Training in progress, epoch 2

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f11e54099794ae11456f3957bfce79a6ee075a2b5ce1509ec5603694095b00e1
+oid sha256:d80c8bfa2833452487c41e8c35d2e402dbfd01382580ad8d4932da7eec322b7f
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e60410aab2a121ee30ee376e60bc0f1ec7660af1743d11361957d7c0a8e4b6d2
+oid sha256:b0cd617648a8e45c0a6ce88477615bef576e1815e68165d69cda4679306c9f2d
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8de622084f990bc14bf437de2ea2d7f6b1c4e2bd5bed2881df2bf9de4a41f319
+oid sha256:ba169ab715eac39f8164acaae05a13f4c30149b780204ec6e6f559a6f408dd0c
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:123d3cf9f470ab88be0bc7c7bcc3588f223827f581e7e438c6278aaacccdc584
+oid sha256:d0e8dadc0d417056c4832561dcac41cf8ce999acef6bba10ff5b7d79d5802b2a
 size 1089994880

trainer_log.jsonl

@@ -346,3 +346,175 @@
 {"current_steps": 346, "total_steps": 519, "loss": 0.3812, "lr": 2.416596096747564e-05, "epoch": 1.997113997113997, "percentage": 66.67, "elapsed_time": "5:06:25", "remaining_time": "2:33:12"}
 {"current_steps": 347, "total_steps": 519, "loss": 0.6364, "lr": 2.3919214397681137e-05, "epoch": 2.0028860028860027, "percentage": 66.86, "elapsed_time": "5:08:38", "remaining_time": "2:32:59"}
 {"current_steps": 348, "total_steps": 519, "loss": 0.3528, "lr": 2.3673195560114365e-05, "epoch": 2.0086580086580086, "percentage": 67.05, "elapsed_time": "5:09:30", "remaining_time": "2:32:04"}
+{"current_steps": 349, "total_steps": 519, "loss": 0.3535, "lr": 2.3427915588300846e-05, "epoch": 2.0144300144300145, "percentage": 67.24, "elapsed_time": "5:10:21", "remaining_time": "2:31:10"}
+{"current_steps": 350, "total_steps": 519, "loss": 0.3491, "lr": 2.31833855823289e-05, "epoch": 2.0202020202020203, "percentage": 67.44, "elapsed_time": "5:11:13", "remaining_time": "2:30:16"}
+{"current_steps": 351, "total_steps": 519, "loss": 0.3554, "lr": 2.2939616608347316e-05, "epoch": 2.0259740259740258, "percentage": 67.63, "elapsed_time": "5:12:05", "remaining_time": "2:29:22"}
+{"current_steps": 352, "total_steps": 519, "loss": 0.3493, "lr": 2.2696619698064515e-05, "epoch": 2.0317460317460316, "percentage": 67.82, "elapsed_time": "5:12:57", "remaining_time": "2:28:28"}
+{"current_steps": 353, "total_steps": 519, "loss": 0.347, "lr": 2.245440584824941e-05, "epoch": 2.0375180375180375, "percentage": 68.02, "elapsed_time": "5:13:51", "remaining_time": "2:27:35"}
+{"current_steps": 354, "total_steps": 519, "loss": 0.3493, "lr": 2.2212986020233683e-05, "epoch": 2.0432900432900434, "percentage": 68.21, "elapsed_time": "5:14:54", "remaining_time": "2:26:46"}
+{"current_steps": 355, "total_steps": 519, "loss": 0.3553, "lr": 2.1972371139415705e-05, "epoch": 2.0490620490620493, "percentage": 68.4, "elapsed_time": "5:15:52", "remaining_time": "2:25:55"}
+{"current_steps": 356, "total_steps": 519, "loss": 0.3532, "lr": 2.1732572094766188e-05, "epoch": 2.0548340548340547, "percentage": 68.59, "elapsed_time": "5:16:43", "remaining_time": "2:25:01"}
+{"current_steps": 357, "total_steps": 519, "loss": 0.3553, "lr": 2.1493599738335358e-05, "epoch": 2.0606060606060606, "percentage": 68.79, "elapsed_time": "5:17:35", "remaining_time": "2:24:07"}
+{"current_steps": 358, "total_steps": 519, "loss": 0.3574, "lr": 2.1255464884761863e-05, "epoch": 2.0663780663780664, "percentage": 68.98, "elapsed_time": "5:18:27", "remaining_time": "2:23:12"}
+{"current_steps": 359, "total_steps": 519, "loss": 0.3467, "lr": 2.1018178310783322e-05, "epoch": 2.0721500721500723, "percentage": 69.17, "elapsed_time": "5:19:19", "remaining_time": "2:22:18"}
+{"current_steps": 360, "total_steps": 519, "loss": 0.3518, "lr": 2.07817507547487e-05, "epoch": 2.0779220779220777, "percentage": 69.36, "elapsed_time": "5:20:11", "remaining_time": "2:21:24"}
+{"current_steps": 361, "total_steps": 519, "loss": 0.3454, "lr": 2.054619291613226e-05, "epoch": 2.0836940836940836, "percentage": 69.56, "elapsed_time": "5:21:02", "remaining_time": "2:20:30"}
+{"current_steps": 362, "total_steps": 519, "loss": 0.3509, "lr": 2.031151545504944e-05, "epoch": 2.0894660894660895, "percentage": 69.75, "elapsed_time": "5:21:54", "remaining_time": "2:19:36"}
+{"current_steps": 363, "total_steps": 519, "loss": 0.3591, "lr": 2.007772899177438e-05, "epoch": 2.0952380952380953, "percentage": 69.94, "elapsed_time": "5:22:46", "remaining_time": "2:18:42"}
