
| "step": 39800 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002479397191441283, | |
| "loss": 0.116, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002468725535740183, | |
| "loss": 0.1072, | |
| "step": 40200 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002458030707311092, | |
| "loss": 0.1192, | |
| "step": 40400 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00024473131072087144, | |
| "loss": 0.1223, | |
| "step": 40600 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00024365731373416938, | |
| "loss": 0.1101, | |
| "step": 40800 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00024258112004575335, | |
| "loss": 0.1199, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00024150277001274987, | |
| "loss": 0.1189, | |
| "step": 41200 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00024042771159917315, | |
| "loss": 0.131, | |
| "step": 41400 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00023934518054633867, | |
| "loss": 0.13, | |
| "step": 41600 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002382606144962366, | |
| "loss": 0.1309, | |
| "step": 41800 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.000237174054119956, | |
| "loss": 0.1413, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00023608554016337285, | |
| "loss": 0.1271, | |
| "step": 42200 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00023499511344562222, | |
| "loss": 0.1061, | |
| "step": 42400 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00023390281485756732, | |
| "loss": 0.1125, | |
| "step": 42600 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00023280868536026635, | |
| "loss": 0.1142, | |
| "step": 42800 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00023171276598343645, | |
| "loss": 0.1261, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00023061509782391492, | |
| "loss": 0.1276, | |
| "step": 43200 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00022951572204411814, | |
| "loss": 0.1173, | |
| "step": 43400 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00022841467987049814, | |
| "loss": 0.1212, | |
| "step": 43600 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00022731201259199648, | |
| "loss": 0.1319, | |
| "step": 43800 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00022620776155849595, | |
| "loss": 0.1208, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00022510196817926987, | |
| "loss": 0.113, | |
| "step": 44200 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00022399467392142941, | |
| "loss": 0.116, | |
| "step": 44400 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00022288592030836858, | |
| "loss": 0.1267, | |
| "step": 44600 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00022177574891820678, | |
| "loss": 0.1154, | |
| "step": 44800 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00022066420138223012, | |
| "loss": 0.1262, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00021955131938332988, | |
| "loss": 0.1135, | |
| "step": 45200 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00021843714465443953, | |
| "loss": 0.1089, | |
| "step": 45400 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002173272991479189, | |
| "loss": 0.1428, | |
| "step": 45600 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00021621067029166857, | |
| "loss": 0.1275, | |
| "step": 45800 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002150928739793444, | |
| "loss": 0.1114, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000213973952128166, | |
| "loss": 0.1203, | |
| "step": 46200 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00021285394669756053, | |
| "loss": 0.1132, | |
| "step": 46400 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00021173289968758944, | |
| "loss": 0.1308, | |
| "step": 46600 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00021061085313737289, | |
| "loss": 0.1151, | |
| "step": 46800 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00020948784912351397, | |
| "loss": 0.1222, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002083639297585204, | |
| "loss": 0.1159, | |
| "step": 47200 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002072391371892256, | |
| "loss": 0.1161, | |
| "step": 47400 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002061135135952078, | |
| "loss": 0.1266, | |
| "step": 47600 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00020498710118720882, | |
| "loss": 0.1161, | |
| "step": 47800 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00020385994220555061, | |
| "loss": 0.1103, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00020273207891855188, | |
| "loss": 0.1132, | |
| "step": 48200 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002016035536209424, | |
| "loss": 0.1168, | |
| "step": 48400 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002004744086322774, | |
| "loss": 0.1199, | |
| "step": 48600 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0001993446862953505, | |
| "loss": 0.1193, | |
| "step": 48800 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0001982144289746056, | |
| "loss": 0.1066, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00019708367905454856, | |
| "loss": 0.1196, | |
| "step": 49200 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00019595247893815766, | |
| "loss": 0.1031, | |
| "step": 49400 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00019482087104529363, | |
| "loss": 0.1157, | |
| "step": 49600 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00019369455851585274, | |
| "loss": 0.1077, | |
| "step": 49800 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00019256792610462733, | |
| "loss": 0.1228, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0001914353521405628, | |
| "loss": 0.1075, | |
| "step": 50200 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0001903025397917058, | |
| "loss": 0.1213, | |
| "step": 50400 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0001891695315383753, | |
| "loss": 0.1137, | |
| "step": 50600 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00018803636986823677, | |
| "loss": 0.1242, | |
| "step": 50800 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00018690309727470875, | |
| "loss": 0.1127, | |
| "step": 51000 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0001857697562553696, | |
| "loss": 0.1246, | |
| "step": 51200 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0001846363893103633, | |
| "loss": 0.129, | |
| "step": 51400 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00018350303894080632, | |
| "loss": 0.1255, | |
| "step": 51600 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00018236974764719338, | |
| "loss": 0.123, | |
| "step": 51800 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00018123655792780396, | |
| "loss": 0.1103, | |
| "step": 52000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00018010351227710852, | |
| "loss": 0.1122, | |
| "step": 52200 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00017897065318417488, | |
| "loss": 0.1195, | |
| "step": 52400 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00017784368564134091, | |
| "loss": 0.1289, | |
| "step": 52600 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0001767113256383547, | |
| "loss": 0.1103, | |
| "step": 52800 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00017557927939969802, | |
| "loss": 0.1293, | |
| "step": 53000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0001744475893769609, | |
| "loss": 0.126, | |
| "step": 53200 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00017331629800837548, | |
| "loss": 0.1211, | |
| "step": 53400 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00017218544771722413, | |
| "loss": 0.1117, | |
| "step": 53600 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00017105508091024905, | |
| "loss": 0.1241, | |
| "step": 53800 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00016992523997606183, | |
| "loss": 0.1213, | |
| "step": 54000 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00016879596728355386, | |
| "loss": 0.1164, | |
| "step": 54200 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00016766730518030752, | |
| "loss": 0.1271, | |
| "step": 54400 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00016653929599100834, | |
| "loss": 0.1054, | |
| "step": 54600 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00016541198201585735, | |
| "loss": 0.1186, | |
| "step": 54800 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001642854055289854, | |
| "loss": 0.0995, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001631596087768676, | |
| "loss": 0.1173, | |
| "step": 55200 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00016203463397673897, | |
| "loss": 0.1103, | |
| "step": 55400 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00016091052331501155, | |
| "loss": 0.1248, | |
| "step": 55600 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001597873189456922, | |
| "loss": 0.1226, | |
| "step": 55800 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.000158665062988802, | |
| "loss": 0.1183, | |
| "step": 56000 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00015754940132266577, | |
| "loss": 0.1307, | |
| "step": 56200 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00015642916313960735, | |
| "loss": 0.1107, | |
| "step": 56400 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00015530999929939393, | |
| "loss": 0.1202, | |
| "step": 56600 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00015419195177052725, | |
| "loss": 0.113, | |
| "step": 56800 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00015307506247964756, | |
| "loss": 0.1186, | |
| "step": 57000 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00015195937330996122, | |
| "loss": 0.1157, | |
| "step": 57200 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00015084492609967028, | |
| "loss": 0.1107, | |
| "step": 57400 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001497317626404036, | |
| "loss": 0.1058, | |
| "step": 57600 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00014861992467564924, | |
| "loss": 0.1106, | |
| "step": 57800 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00014750945389918954, | |
| "loss": 0.1239, | |
| "step": 58000 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00014640039195353761, | |
| "loss": 0.1163, | |
| "step": 58200 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00014529831480934995, | |
| "loss": 0.1225, | |
| "step": 58400 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00014419218767696256, | |