+{"current_steps": 364, "total_steps": 519, "loss": 0.3462, "lr": 1.984484410625935e-05, "epoch": 2.101010101010101, "percentage": 70.13, "elapsed_time": "5:23:38", "remaining_time": "2:17:48"}
+{"current_steps": 365, "total_steps": 519, "loss": 0.3474, "lr": 1.9612871337655882e-05, "epoch": 2.1067821067821066, "percentage": 70.33, "elapsed_time": "5:24:29", "remaining_time": "2:16:54"}
+{"current_steps": 366, "total_steps": 519, "loss": 0.3532, "lr": 1.938182118383787e-05, "epoch": 2.1125541125541125, "percentage": 70.52, "elapsed_time": "5:25:21", "remaining_time": "2:16:00"}
+{"current_steps": 367, "total_steps": 519, "loss": 0.348, "lr": 1.9151704100926536e-05, "epoch": 2.1183261183261184, "percentage": 70.71, "elapsed_time": "5:26:13", "remaining_time": "2:15:06"}
+{"current_steps": 368, "total_steps": 519, "loss": 0.349, "lr": 1.892253050281715e-05, "epoch": 2.1240981240981243, "percentage": 70.91, "elapsed_time": "5:27:07", "remaining_time": "2:14:13"}
+{"current_steps": 369, "total_steps": 519, "loss": 0.346, "lr": 1.8694310760707824e-05, "epoch": 2.1298701298701297, "percentage": 71.1, "elapsed_time": "5:28:14", "remaining_time": "2:13:26"}
+{"current_steps": 370, "total_steps": 519, "loss": 0.3472, "lr": 1.8467055202630137e-05, "epoch": 2.1356421356421356, "percentage": 71.29, "elapsed_time": "5:29:10", "remaining_time": "2:12:33"}
+{"current_steps": 371, "total_steps": 519, "loss": 0.3482, "lr": 1.8240774112981702e-05, "epoch": 2.1414141414141414, "percentage": 71.48, "elapsed_time": "5:30:02", "remaining_time": "2:11:39"}
+{"current_steps": 372, "total_steps": 519, "loss": 0.3461, "lr": 1.8015477732060853e-05, "epoch": 2.1471861471861473, "percentage": 71.68, "elapsed_time": "5:30:53", "remaining_time": "2:10:45"}
+{"current_steps": 373, "total_steps": 519, "loss": 0.3534, "lr": 1.779117625560309e-05, "epoch": 2.1529581529581527, "percentage": 71.87, "elapsed_time": "5:31:45", "remaining_time": "2:09:51"}
+{"current_steps": 374, "total_steps": 519, "loss": 0.3541, "lr": 1.7567879834319793e-05, "epoch": 2.1587301587301586, "percentage": 72.06, "elapsed_time": "5:32:37", "remaining_time": "2:08:57"}
+{"current_steps": 375, "total_steps": 519, "loss": 0.3512, "lr": 1.734559857343878e-05, "epoch": 2.1645021645021645, "percentage": 72.25, "elapsed_time": "5:33:29", "remaining_time": "2:08:03"}
+{"current_steps": 376, "total_steps": 519, "loss": 0.3453, "lr": 1.7124342532247057e-05, "epoch": 2.1702741702741704, "percentage": 72.45, "elapsed_time": "5:34:21", "remaining_time": "2:07:09"}
+{"current_steps": 377, "total_steps": 519, "loss": 0.3513, "lr": 1.6904121723635477e-05, "epoch": 2.1760461760461762, "percentage": 72.64, "elapsed_time": "5:35:13", "remaining_time": "2:06:15"}
+{"current_steps": 378, "total_steps": 519, "loss": 0.3406, "lr": 1.6684946113645746e-05, "epoch": 2.1818181818181817, "percentage": 72.83, "elapsed_time": "5:36:04", "remaining_time": "2:05:21"}
+{"current_steps": 379, "total_steps": 519, "loss": 0.3469, "lr": 1.6466825621019337e-05, "epoch": 2.1875901875901875, "percentage": 73.03, "elapsed_time": "5:36:56", "remaining_time": "2:04:27"}
+{"current_steps": 380, "total_steps": 519, "loss": 0.3485, "lr": 1.6249770116748604e-05, "epoch": 2.1933621933621934, "percentage": 73.22, "elapsed_time": "5:37:48", "remaining_time": "2:03:34"}
+{"current_steps": 381, "total_steps": 519, "loss": 0.3499, "lr": 1.6033789423630134e-05, "epoch": 2.1991341991341993, "percentage": 73.41, "elapsed_time": "5:38:40", "remaining_time": "2:02:40"}
+{"current_steps": 382, "total_steps": 519, "loss": 0.3467, "lr": 1.5818893315820144e-05, "epoch": 2.2049062049062047, "percentage": 73.6, "elapsed_time": "5:39:32", "remaining_time": "2:01:46"}
+{"current_steps": 383, "total_steps": 519, "loss": 0.3431, "lr": 1.5605091518392234e-05, "epoch": 2.2106782106782106, "percentage": 73.8, "elapsed_time": "5:40:29", "remaining_time": "2:00:54"}
+{"current_steps": 384, "total_steps": 519, "loss": 0.3483, "lr": 1.5392393706897224e-05, "epoch": 2.2164502164502164, "percentage": 73.99, "elapsed_time": "5:41:34", "remaining_time": "2:00:05"}
+{"current_steps": 385, "total_steps": 519, "loss": 0.3532, "lr": 1.5180809506925306e-05, "epoch": 2.2222222222222223, "percentage": 74.18, "elapsed_time": "5:42:29", "remaining_time": "1:59:12"}