| "loss": 0.1178, | |
| "step": 58600 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00014308759377244454, | |
| "loss": 0.1216, | |
| "step": 58800 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00014198457451792686, | |
| "loss": 0.1145, | |
| "step": 59000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001408831712764913, | |
| "loss": 0.1402, | |
| "step": 59200 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00013978342535061943, | |
| "loss": 0.1088, | |
| "step": 59400 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00013868537798064366, | |
| "loss": 0.112, | |
| "step": 59600 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00013758907034320075, | |
| "loss": 0.1192, | |
| "step": 59800 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00013649454354968795, | |
| "loss": 0.1125, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00013540183864472056, | |
| "loss": 0.1166, | |
| "step": 60200 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001343109966045938, | |
| "loss": 0.127, | |
| "step": 60400 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00013322749822387983, | |
| "loss": 0.1224, | |
| "step": 60600 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00013214049473686784, | |
| "loss": 0.117, | |
| "step": 60800 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00013105547641467842, | |
| "loss": 0.1259, | |
| "step": 61000 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00012997248394536063, | |
| "loss": 0.1089, | |
| "step": 61200 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0001288915579409946, | |
| "loss": 0.1065, | |
| "step": 61400 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00012781273893616794, | |
| "loss": 0.1287, | |
| "step": 61600 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00012673606738645624, | |
| "loss": 0.1155, | |
| "step": 61800 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00012566158366690577, | |
| "loss": 0.1243, | |
| "step": 62000 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00012458932807051936, | |
| "loss": 0.1014, | |
| "step": 62200 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00012351934080674567, | |
| "loss": 0.1199, | |
| "step": 62400 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00012245699458545163, | |
| "loss": 0.1227, | |
| "step": 62600 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00012139165243157235, | |
| "loss": 0.1099, | |
| "step": 62800 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00012032869852273575, | |
| "loss": 0.1231, | |
| "step": 63000 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00011927346924285653, | |
| "loss": 0.137, | |
| "step": 63200 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00011821539887681032, | |
| "loss": 0.1084, | |
| "step": 63400 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00011715983586491115, | |
| "loss": 0.1229, | |
| "step": 63600 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00011610681979063796, | |
| "loss": 0.1115, | |
| "step": 63800 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00011505639014195968, | |
| "loss": 0.1147, | |
| "step": 64000 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00011400858630985452, | |
| "loss": 0.1076, | |
| "step": 64200 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00011296344758683303, | |
| "loss": 0.111, | |
| "step": 64400 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001119210131654645, | |
| "loss": 0.1173, | |
| "step": 64600 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00011088132213690715, | |
| "loss": 0.119, | |
| "step": 64800 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00010984441348944241, | |
| "loss": 0.1142, | |
| "step": 65000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00010881032610701282, | |
| "loss": 0.1105, | |
| "step": 65200 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00010777909876776373, | |
| "loss": 0.1116, | |
| "step": 65400 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00010675077014258922, | |
| "loss": 0.1114, | |
| "step": 65600 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.000105725378793682, | |
| "loss": 0.1256, | |
| "step": 65800 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00010470296317308734, | |
| "loss": 0.1143, | |
| "step": 66000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00010368356162126112, | |
| "loss": 0.1209, | |
| "step": 66200 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00010266721236563191, | |
| "loss": 0.1147, | |
| "step": 66400 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00010165395351916748, | |
| "loss": 0.1155, | |
| "step": 66600 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00010064382307894596, | |
| "loss": 0.1134, | |
| "step": 66800 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 9.963685892473035e-05, | |
| "loss": 0.1036, | |
| "step": 67000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 9.863309881754838e-05, | |
| "loss": 0.11, | |
| "step": 67200 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 9.763258039827656e-05, | |
| "loss": 0.115, | |
| "step": 67400 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 9.663534118622836e-05, | |
| "loss": 0.109, | |
| "step": 67600 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 9.564141857774755e-05, | |
| "loss": 0.1132, | |
| "step": 67800 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 9.465579428420546e-05, | |
| "loss": 0.1048, | |
| "step": 68000 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 9.366859952570371e-05, | |
| "loss": 0.1065, | |
| "step": 68200 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.268974286984645e-05, | |
| "loss": 0.1297, | |
| "step": 68400 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.170942329919546e-05, | |
| "loss": 0.1168, | |
| "step": 68600 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.073260505340629e-05, | |
| "loss": 0.1157, | |
| "step": 68800 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 8.97593247630392e-05, | |
| "loss": 0.1067, | |
| "step": 69000 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 8.87896189259812e-05, | |
| "loss": 0.1138, | |
| "step": 69200 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 8.7823523906078e-05, | |
| "loss": 0.1163, | |
| "step": 69400 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 8.686107593176991e-05, | |
| "loss": 0.1174, | |
| "step": 69600 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 8.590231109473369e-05, | |
| "loss": 0.1238, | |
| "step": 69800 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 8.494726534852897e-05, | |
| "loss": 0.0981, | |
| "step": 70000 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 8.39959745072497e-05, | |
| "loss": 0.1114, | |
| "step": 70200 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 8.304847424418146e-05, | |
| "loss": 0.0995, | |
| "step": 70400 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 8.210480009046381e-05, | |
| "loss": 0.1124, | |
| "step": 70600 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 8.11649874337575e-05, | |
| "loss": 0.1141, | |
| "step": 70800 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 8.022907151691759e-05, | |
| "loss": 0.1046, | |
| "step": 71000 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 7.929708743667212e-05, | |
| "loss": 0.119, | |
| "step": 71200 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 7.836907014230553e-05, | |
| "loss": 0.1083, | |
| "step": 71400 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 7.744505443434856e-05, | |
| "loss": 0.1217, | |
| "step": 71600 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 7.652507496327256e-05, | |
| "loss": 0.1101, | |
| "step": 71800 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 7.560916622819085e-05, | |
| "loss": 0.1084, | |
| "step": 72000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 7.469736257556466e-05, | |
| "loss": 0.1085, | |
| "step": 72200 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.379422616698777e-05, | |
| "loss": 0.1149, | |
| "step": 72400 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.289071415064249e-05, | |
| "loss": 0.1056, | |
| "step": 72600 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.199140915836003e-05, | |
| "loss": 0.1144, | |
| "step": 72800 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.109634491396327e-05, | |
| "loss": 0.1172, | |
| "step": 73000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.02055549822475e-05, | |
| "loss": 0.1094, | |
| "step": 73200 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 6.931907276772206e-05, | |
| "loss": 0.1151, | |
| "step": 73400 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 6.843693151335722e-05, | |
| "loss": 0.1298, | |
| "step": 73600 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 6.755916429933786e-05, | |
| "loss": 0.1138, | |
| "step": 73800 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 6.668580404182292e-05, | |
| "loss": 0.1175, | |
| "step": 74000 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 6.581688349171117e-05, | |
| "loss": 0.1159, | |
| "step": 74200 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 6.495243523341272e-05, | |
| "loss": 0.1143, | |
| "step": 74400 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 6.409249168362724e-05, | |
| "loss": 0.1124, | |
| "step": 74600 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 6.32370850901287e-05, | |
| "loss": 0.1094, | |
| "step": 74800 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 6.238624753055555e-05, | |
| "loss": 0.1139, | |
| "step": 75000 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 6.154001091120815e-05, | |
| "loss": 0.1089, | |
| "step": 75200 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 6.069840696585203e-05, | |
| "loss": 0.1044, | |
| "step": 75400 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 5.986146725452824e-05, | |
| "loss": 0.1005, | |
| "step": 75600 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 5.902922316236964e-05, | |
| "loss": 0.1268, | |
| "step": 75800 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 5.820583167535201e-05, | |
| "loss": 0.1176, | |
| "step": 76000 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 5.738304840521423e-05, | |
| "loss": 0.1126, | |
| "step": 76200 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 5.65650536946333e-05, | |
| "loss": 0.1096, | |
| "step": 76400 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 5.575187821830701e-05, | |
| "loss": 0.1212, | |