+{"current_steps": 386, "total_steps": 519, "loss": 0.3483, "lr": 1.4970348493670454e-05, "epoch": 2.227994227994228, "percentage": 74.37, "elapsed_time": "5:43:21", "remaining_time": "1:58:18"}
+{"current_steps": 387, "total_steps": 519, "loss": 0.3516, "lr": 1.476102019149707e-05, "epoch": 2.2337662337662336, "percentage": 74.57, "elapsed_time": "5:44:13", "remaining_time": "1:57:24"}
+{"current_steps": 388, "total_steps": 519, "loss": 0.3475, "lr": 1.4552834073508958e-05, "epoch": 2.2395382395382395, "percentage": 74.76, "elapsed_time": "5:45:04", "remaining_time": "1:56:30"}
+{"current_steps": 389, "total_steps": 519, "loss": 0.3453, "lr": 1.434579956112069e-05, "epoch": 2.2453102453102454, "percentage": 74.95, "elapsed_time": "5:45:56", "remaining_time": "1:55:36"}
+{"current_steps": 390, "total_steps": 519, "loss": 0.3405, "lr": 1.4139926023631168e-05, "epoch": 2.2510822510822512, "percentage": 75.14, "elapsed_time": "5:46:48", "remaining_time": "1:54:42"}
+{"current_steps": 391, "total_steps": 519, "loss": 0.3467, "lr": 1.3935222777799652e-05, "epoch": 2.2568542568542567, "percentage": 75.34, "elapsed_time": "5:47:40", "remaining_time": "1:53:49"}
+{"current_steps": 392, "total_steps": 519, "loss": 0.3493, "lr": 1.3731699087424115e-05, "epoch": 2.2626262626262625, "percentage": 75.53, "elapsed_time": "5:48:32", "remaining_time": "1:52:55"}
+{"current_steps": 393, "total_steps": 519, "loss": 0.3441, "lr": 1.3529364162922004e-05, "epoch": 2.2683982683982684, "percentage": 75.72, "elapsed_time": "5:49:24", "remaining_time": "1:52:01"}
+{"current_steps": 394, "total_steps": 519, "loss": 0.3432, "lr": 1.3328227160913478e-05, "epoch": 2.2741702741702743, "percentage": 75.92, "elapsed_time": "5:50:15", "remaining_time": "1:51:07"}
+{"current_steps": 395, "total_steps": 519, "loss": 0.3454, "lr": 1.3128297183806976e-05, "epoch": 2.27994227994228, "percentage": 76.11, "elapsed_time": "5:51:07", "remaining_time": "1:50:13"}
+{"current_steps": 396, "total_steps": 519, "loss": 0.3437, "lr": 1.2929583279387296e-05, "epoch": 2.2857142857142856, "percentage": 76.3, "elapsed_time": "5:51:59", "remaining_time": "1:49:19"}
+{"current_steps": 397, "total_steps": 519, "loss": 0.352, "lr": 1.2732094440406177e-05, "epoch": 2.2914862914862915, "percentage": 76.49, "elapsed_time": "5:52:51", "remaining_time": "1:48:26"}
+{"current_steps": 398, "total_steps": 519, "loss": 0.3486, "lr": 1.2535839604175255e-05, "epoch": 2.2972582972582973, "percentage": 76.69, "elapsed_time": "5:53:51", "remaining_time": "1:47:34"}
+{"current_steps": 399, "total_steps": 519, "loss": 0.3452, "lr": 1.2340827652161713e-05, "epoch": 2.303030303030303, "percentage": 76.88, "elapsed_time": "5:54:55", "remaining_time": "1:46:44"}
+{"current_steps": 400, "total_steps": 519, "loss": 0.348, "lr": 1.2147067409586243e-05, "epoch": 2.3088023088023086, "percentage": 77.07, "elapsed_time": "5:55:48", "remaining_time": "1:45:51"}
+{"current_steps": 401, "total_steps": 519, "loss": 0.3535, "lr": 1.1954567645023759e-05, "epoch": 2.3145743145743145, "percentage": 77.26, "elapsed_time": "5:56:40", "remaining_time": "1:44:57"}
+{"current_steps": 402, "total_steps": 519, "loss": 0.3495, "lr": 1.1763337070006515e-05, "epoch": 2.3203463203463204, "percentage": 77.46, "elapsed_time": "5:57:32", "remaining_time": "1:44:03"}
+{"current_steps": 403, "total_steps": 519, "loss": 0.3518, "lr": 1.1573384338629908e-05, "epoch": 2.3261183261183263, "percentage": 77.65, "elapsed_time": "5:58:24", "remaining_time": "1:43:09"}
+{"current_steps": 404, "total_steps": 519, "loss": 0.349, "lr": 1.138471804716076e-05, "epoch": 2.3318903318903317, "percentage": 77.84, "elapsed_time": "5:59:16", "remaining_time": "1:42:16"}
+{"current_steps": 405, "total_steps": 519, "loss": 0.3472, "lr": 1.1197346733648424e-05, "epoch": 2.3376623376623376, "percentage": 78.03, "elapsed_time": "6:00:08", "remaining_time": "1:41:22"}
+{"current_steps": 406, "total_steps": 519, "loss": 0.3517, "lr": 1.1011278877538292e-05, "epoch": 2.3434343434343434, "percentage": 78.23, "elapsed_time": "6:00:59", "remaining_time": "1:40:28"}
+{"current_steps": 407, "total_steps": 519, "loss": 0.3447, "lr": 1.0826522899288077e-05, "epoch": 2.3492063492063493, "percentage": 78.42, "elapsed_time": "6:01:51", "remaining_time": "1:39:34"}
+{"current_steps": 408, "total_steps": 519, "loss": 0.3497, "lr": 1.0643087159986796e-05, "epoch": 2.354978354978355, "percentage": 78.61, "elapsed_time": "6:02:43", "remaining_time": "1:38:40"}
+{"current_steps": 409, "total_steps": 519, "loss": 0.3399, "lr": 1.046097996097633e-05, "epoch": 2.3607503607503606, "percentage": 78.81, "elapsed_time": "6:03:35", "remaining_time": "1:37:47"}
+{"current_steps": 410, "total_steps": 519, "loss": 0.3469, "lr": 1.0280209543475799e-05, "epoch": 2.3665223665223665, "percentage": 79.0, "elapsed_time": "6:04:26", "remaining_time": "1:36:53"}
+{"current_steps": 411, "total_steps": 519, "loss": 0.3498, "lr": 1.0100784088208569e-05, "epoch": 2.3722943722943723, "percentage": 79.19, "elapsed_time": "6:05:18", "remaining_time": "1:35:59"}
+{"current_steps": 412, "total_steps": 519, "loss": 0.3479, "lr": 9.922711715032069e-06, "epoch": 2.378066378066378, "percentage": 79.38, "elapsed_time": "6:06:11", "remaining_time": "1:35:06"}
+{"current_steps": 413, "total_steps": 519, "loss": 0.3507, "lr": 9.746000482570306e-06, "epoch": 2.3838383838383836, "percentage": 79.58, "elapsed_time": "6:07:14", "remaining_time": "1:34:15"}
+{"current_steps": 414, "total_steps": 519, "loss": 0.3396, "lr": 9.570658387849163e-06, "epoch": 2.3896103896103895, "percentage": 79.77, "elapsed_time": "6:08:16", "remaining_time": "1:33:24"}
+{"current_steps": 415, "total_steps": 519, "loss": 0.3409, "lr": 9.396693365934517e-06, "epoch": 2.3953823953823954, "percentage": 79.96, "elapsed_time": "6:09:09", "remaining_time": "1:32:30"}
+{"current_steps": 416, "total_steps": 519, "loss": 0.3476, "lr": 9.22411328957315e-06, "epoch": 2.4011544011544013, "percentage": 80.15, "elapsed_time": "6:10:01", "remaining_time": "1:31:36"}
+{"current_steps": 417, "total_steps": 519, "loss": 0.343, "lr": 9.052925968836442e-06, "epoch": 2.4069264069264067, "percentage": 80.35, "elapsed_time": "6:10:52", "remaining_time": "1:30:43"}
+{"current_steps": 418, "total_steps": 519, "loss": 0.3457, "lr": 8.883139150766941e-06, "epoch": 2.4126984126984126, "percentage": 80.54, "elapsed_time": "6:11:44", "remaining_time": "1:29:49"}
+{"current_steps": 419, "total_steps": 519, "loss": 0.3491, "lr": 8.714760519027789e-06, "epoch": 2.4184704184704184, "percentage": 80.73, "elapsed_time": "6:12:36", "remaining_time": "1:28:55"}
+{"current_steps": 420, "total_steps": 519, "loss": 0.3486, "lr": 8.547797693554946e-06, "epoch": 2.4242424242424243, "percentage": 80.92, "elapsed_time": "6:13:28", "remaining_time": "1:28:02"}
+{"current_steps": 421, "total_steps": 519, "loss": 0.3456, "lr": 8.3822582302124e-06, "epoch": 2.43001443001443, "percentage": 81.12, "elapsed_time": "6:14:20", "remaining_time": "1:27:08"}
+{"current_steps": 422, "total_steps": 519, "loss": 0.3411, "lr": 8.218149620450235e-06, "epoch": 2.4357864357864356, "percentage": 81.31, "elapsed_time": "6:15:12", "remaining_time": "1:26:14"}
+{"current_steps": 423, "total_steps": 519, "loss": 0.3414, "lr": 8.055479290965569e-06, "epoch": 2.4415584415584415, "percentage": 81.5, "elapsed_time": "6:16:03", "remaining_time": "1:25:20"}
+{"current_steps": 424, "total_steps": 519, "loss": 0.3382, "lr": 7.894254603366497e-06, "epoch": 2.4473304473304474, "percentage": 81.7, "elapsed_time": "6:16:55", "remaining_time": "1:24:27"}
+{"current_steps": 425, "total_steps": 519, "loss": 0.35, "lr": 7.734482853838892e-06, "epoch": 2.4531024531024532, "percentage": 81.89, "elapsed_time": "6:17:47", "remaining_time": "1:23:33"}
+{"current_steps": 426, "total_steps": 519, "loss": 0.3414, "lr": 7.576171272816286e-06, "epoch": 2.458874458874459, "percentage": 82.08, "elapsed_time": "6:18:39", "remaining_time": "1:22:39"}
+{"current_steps": 427, "total_steps": 519, "loss": 0.3412, "lr": 7.4193270246526135e-06, "epoch": 2.4646464646464645, "percentage": 82.27, "elapsed_time": "6:19:32", "remaining_time": "1:21:46"}
+{"current_steps": 428, "total_steps": 519, "loss": 0.3433, "lr": 7.263957207297978e-06, "epoch": 2.4704184704184704, "percentage": 82.47, "elapsed_time": "6:20:36", "remaining_time": "1:20:55"}
+{"current_steps": 429, "total_steps": 519, "loss": 0.3356, "lr": 7.1100688519774876e-06, "epoch": 2.4761904761904763, "percentage": 82.66, "elapsed_time": "6:21:37", "remaining_time": "1:20:03"}
+{"current_steps": 430, "total_steps": 519, "loss": 0.3526, "lr": 6.957668922873031e-06, "epoch": 2.481962481962482, "percentage": 82.85, "elapsed_time": "6:22:30", "remaining_time": "1:19:10"}
+{"current_steps": 431, "total_steps": 519, "loss": 0.3488, "lr": 6.8067643168080766e-06, "epoch": 2.4877344877344876, "percentage": 83.04, "elapsed_time": "6:23:21", "remaining_time": "1:18:16"}
+{"current_steps": 432, "total_steps": 519, "loss": 0.3452, "lr": 6.6573618629356274e-06, "epoch": 2.4935064935064934, "percentage": 83.24, "elapsed_time": "6:24:13", "remaining_time": "1:17:22"}
+{"current_steps": 433, "total_steps": 519, "loss": 0.347, "lr": 6.509468322429131e-06, "epoch": 2.4992784992784993, "percentage": 83.43, "elapsed_time": "6:25:05", "remaining_time": "1:16:29"}
+{"current_steps": 434, "total_steps": 519, "loss": 0.3455, "lr": 6.36309038817652e-06, "epoch": 2.505050505050505, "percentage": 83.62, "elapsed_time": "6:25:57", "remaining_time": "1:15:35"}
+{"current_steps": 435, "total_steps": 519, "loss": 0.3439, "lr": 6.218234684477278e-06, "epoch": 2.5108225108225106, "percentage": 83.82, "elapsed_time": "6:26:49", "remaining_time": "1:14:41"}
+{"current_steps": 436, "total_steps": 519, "loss": 0.3378, "lr": 6.07490776674275e-06, "epoch": 2.5165945165945165, "percentage": 84.01, "elapsed_time": "6:27:41", "remaining_time": "1:13:48"}
+{"current_steps": 437, "total_steps": 519, "loss": 0.3452, "lr": 5.933116121199378e-06, "epoch": 2.5223665223665224, "percentage": 84.2, "elapsed_time": "6:28:32", "remaining_time": "1:12:54"}
+{"current_steps": 438, "total_steps": 519, "loss": 0.3368, "lr": 5.792866164595245e-06, "epoch": 2.5281385281385282, "percentage": 84.39, "elapsed_time": "6:29:24", "remaining_time": "1:12:00"}
+{"current_steps": 439, "total_steps": 519, "loss": 0.3396, "lr": 5.654164243909659e-06, "epoch": 2.533910533910534, "percentage": 84.59, "elapsed_time": "6:30:16", "remaining_time": "1:11:07"}
+{"current_steps": 440, "total_steps": 519, "loss": 0.3432, "lr": 5.517016636065915e-06, "epoch": 2.5396825396825395, "percentage": 84.78, "elapsed_time": "6:31:08", "remaining_time": "1:10:13"}
+{"current_steps": 441, "total_steps": 519, "loss": 0.3499, "lr": 5.381429547647239e-06, "epoch": 2.5454545454545454, "percentage": 84.97, "elapsed_time": "6:31:59", "remaining_time": "1:09:19"}
+{"current_steps": 442, "total_steps": 519, "loss": 0.3407, "lr": 5.247409114615898e-06, "epoch": 2.5512265512265513, "percentage": 85.16, "elapsed_time": "6:32:53", "remaining_time": "1:08:26"}
+{"current_steps": 443, "total_steps": 519, "loss": 0.3424, "lr": 5.114961402035552e-06, "epoch": 2.556998556998557, "percentage": 85.36, "elapsed_time": "6:33:57", "remaining_time": "1:07:35"}
+{"current_steps": 444, "total_steps": 519, "loss": 0.3413, "lr": 4.984092403796763e-06, "epoch": 2.562770562770563, "percentage": 85.55, "elapsed_time": "6:34:56", "remaining_time": "1:06:42"}
+{"current_steps": 445, "total_steps": 519, "loss": 0.3373, "lr": 4.854808042345736e-06, "epoch": 2.5685425685425685, "percentage": 85.74, "elapsed_time": "6:35:49", "remaining_time": "1:05:49"}
+{"current_steps": 446, "total_steps": 519, "loss": 0.3486, "lr": 4.7271141684163e-06, "epoch": 2.5743145743145743, "percentage": 85.93, "elapsed_time": "6:36:40", "remaining_time": "1:04:55"}
+{"current_steps": 447, "total_steps": 519, "loss": 0.3419, "lr": 4.601016560765144e-06, "epoch": 2.58008658008658, "percentage": 86.13, "elapsed_time": "6:37:32", "remaining_time": "1:04:02"}
+{"current_steps": 448, "total_steps": 519, "loss": 0.3366, "lr": 4.476520925910284e-06, "epoch": 2.5858585858585856, "percentage": 86.32, "elapsed_time": "6:38:24", "remaining_time": "1:03:08"}
+{"current_steps": 449, "total_steps": 519, "loss": 0.3425, "lr": 4.353632897872846e-06, "epoch": 2.5916305916305915, "percentage": 86.51, "elapsed_time": "6:39:16", "remaining_time": "1:02:14"}
+{"current_steps": 450, "total_steps": 519, "loss": 0.3484, "lr": 4.2323580379220755e-06, "epoch": 2.5974025974025974, "percentage": 86.71, "elapsed_time": "6:40:08", "remaining_time": "1:01:21"}
+{"current_steps": 451, "total_steps": 519, "loss": 0.3429, "lr": 4.112701834323672e-06, "epoch": 2.6031746031746033, "percentage": 86.9, "elapsed_time": "6:40:59", "remaining_time": "1:00:27"}
+{"current_steps": 452, "total_steps": 519, "loss": 0.3351, "lr": 3.994669702091418e-06, "epoch": 2.608946608946609, "percentage": 87.09, "elapsed_time": "6:41:51", "remaining_time": "0:59:34"}
+{"current_steps": 453, "total_steps": 519, "loss": 0.3464, "lr": 3.878266982742109e-06, "epoch": 2.6147186147186146, "percentage": 87.28, "elapsed_time": "6:42:43", "remaining_time": "0:58:40"}
+{"current_steps": 454, "total_steps": 519, "loss": 0.3493, "lr": 3.7634989440538603e-06, "epoch": 2.6204906204906204, "percentage": 87.48, "elapsed_time": "6:43:35", "remaining_time": "0:57:46"}
+{"current_steps": 455, "total_steps": 519, "loss": 0.3407, "lr": 3.650370779827661e-06, "epoch": 2.6262626262626263, "percentage": 87.67, "elapsed_time": "6:44:27", "remaining_time": "0:56:53"}
+{"current_steps": 456, "total_steps": 519, "loss": 0.3378, "lr": 3.538887609652388e-06, "epoch": 2.632034632034632, "percentage": 87.86, "elapsed_time": "6:45:18", "remaining_time": "0:55:59"}
+{"current_steps": 457, "total_steps": 519, "loss": 0.3414, "lr": 3.4290544786730996e-06, "epoch": 2.637806637806638, "percentage": 88.05, "elapsed_time": "6:46:15", "remaining_time": "0:55:06"}
+{"current_steps": 458, "total_steps": 519, "loss": 0.3423, "lr": 3.3208763573626766e-06, "epoch": 2.6435786435786435, "percentage": 88.25, "elapsed_time": "6:47:21", "remaining_time": "0:54:15"}
+{"current_steps": 459, "total_steps": 519, "loss": 0.3454, "lr": 3.214358141296949e-06, "epoch": 2.6493506493506493, "percentage": 88.44, "elapsed_time": "6:48:17", "remaining_time": "0:53:22"}
+{"current_steps": 460, "total_steps": 519, "loss": 0.3406, "lr": 3.1095046509331196e-06, "epoch": 2.655122655122655, "percentage": 88.63, "elapsed_time": "6:49:09", "remaining_time": "0:52:28"}
+{"current_steps": 461, "total_steps": 519, "loss": 0.3476, "lr": 3.006320631391613e-06, "epoch": 2.6608946608946606, "percentage": 88.82, "elapsed_time": "6:50:01", "remaining_time": "0:51:35"}
+{"current_steps": 462, "total_steps": 519, "loss": 0.3351, "lr": 2.904810752241316e-06, "epoch": 2.6666666666666665, "percentage": 89.02, "elapsed_time": "6:50:53", "remaining_time": "0:50:41"}
+{"current_steps": 463, "total_steps": 519, "loss": 0.3364, "lr": 2.804979607288312e-06, "epoch": 2.6724386724386724, "percentage": 89.21, "elapsed_time": "6:51:45", "remaining_time": "0:49:48"}
+{"current_steps": 464, "total_steps": 519, "loss": 0.351, "lr": 2.706831714367932e-06, "epoch": 2.6782106782106783, "percentage": 89.4, "elapsed_time": "6:52:37", "remaining_time": "0:48:54"}
+{"current_steps": 465, "total_steps": 519, "loss": 0.3432, "lr": 2.6103715151403463e-06, "epoch": 2.683982683982684, "percentage": 89.6, "elapsed_time": "6:53:28", "remaining_time": "0:48:01"}
+{"current_steps": 466, "total_steps": 519, "loss": 0.3444, "lr": 2.515603374889519e-06, "epoch": 2.6897546897546896, "percentage": 89.79, "elapsed_time": "6:54:20", "remaining_time": "0:47:07"}
+{"current_steps": 467, "total_steps": 519, "loss": 0.3396, "lr": 2.4225315823257e-06, "epoch": 2.6955266955266954, "percentage": 89.98, "elapsed_time": "6:55:12", "remaining_time": "0:46:13"}
+{"current_steps": 468, "total_steps": 519, "loss": 0.3452, "lr": 2.3311603493912927e-06, "epoch": 2.7012987012987013, "percentage": 90.17, "elapsed_time": "6:56:04", "remaining_time": "0:45:20"}
+{"current_steps": 469, "total_steps": 519, "loss": 0.3468, "lr": 2.241493811070279e-06, "epoch": 2.707070707070707, "percentage": 90.37, "elapsed_time": "6:56:55", "remaining_time": "0:44:26"}
+{"current_steps": 470, "total_steps": 519, "loss": 0.3399, "lr": 2.1535360252010885e-06, "epoch": 2.712842712842713, "percentage": 90.56, "elapsed_time": "6:57:47", "remaining_time": "0:43:33"}
+{"current_steps": 471, "total_steps": 519, "loss": 0.3356, "lr": 2.0672909722929412e-06, "epoch": 2.7186147186147185, "percentage": 90.75, "elapsed_time": "6:58:39", "remaining_time": "0:42:39"}
+{"current_steps": 472, "total_steps": 519, "loss": 0.3458, "lr": 1.982762555345734e-06, "epoch": 2.7243867243867244, "percentage": 90.94, "elapsed_time": "6:59:37", "remaining_time": "0:41:47"}
+{"current_steps": 473, "total_steps": 519, "loss": 0.3379, "lr": 1.8999545996734037e-06, "epoch": 2.7301587301587302, "percentage": 91.14, "elapsed_time": "7:00:43", "remaining_time": "0:40:54"}
+{"current_steps": 474, "total_steps": 519, "loss": 0.3419, "lr": 1.8188708527307942e-06, "epoch": 2.7359307359307357, "percentage": 91.33, "elapsed_time": "7:01:39", "remaining_time": "0:40:01"}
+{"current_steps": 475, "total_steps": 519, "loss": 0.3531, "lr": 1.7395149839440994e-06, "epoch": 2.741702741702742, "percentage": 91.52, "elapsed_time": "7:02:31", "remaining_time": "0:39:08"}
+{"current_steps": 476, "total_steps": 519, "loss": 0.339, "lr": 1.6618905845447697e-06, "epoch": 2.7474747474747474, "percentage": 91.71, "elapsed_time": "7:03:22", "remaining_time": "0:38:14"}
+{"current_steps": 477, "total_steps": 519, "loss": 0.3501, "lr": 1.5860011674070275e-06, "epoch": 2.7532467532467533, "percentage": 91.91, "elapsed_time": "7:04:14", "remaining_time": "0:37:21"}
+{"current_steps": 478, "total_steps": 519, "loss": 0.3398, "lr": 1.5118501668888664e-06, "epoch": 2.759018759018759, "percentage": 92.1, "elapsed_time": "7:05:06", "remaining_time": "0:36:27"}
+{"current_steps": 479, "total_steps": 519, "loss": 0.3463, "lr": 1.4394409386766461e-06, "epoch": 2.7647907647907646, "percentage": 92.29, "elapsed_time": "7:05:58", "remaining_time": "0:35:34"}
+{"current_steps": 480, "total_steps": 519, "loss": 0.3353, "lr": 1.3687767596332058e-06, "epoch": 2.7705627705627704, "percentage": 92.49, "elapsed_time": "7:06:49", "remaining_time": "0:34:40"}
+{"current_steps": 481, "total_steps": 519, "loss": 0.3454, "lr": 1.2998608276496128e-06, "epoch": 2.7763347763347763, "percentage": 92.68, "elapsed_time": "7:07:41", "remaining_time": "0:33:47"}
+{"current_steps": 482, "total_steps": 519, "loss": 0.3379, "lr": 1.232696261500399e-06, "epoch": 2.782106782106782, "percentage": 92.87, "elapsed_time": "7:08:33", "remaining_time": "0:32:53"}
+{"current_steps": 483, "total_steps": 519, "loss": 0.3445, "lr": 1.1672861007024472e-06, "epoch": 2.787878787878788, "percentage": 93.06, "elapsed_time": "7:09:25", "remaining_time": "0:32:00"}
+{"current_steps": 484, "total_steps": 519, "loss": 0.3381, "lr": 1.1036333053774429e-06, "epoch": 2.7936507936507935, "percentage": 93.26, "elapsed_time": "7:10:17", "remaining_time": "0:31:06"}
+{"current_steps": 485, "total_steps": 519, "loss": 0.346, "lr": 1.0417407561178839e-06, "epoch": 2.7994227994227994, "percentage": 93.45, "elapsed_time": "7:11:09", "remaining_time": "0:30:13"}
+{"current_steps": 486, "total_steps": 519, "loss": 0.3417, "lr": 9.816112538567578e-07, "epoch": 2.8051948051948052, "percentage": 93.64, "elapsed_time": "7:12:00", "remaining_time": "0:29:20"}
+{"current_steps": 487, "total_steps": 519, "loss": 0.3418, "lr": 9.232475197407553e-07, "epoch": 2.810966810966811, "percentage": 93.83, "elapsed_time": "7:13:01", "remaining_time": "0:28:27"}
+{"current_steps": 488, "total_steps": 519, "loss": 0.3361, "lr": 8.666521950071538e-07, "epoch": 2.816738816738817, "percentage": 94.03, "elapsed_time": "7:14:05", "remaining_time": "0:27:34"}
+{"current_steps": 489, "total_steps": 519, "loss": 0.3417, "lr": 8.118278408642655e-07, "epoch": 2.8225108225108224, "percentage": 94.22, "elapsed_time": "7:14:59", "remaining_time": "0:26:41"}
+{"current_steps": 490, "total_steps": 519, "loss": 0.3476, "lr": 7.587769383755383e-07, "epoch": 2.8282828282828283, "percentage": 94.41, "elapsed_time": "7:15:51", "remaining_time": "0:25:47"}
+{"current_steps": 491, "total_steps": 519, "loss": 0.3498, "lr": 7.07501888347264e-07, "epoch": 2.834054834054834, "percentage": 94.61, "elapsed_time": "7:16:42", "remaining_time": "0:24:54"}
+{"current_steps": 492, "total_steps": 519, "loss": 0.3358, "lr": 6.580050112199709e-07, "epoch": 2.8398268398268396, "percentage": 94.8, "elapsed_time": "7:17:35", "remaining_time": "0:24:00"}
+{"current_steps": 493, "total_steps": 519, "loss": 0.3427, "lr": 6.10288546963358e-07, "epoch": 2.8455988455988455, "percentage": 94.99, "elapsed_time": "7:18:26", "remaining_time": "0:23:07"}
+{"current_steps": 494, "total_steps": 519, "loss": 0.3361, "lr": 5.643546549749656e-07, "epoch": 2.8513708513708513, "percentage": 95.18, "elapsed_time": "7:19:18", "remaining_time": "0:22:13"}
+{"current_steps": 495, "total_steps": 519, "loss": 0.3416, "lr": 5.202054139824419e-07, "epoch": 2.857142857142857, "percentage": 95.38, "elapsed_time": "7:20:10", "remaining_time": "0:21:20"}
+{"current_steps": 496, "total_steps": 519, "loss": 0.3431, "lr": 4.778428219494613e-07, "epoch": 2.862914862914863, "percentage": 95.57, "elapsed_time": "7:21:02", "remaining_time": "0:20:27"}
+{"current_steps": 497, "total_steps": 519, "loss": 0.3372, "lr": 4.372687959853039e-07, "epoch": 2.8686868686868685, "percentage": 95.76, "elapsed_time": "7:21:54", "remaining_time": "0:19:33"}
+{"current_steps": 498, "total_steps": 519, "loss": 0.3337, "lr": 3.984851722581251e-07, "epoch": 2.8744588744588744, "percentage": 95.95, "elapsed_time": "7:22:46", "remaining_time": "0:18:40"}
+{"current_steps": 499, "total_steps": 519, "loss": 0.3453, "lr": 3.614937059118351e-07, "epoch": 2.8802308802308803, "percentage": 96.15, "elapsed_time": "7:23:37", "remaining_time": "0:17:46"}
+{"current_steps": 500, "total_steps": 519, "loss": 0.3392, "lr": 3.262960709866692e-07, "epoch": 2.886002886002886, "percentage": 96.34, "elapsed_time": "7:24:29", "remaining_time": "0:16:53"}
+{"current_steps": 501, "total_steps": 519, "loss": 0.3441, "lr": 2.9289386034344835e-07, "epoch": 2.891774891774892, "percentage": 96.53, "elapsed_time": "7:25:21", "remaining_time": "0:16:00"}
+{"current_steps": 502, "total_steps": 519, "loss": 0.3376, "lr": 2.6128858559147705e-07, "epoch": 2.8975468975468974, "percentage": 96.72, "elapsed_time": "7:26:22", "remaining_time": "0:15:06"}
+{"current_steps": 503, "total_steps": 519, "loss": 0.3393, "lr": 2.3148167702014e-07, "epoch": 2.9033189033189033, "percentage": 96.92, "elapsed_time": "7:27:24", "remaining_time": "0:14:13"}
+{"current_steps": 504, "total_steps": 519, "loss": 0.34, "lr": 2.0347448353418508e-07, "epoch": 2.909090909090909, "percentage": 97.11, "elapsed_time": "7:28:19", "remaining_time": "0:13:20"}
+{"current_steps": 505, "total_steps": 519, "loss": 0.336, "lr": 1.7726827259266554e-07, "epoch": 2.9148629148629146, "percentage": 97.3, "elapsed_time": "7:29:11", "remaining_time": "0:12:27"}
+{"current_steps": 506, "total_steps": 519, "loss": 0.3353, "lr": 1.5286423015158591e-07, "epoch": 2.9206349206349205, "percentage": 97.5, "elapsed_time": "7:30:03", "remaining_time": "0:11:33"}
+{"current_steps": 507, "total_steps": 519, "loss": 0.3412, "lr": 1.302634606102382e-07, "epoch": 2.9264069264069263, "percentage": 97.69, "elapsed_time": "7:30:54", "remaining_time": "0:10:40"}
+{"current_steps": 508, "total_steps": 519, "loss": 0.3451, "lr": 1.0946698676120637e-07, "epoch": 2.932178932178932, "percentage": 97.88, "elapsed_time": "7:31:46", "remaining_time": "0:09:46"}
+{"current_steps": 509, "total_steps": 519, "loss": 0.3426, "lr": 9.047574974410555e-08, "epoch": 2.937950937950938, "percentage": 98.07, "elapsed_time": "7:32:38", "remaining_time": "0:08:53"}
+{"current_steps": 510, "total_steps": 519, "loss": 0.3424, "lr": 7.32906090029628e-08, "epoch": 2.9437229437229435, "percentage": 98.27, "elapsed_time": "7:33:30", "remaining_time": "0:08:00"}
+{"current_steps": 511, "total_steps": 519, "loss": 0.3378, "lr": 5.791234224734155e-08, "epoch": 2.9494949494949494, "percentage": 98.46, "elapsed_time": "7:34:21", "remaining_time": "0:07:06"}
+{"current_steps": 512, "total_steps": 519, "loss": 0.344, "lr": 4.434164541714747e-08, "epoch": 2.9552669552669553, "percentage": 98.65, "elapsed_time": "7:35:13", "remaining_time": "0:06:13"}
+{"current_steps": 513, "total_steps": 519, "loss": 0.3362, "lr": 3.2579132651111566e-08, "epoch": 2.961038961038961, "percentage": 98.84, "elapsed_time": "7:36:05", "remaining_time": "0:05:20"}
+{"current_steps": 514, "total_steps": 519, "loss": 0.3418, "lr": 2.2625336259038955e-08, "epoch": 2.966810966810967, "percentage": 99.04, "elapsed_time": "7:36:57", "remaining_time": "0:04:26"}
+{"current_steps": 515, "total_steps": 519, "loss": 0.3437, "lr": 1.4480706697668213e-08, "epoch": 2.9725829725829724, "percentage": 99.23, "elapsed_time": "7:37:49", "remaining_time": "0:03:33"}
+{"current_steps": 516, "total_steps": 519, "loss": 0.3529, "lr": 8.145612550327642e-09, "epoch": 2.9783549783549783, "percentage": 99.42, "elapsed_time": "7:38:41", "remaining_time": "0:02:40"}
+{"current_steps": 517, "total_steps": 519, "loss": 0.3441, "lr": 3.620340510233078e-09, "epoch": 2.984126984126984, "percentage": 99.61, "elapsed_time": "7:39:45", "remaining_time": "0:01:46"}
+{"current_steps": 518, "total_steps": 519, "loss": 0.3343, "lr": 9.050953675293627e-10, "epoch": 2.98989898989899, "percentage": 99.81, "elapsed_time": "7:40:45", "remaining_time": "0:00:53"}
+{"current_steps": 519, "total_steps": 519, "loss": 0.3398, "lr": 0.0, "epoch": 2.995670995670996, "percentage": 100.0, "elapsed_time": "7:41:38", "remaining_time": "0:00:00"}
+{"current_steps": 519, "total_steps": 519, "epoch": 2.995670995670996, "percentage": 100.0, "elapsed_time": "7:43:07", "remaining_time": "0:00:00"}