| "step": 76600 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 5.4943552470212244e-05, | |
| "loss": 0.1083, | |
| "step": 76800 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 5.4140106762461915e-05, | |
| "loss": 0.0968, | |
| "step": 77000 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 5.3341571224167844e-05, | |
| "loss": 0.1115, | |
| "step": 77200 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 5.25479758003112e-05, | |
| "loss": 0.1189, | |
| "step": 77400 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.1759350250619554e-05, | |
| "loss": 0.0987, | |
| "step": 77600 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.097572414845076e-05, | |
| "loss": 0.1092, | |
| "step": 77800 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.0197126879683965e-05, | |
| "loss": 0.1163, | |
| "step": 78000 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 4.9423587641617866e-05, | |
| "loss": 0.1061, | |
| "step": 78200 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 4.865513544187548e-05, | |
| "loss": 0.1132, | |
| "step": 78400 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.78917990973167e-05, | |
| "loss": 0.1154, | |
| "step": 78600 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.7133607232957376e-05, | |
| "loss": 0.1153, | |
| "step": 78800 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.638058828089595e-05, | |
| "loss": 0.1169, | |
| "step": 79000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.563277047924755e-05, | |
| "loss": 0.1067, | |
| "step": 79200 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.489758178105981e-05, | |
| "loss": 0.1128, | |
| "step": 79400 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.416019750575994e-05, | |
| "loss": 0.1068, | |
| "step": 79600 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.342809764524491e-05, | |
| "loss": 0.1108, | |
| "step": 79800 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.270130965316634e-05, | |
| "loss": 0.1095, | |
| "step": 80000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.197986078398151e-05, | |
| "loss": 0.1002, | |
| "step": 80200 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.1263778091931396e-05, | |
| "loss": 0.1071, | |
| "step": 80400 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.055308843002584e-05, | |
| "loss": 0.1112, | |
| "step": 80600 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.98478184490369e-05, | |
| "loss": 0.1202, | |
| "step": 80800 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.91479945964992e-05, | |
| "loss": 0.1159, | |
| "step": 81000 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.8453643115718156e-05, | |
| "loss": 0.1118, | |
| "step": 81200 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.776479004478614e-05, | |
| "loss": 0.1211, | |
| "step": 81400 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.708146121560571e-05, | |
| "loss": 0.1046, | |
| "step": 81600 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.640368225292103e-05, | |
| "loss": 0.099, | |
| "step": 81800 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.573147857335717e-05, | |
| "loss": 0.1109, | |
| "step": 82000 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.506487538446655e-05, | |
| "loss": 0.1074, | |
| "step": 82200 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.4403897683784094e-05, | |
| "loss": 0.0999, | |
| "step": 82400 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.374857025788953e-05, | |
| "loss": 0.1123, | |
| "step": 82600 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.309891768147796e-05, | |
| "loss": 0.1087, | |
| "step": 82800 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.245496431643853e-05, | |
| "loss": 0.1161, | |
| "step": 83000 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.18167343109405e-05, | |
| "loss": 0.096, | |
| "step": 83200 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.118739967637463e-05, | |
| "loss": 0.121, | |
| "step": 83400 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.0560659061372356e-05, | |
| "loss": 0.1107, | |
| "step": 83600 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.9939712842117768e-05, | |
| "loss": 0.1186, | |
| "step": 83800 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.9324584304015175e-05, | |
| "loss": 0.1105, | |
| "step": 84000 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.8715296514306542e-05, | |
| "loss": 0.1072, | |
| "step": 84200 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.8111872321206513e-05, | |
| "loss": 0.1169, | |
| "step": 84400 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.751433435304559e-05, | |
| "loss": 0.108, | |
| "step": 84600 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.6922705017421654e-05, | |
| "loss": 0.1022, | |
| "step": 84800 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.6337006500359684e-05, | |
| "loss": 0.1092, | |
| "step": 85000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.575726076547943e-05, | |
| "loss": 0.1066, | |
| "step": 85200 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.5183489553172356e-05, | |
| "loss": 0.1025, | |
| "step": 85400 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.4618538305192336e-05, | |
| "loss": 0.1107, | |
| "step": 85600 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.405675032298176e-05, | |
| "loss": 0.1129, | |
| "step": 85800 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.350100063226973e-05, | |
| "loss": 0.1023, | |
| "step": 86000 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.2951310073599112e-05, | |
| "loss": 0.1074, | |
| "step": 86200 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.240769926029589e-05, | |
| "loss": 0.1178, | |
| "step": 86400 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.187018857769647e-05, | |
| "loss": 0.0992, | |
| "step": 86600 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.1338798182383172e-05, | |
| "loss": 0.1141, | |
| "step": 86800 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.081354800142823e-05, | |
| "loss": 0.1064, | |
| "step": 87000 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.029445773164665e-05, | |
| "loss": 0.112, | |
| "step": 87200 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.978154683885758e-05, | |
| "loss": 0.1155, | |
| "step": 87400 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.927735266798869e-05, | |
| "loss": 0.1068, | |
| "step": 87600 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.8776826864072523e-05, | |
| "loss": 0.0999, | |
| "step": 87800 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.8282537348113476e-05, | |
| "loss": 0.1178, | |
| "step": 88000 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.7794502655905586e-05, | |
| "loss": 0.1244, | |
| "step": 88200 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.731274108868771e-05, | |
| "loss": 0.106, | |
| "step": 88400 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.6837270712457434e-05, | |
| "loss": 0.1126, | |
| "step": 88600 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.6368109357293467e-05, | |
| "loss": 0.1083, | |
| "step": 88800 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.5905274616686885e-05, | |
| "loss": 0.1053, | |
| "step": 89000 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.5448783846881664e-05, | |
| "loss": 0.1096, | |
| "step": 89200 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4998654166223707e-05, | |
| "loss": 0.1076, | |
| "step": 89400 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4554902454518763e-05, | |
| "loss": 0.094, | |
| "step": 89600 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4119716204088943e-05, | |
| "loss": 0.1132, | |
| "step": 89800 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.3688738016939574e-05, | |
| "loss": 0.11, | |
| "step": 90000 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.3264186920432302e-05, | |
| "loss": 0.0963, | |
| "step": 90200 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.2846078835179235e-05, | |
| "loss": 0.1174, | |
| "step": 90400 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.2434429440180526e-05, | |
| "loss": 0.1106, | |
| "step": 90600 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.2029254172236074e-05, | |
| "loss": 0.1202, | |
| "step": 90800 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.1630568225366933e-05, | |
| "loss": 0.1337, | |
| "step": 91000 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.1238386550245468e-05, | |
| "loss": 0.098, | |
| "step": 91200 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.0852723853634612e-05, | |
| "loss": 0.1105, | |
| "step": 91400 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.0473594597836475e-05, | |
| "loss": 0.115, | |
| "step": 91600 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.0101013000149961e-05, | |
| "loss": 0.1074, | |
| "step": 91800 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 9.73499303233757e-06, | |
| "loss": 0.1182, | |
| "step": 92000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 9.377329264566203e-06, | |
| "loss": 0.1063, | |
| "step": 92200 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 9.024440509739605e-06, | |
| "loss": 0.0997, | |
| "step": 92400 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 8.67815375611877e-06, | |
| "loss": 0.1185, | |
| "step": 92600 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 8.338481989412805e-06, | |
| "loss": 0.1013, | |
| "step": 92800 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.005437947269865e-06, | |
| "loss": 0.1094, | |
| "step": 93000 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 7.679034118799212e-06, | |
| "loss": 0.1005, | |
| "step": 93200 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 7.359282744102871e-06, | |
| "loss": 0.1112, | |
| "step": 93400 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 7.046195813816774e-06, | |
| "loss": 0.118, | |
| "step": 93600 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 6.7397850686610074e-06, | |
| "loss": 0.1206, | |
| "step": 93800 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.440061998999538e-06, | |
| "loss": 0.1076, | |
| "step": 94000 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.147037844409444e-06, | |
| "loss": 0.1129, | |
| "step": 94200 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 5.8621384557648585e-06, | |
| "loss": 0.1074, | |
| "step": 94400 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 5.5825112153106485e-06, | |
| "loss": 0.1086, | |
| "step": 94600 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 5.309615047973128e-06, | |
| "loss": 0.1187, | |
| "step": 94800 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.043460187323934e-06, | |
| "loss": 0.1103, | |
| "step": 95000 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.7840566141364325e-06, | |
| "loss": 0.0992, | |
| "step": 95200 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.531414056011703e-06, | |
| "loss": 0.1075, | |
| "step": 95400 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.285541987013553e-06, | |
| "loss": 0.1032, | |
| "step": 95600 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.04644962731328e-06, | |
| "loss": 0.1178, | |
| "step": 95800 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.8141459428440006e-06, | |
| "loss": 0.1096, | |
| "step": 96000 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.588639644964385e-06, | |
| "loss": 0.1006, | |
| "step": 96200 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.3699391901318616e-06, | |
| "loss": 0.1058, | |
| "step": 96400 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.15805277958575e-06, | |
| "loss": 0.1066, | |
| "step": 96600 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.9539966986151773e-06, | |
| "loss": 0.1004, | |
| "step": 96800 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.755727790853974e-06, | |
| "loss": 0.121, | |
| "step": 97000 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.5642959602285565e-06, | |
| "loss": 0.1159, | |
| "step": 97200 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.379708385408211e-06, | |
| "loss": 0.1043, | |
| "step": 97400 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.20197198840359e-06, | |
| "loss": 0.1057, | |
| "step": 97600 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.0310934343069233e-06, | |
| "loss": 0.1079, | |
| "step": 97800 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.8670791310423448e-06, | |
| "loss": 0.1123, | |
| "step": 98000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7099352291252954e-06, | |
| "loss": 0.109, | |
| "step": 98200 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.5596676214321622e-06, | |
| "loss": 0.1162, | |
| "step": 98400 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.416281942979123e-06, | |
| "loss": 0.0953, | |
| "step": 98600 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2804489218234843e-06, | |
| "loss": 0.1049, | |
| "step": 98800 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.1508084999821759e-06, | |
| "loss": 0.1113, | |
| "step": 99000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.0280653395460467e-06, | |
| "loss": 0.1145, | |
| "step": 99200 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.122240433680127e-07, | |
| "loss": 0.1036, | |
| "step": 99400 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.032889554819566e-07, | |
| "loss": 0.1099, | |
| "step": 99600 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.01264160939824e-07, | |
| "loss": 0.1211, | |
| "step": 99800 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.061534856584398e-07, | |
| "loss": 0.1108, | |
| "step": 100000 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 5.179604962760226e-07, | |
| "loss": 0.1043, | |
| "step": 100200 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.366885000184595e-07, | |
| "loss": 0.1033, | |
| "step": 100400 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.6234054457534074e-07, | |
| "loss": 0.1075, | |
| "step": 100600 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.9491941798546746e-07, | |
| "loss": 0.1061, | |
| "step": 100800 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.3442764853255093e-07, | |
| "loss": 0.1047, | |
| "step": 101000 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.8086750465026087e-07, | |
| "loss": 0.1107, | |
| "step": 101200 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.3424099483717806e-07, | |
| "loss": 0.1096, | |
| "step": 101400 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.454986758148648e-08, | |
| "loss": 0.121, | |
| "step": 101600 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.17956112954053e-08, | |
| "loss": 0.0994, | |
| "step": 101800 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.597945425940143e-08, | |
| "loss": 0.1083, | |
| "step": 102000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.710236457610074e-08, | |
| "loss": 0.0999, | |
| "step": 102200 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 5.1650501339184315e-09, | |
| "loss": 0.1125, | |
| "step": 102400 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.6795858065506786e-10, | |
| "loss": 0.1147, | |
| "step": 102600 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9390755356994795, | |
| "eval_auc": 0.816582265454698, | |
| "eval_f1": 0.3302349377015186, | |
| "eval_loss": 0.5970005989074707, | |
| "eval_mcc": 0.36456376398967993, | |
| "eval_precision": 0.21720888765810265, | |
| "eval_recall": 0.6885007658293798, | |
| "eval_runtime": 6449.9747, | |
| "eval_samples_per_second": 23.861, | |
| "eval_steps_per_second": 3.977, | |
| "step": 102604 | |
| } | |
| ], | |
| "logging_steps": 200, | |
| "max_steps": 102604, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "total_flos": 1.233389886781941e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |