Training in progress, epoch 1

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:69e8407f88eeabdf51292acd68a316b4d2c22750652f7a05fbe4e02c8979df8d
+oid sha256:f513f16d1db7e096575918701a285e626cbba8afa0af2ddc6175006947561acd
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:626803f8224e439bf5f3f1b99b7bd7356b0ee10deebe418475de183b8d5d7e74
+oid sha256:2040fcb22213b468006bc4828b729e9e947dda3973bad4a6bc7662a646e86cf4
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:41aa7e1a58000b519be75ca4d75d76b09824e7fba9d8848c13e41b7ff7f7d4ac
+oid sha256:fb5df45540452f2733f97f41dac5bbe8e5a91c28188fe068e20f02702a2c5c43
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eba11352b7e68d52bab06b90f0f7f51b7d964ba7d5d79e8f06c0a7f9210e3f92
+oid sha256:1433facf80136f13cc2a1657274ffe83722131fad78b1da1eb67666de3b86b57
 size 1089994880

trainer_log.jsonl

@@ -194,3 +194,187 @@
 {"current_steps": 1940, "total_steps": 5547, "loss": 0.4393, "lr": 8.217953707013025e-06, "epoch": 1.0488376284015137, "percentage": 34.97, "elapsed_time": "9:06:31", "remaining_time": "16:56:08"}
 {"current_steps": 1950, "total_steps": 5547, "loss": 0.4383, "lr": 8.193806807284064e-06, "epoch": 1.0542440079293567, "percentage": 35.15, "elapsed_time": "9:09:15", "remaining_time": "16:53:10"}
 {"current_steps": 1960, "total_steps": 5547, "loss": 0.4286, "lr": 8.169533417086673e-06, "epoch": 1.0596503874571994, "percentage": 35.33, "elapsed_time": "9:12:04", "remaining_time": "16:50:21"}
+{"current_steps": 1970, "total_steps": 5547, "loss": 0.4467, "lr": 8.145134497766566e-06, "epoch": 1.0650567669850424, "percentage": 35.51, "elapsed_time": "9:14:57", "remaining_time": "16:47:38"}
+{"current_steps": 1980, "total_steps": 5547, "loss": 0.4363, "lr": 8.120611015641036e-06, "epoch": 1.070463146512885, "percentage": 35.69, "elapsed_time": "9:17:46", "remaining_time": "16:44:49"}
+{"current_steps": 1990, "total_steps": 5547, "loss": 0.437, "lr": 8.095963941960667e-06, "epoch": 1.075869526040728, "percentage": 35.88, "elapsed_time": "9:20:39", "remaining_time": "16:42:07"}
+{"current_steps": 2000, "total_steps": 5547, "loss": 0.432, "lr": 8.071194252870887e-06, "epoch": 1.081275905568571, "percentage": 36.06, "elapsed_time": "9:23:21", "remaining_time": "16:39:07"}
+{"current_steps": 2010, "total_steps": 5547, "loss": 0.4367, "lr": 8.046302929373286e-06, "epoch": 1.0866822850964137, "percentage": 36.24, "elapsed_time": "9:26:06", "remaining_time": "16:36:10"}
+{"current_steps": 2020, "total_steps": 5547, "loss": 0.4352, "lr": 8.021290957286787e-06, "epoch": 1.0920886646242567, "percentage": 36.42, "elapsed_time": "9:28:58", "remaining_time": "16:33:27"}
+{"current_steps": 2030, "total_steps": 5547, "loss": 0.4434, "lr": 7.996159327208581e-06, "epoch": 1.0974950441520994, "percentage": 36.6, "elapsed_time": "9:31:50", "remaining_time": "16:30:42"}
+{"current_steps": 2040, "total_steps": 5547, "loss": 0.4326, "lr": 7.97090903447491e-06, "epoch": 1.1029014236799424, "percentage": 36.78, "elapsed_time": "9:34:40", "remaining_time": "16:27:55"}
+{"current_steps": 2050, "total_steps": 5547, "loss": 0.4485, "lr": 7.945541079121642e-06, "epoch": 1.108307803207785, "percentage": 36.96, "elapsed_time": "9:37:36", "remaining_time": "16:25:19"}
+{"current_steps": 2060, "total_steps": 5547, "loss": 0.4328, "lr": 7.920056465844658e-06, "epoch": 1.113714182735628, "percentage": 37.14, "elapsed_time": "9:40:21", "remaining_time": "16:22:22"}
+{"current_steps": 2070, "total_steps": 5547, "loss": 0.4339, "lr": 7.894456203960075e-06, "epoch": 1.119120562263471, "percentage": 37.32, "elapsed_time": "9:43:10", "remaining_time": "16:19:33"}
+{"current_steps": 2080, "total_steps": 5547, "loss": 0.4307, "lr": 7.868741307364255e-06, "epoch": 1.1245269417913137, "percentage": 37.5, "elapsed_time": "9:46:00", "remaining_time": "16:16:45"}
+{"current_steps": 2090, "total_steps": 5547, "loss": 0.4357, "lr": 7.842912794493667e-06, "epoch": 1.1299333213191567, "percentage": 37.68, "elapsed_time": "9:48:52", "remaining_time": "16:14:02"}
+{"current_steps": 2100, "total_steps": 5547, "loss": 0.4429, "lr": 7.81697168828454e-06, "epoch": 1.1353397008469994, "percentage": 37.86, "elapsed_time": "9:51:42", "remaining_time": "16:11:14"}
+{"current_steps": 2110, "total_steps": 5547, "loss": 0.4435, "lr": 7.790919016132351e-06, "epoch": 1.1407460803748424, "percentage": 38.04, "elapsed_time": "9:54:34", "remaining_time": "16:08:29"}
+{"current_steps": 2120, "total_steps": 5547, "loss": 0.4375, "lr": 7.764755809851141e-06, "epoch": 1.146152459902685, "percentage": 38.22, "elapsed_time": "9:57:18", "remaining_time": "16:05:32"}
+{"current_steps": 2130, "total_steps": 5547, "loss": 0.4408, "lr": 7.738483105632644e-06, "epoch": 1.151558839430528, "percentage": 38.4, "elapsed_time": "10:00:10", "remaining_time": "16:02:48"}
+{"current_steps": 2140, "total_steps": 5547, "loss": 0.442, "lr": 7.712101944005256e-06, "epoch": 1.1569652189583708, "percentage": 38.58, "elapsed_time": "10:03:02", "remaining_time": "16:00:04"}
+{"current_steps": 2150, "total_steps": 5547, "loss": 0.4389, "lr": 7.685613369792815e-06, "epoch": 1.1623715984862137, "percentage": 38.76, "elapsed_time": "10:05:50", "remaining_time": "15:57:13"}
+{"current_steps": 2160, "total_steps": 5547, "loss": 0.4372, "lr": 7.65901843207323e-06, "epoch": 1.1677779780140565, "percentage": 38.94, "elapsed_time": "10:08:35", "remaining_time": "15:54:19"}
+{"current_steps": 2170, "total_steps": 5547, "loss": 0.4313, "lr": 7.63231818413692e-06, "epoch": 1.1731843575418994, "percentage": 39.12, "elapsed_time": "10:11:24", "remaining_time": "15:51:28"}
+{"current_steps": 2180, "total_steps": 5547, "loss": 0.433, "lr": 7.605513683445118e-06, "epoch": 1.1785907370697424, "percentage": 39.3, "elapsed_time": "10:14:15", "remaining_time": "15:48:43"}
+{"current_steps": 2190, "total_steps": 5547, "loss": 0.43, "lr": 7.578605991587974e-06, "epoch": 1.183997116597585, "percentage": 39.48, "elapsed_time": "10:17:07", "remaining_time": "15:45:58"}
+{"current_steps": 2200, "total_steps": 5547, "loss": 0.4357, "lr": 7.5515961742425146e-06, "epoch": 1.189403496125428, "percentage": 39.66, "elapsed_time": "10:19:58", "remaining_time": "15:43:11"}
+{"current_steps": 2210, "total_steps": 5547, "loss": 0.4363, "lr": 7.524485301130443e-06, "epoch": 1.1948098756532708, "percentage": 39.84, "elapsed_time": "10:22:49", "remaining_time": "15:40:25"}
+{"current_steps": 2220, "total_steps": 5547, "loss": 0.4283, "lr": 7.497274445975762e-06, "epoch": 1.2002162551811137, "percentage": 40.02, "elapsed_time": "10:25:36", "remaining_time": "15:37:34"}
+{"current_steps": 2230, "total_steps": 5547, "loss": 0.4416, "lr": 7.469964686462261e-06, "epoch": 1.2056226347089565, "percentage": 40.2, "elapsed_time": "10:28:26", "remaining_time": "15:34:46"}
+{"current_steps": 2240, "total_steps": 5547, "loss": 0.4388, "lr": 7.4425571041908254e-06, "epoch": 1.2110290142367994, "percentage": 40.38, "elapsed_time": "10:31:16", "remaining_time": "15:31:58"}
+{"current_steps": 2250, "total_steps": 5547, "loss": 0.4401, "lr": 7.415052784636603e-06, "epoch": 1.2164353937646424, "percentage": 40.56, "elapsed_time": "10:34:01", "remaining_time": "15:29:03"}
+{"current_steps": 2260, "total_steps": 5547, "loss": 0.4313, "lr": 7.387452817106017e-06, "epoch": 1.2218417732924851, "percentage": 40.74, "elapsed_time": "10:36:55", "remaining_time": "15:26:21"}
+{"current_steps": 2270, "total_steps": 5547, "loss": 0.4392, "lr": 7.359758294693618e-06, "epoch": 1.227248152820328, "percentage": 40.92, "elapsed_time": "10:39:50", "remaining_time": "15:23:40"}
+{"current_steps": 2280, "total_steps": 5547, "loss": 0.4405, "lr": 7.331970314238799e-06, "epoch": 1.2326545323481708, "percentage": 41.1, "elapsed_time": "10:42:37", "remaining_time": "15:20:49"}
+{"current_steps": 2290, "total_steps": 5547, "loss": 0.4401, "lr": 7.304089976282348e-06, "epoch": 1.2380609118760137, "percentage": 41.28, "elapsed_time": "10:45:24", "remaining_time": "15:17:56"}
+{"current_steps": 2300, "total_steps": 5547, "loss": 0.4241, "lr": 7.276118385022865e-06, "epoch": 1.2434672914038565, "percentage": 41.46, "elapsed_time": "10:48:12", "remaining_time": "15:15:06"}
+{"current_steps": 2310, "total_steps": 5547, "loss": 0.4425, "lr": 7.248056648273034e-06, "epoch": 1.2488736709316994, "percentage": 41.64, "elapsed_time": "10:51:05", "remaining_time": "15:12:22"}
+{"current_steps": 2320, "total_steps": 5547, "loss": 0.4276, "lr": 7.2199058774157375e-06, "epoch": 1.2542800504595424, "percentage": 41.82, "elapsed_time": "10:53:50", "remaining_time": "15:09:28"}
+{"current_steps": 2330, "total_steps": 5547, "loss": 0.4312, "lr": 7.1916671873600515e-06, "epoch": 1.2596864299873851, "percentage": 42.0, "elapsed_time": "10:56:35", "remaining_time": "15:06:32"}
+{"current_steps": 2340, "total_steps": 5547, "loss": 0.4405, "lr": 7.163341696497084e-06, "epoch": 1.2650928095152278, "percentage": 42.18, "elapsed_time": "10:59:27", "remaining_time": "15:03:47"}
+{"current_steps": 2350, "total_steps": 5547, "loss": 0.4347, "lr": 7.134930526655679e-06, "epoch": 1.2704991890430708, "percentage": 42.37, "elapsed_time": "11:02:14", "remaining_time": "15:00:55"}
+{"current_steps": 2360, "total_steps": 5547, "loss": 0.4392, "lr": 7.106434803057998e-06, "epoch": 1.2759055685709138, "percentage": 42.55, "elapsed_time": "11:04:59", "remaining_time": "14:58:01"}
+{"current_steps": 2370, "total_steps": 5547, "loss": 0.4329, "lr": 7.077855654274939e-06, "epoch": 1.2813119480987565, "percentage": 42.73, "elapsed_time": "11:07:45", "remaining_time": "14:55:08"}
+{"current_steps": 2380, "total_steps": 5547, "loss": 0.4344, "lr": 7.04919421218145e-06, "epoch": 1.2867183276265994, "percentage": 42.91, "elapsed_time": "11:10:35", "remaining_time": "14:52:19"}
+{"current_steps": 2390, "total_steps": 5547, "loss": 0.4274, "lr": 7.020451611911703e-06, "epoch": 1.2921247071544422, "percentage": 43.09, "elapsed_time": "11:13:26", "remaining_time": "14:49:34"}
+{"current_steps": 2400, "total_steps": 5547, "loss": 0.4383, "lr": 6.9916289918141265e-06, "epoch": 1.2975310866822851, "percentage": 43.27, "elapsed_time": "11:16:06", "remaining_time": "14:46:32"}
+{"current_steps": 2410, "total_steps": 5547, "loss": 0.4363, "lr": 6.962727493406335e-06, "epoch": 1.3029374662101278, "percentage": 43.45, "elapsed_time": "11:18:57", "remaining_time": "14:43:46"}
+{"current_steps": 2420, "total_steps": 5547, "loss": 0.4251, "lr": 6.9337482613299065e-06, "epoch": 1.3083438457379708, "percentage": 43.63, "elapsed_time": "11:21:49", "remaining_time": "14:41:01"}
+{"current_steps": 2430, "total_steps": 5547, "loss": 0.439, "lr": 6.904692443305059e-06, "epoch": 1.3137502252658138, "percentage": 43.81, "elapsed_time": "11:24:38", "remaining_time": "14:38:12"}
+{"current_steps": 2440, "total_steps": 5547, "loss": 0.4268, "lr": 6.87556119008519e-06, "epoch": 1.3191566047936565, "percentage": 43.99, "elapsed_time": "11:27:23", "remaining_time": "14:35:17"}
+{"current_steps": 2450, "total_steps": 5547, "loss": 0.4353, "lr": 6.8463556554113005e-06, "epoch": 1.3245629843214994, "percentage": 44.17, "elapsed_time": "11:30:13", "remaining_time": "14:32:30"}
+{"current_steps": 2460, "total_steps": 5547, "loss": 0.4292, "lr": 6.8170769959663045e-06, "epoch": 1.3299693638493422, "percentage": 44.35, "elapsed_time": "11:33:07", "remaining_time": "14:29:47"}
+{"current_steps": 2470, "total_steps": 5547, "loss": 0.4402, "lr": 6.787726371329214e-06, "epoch": 1.3353757433771851, "percentage": 44.53, "elapsed_time": "11:35:55", "remaining_time": "14:26:56"}
+{"current_steps": 2480, "total_steps": 5547, "loss": 0.4369, "lr": 6.7583049439292205e-06, "epoch": 1.3407821229050279, "percentage": 44.71, "elapsed_time": "11:38:45", "remaining_time": "14:24:08"}
+{"current_steps": 2490, "total_steps": 5547, "loss": 0.4377, "lr": 6.728813878999652e-06, "epoch": 1.3461885024328708, "percentage": 44.89, "elapsed_time": "11:41:33", "remaining_time": "14:21:18"}
+{"current_steps": 2500, "total_steps": 5547, "loss": 0.4309, "lr": 6.699254344531821e-06, "epoch": 1.3515948819607138, "percentage": 45.07, "elapsed_time": "11:44:23", "remaining_time": "14:18:30"}
+{"current_steps": 2510, "total_steps": 5547, "loss": 0.434, "lr": 6.669627511228778e-06, "epoch": 1.3570012614885565, "percentage": 45.25, "elapsed_time": "11:47:11", "remaining_time": "14:15:40"}
+{"current_steps": 2520, "total_steps": 5547, "loss": 0.4401, "lr": 6.6399345524589366e-06, "epoch": 1.3624076410163992, "percentage": 45.43, "elapsed_time": "11:50:01", "remaining_time": "14:12:52"}
+{"current_steps": 2530, "total_steps": 5547, "loss": 0.4266, "lr": 6.610176644209602e-06, "epoch": 1.3678140205442422, "percentage": 45.61, "elapsed_time": "11:52:49", "remaining_time": "14:10:01"}
+{"current_steps": 2540, "total_steps": 5547, "loss": 0.4393, "lr": 6.580354965040396e-06, "epoch": 1.3732204000720851, "percentage": 45.79, "elapsed_time": "11:55:38", "remaining_time": "14:07:12"}
+{"current_steps": 2550, "total_steps": 5547, "loss": 0.4276, "lr": 6.550470696036591e-06, "epoch": 1.3786267795999279, "percentage": 45.97, "elapsed_time": "11:58:26", "remaining_time": "14:04:23"}
+{"current_steps": 2560, "total_steps": 5547, "loss": 0.4374, "lr": 6.520525020762318e-06, "epoch": 1.3840331591277708, "percentage": 46.15, "elapsed_time": "12:01:18", "remaining_time": "14:01:37"}
+{"current_steps": 2570, "total_steps": 5547, "loss": 0.44, "lr": 6.490519125213701e-06, "epoch": 1.3894395386556138, "percentage": 46.33, "elapsed_time": "12:03:58", "remaining_time": "13:58:37"}
+{"current_steps": 2580, "total_steps": 5547, "loss": 0.4347, "lr": 6.460454197771881e-06, "epoch": 1.3948459181834565, "percentage": 46.51, "elapsed_time": "12:06:48", "remaining_time": "13:55:50"}
+{"current_steps": 2590, "total_steps": 5547, "loss": 0.4398, "lr": 6.430331429155956e-06, "epoch": 1.4002522977112992, "percentage": 46.69, "elapsed_time": "12:09:36", "remaining_time": "13:52:59"}
+{"current_steps": 2600, "total_steps": 5547, "loss": 0.4361, "lr": 6.400152012375818e-06, "epoch": 1.4056586772391422, "percentage": 46.87, "elapsed_time": "12:12:25", "remaining_time": "13:50:10"}
+{"current_steps": 2610, "total_steps": 5547, "loss": 0.433, "lr": 6.3699171426849036e-06, "epoch": 1.4110650567669851, "percentage": 47.05, "elapsed_time": "12:15:13", "remaining_time": "13:47:19"}
+{"current_steps": 2620, "total_steps": 5547, "loss": 0.4305, "lr": 6.339628017532858e-06, "epoch": 1.4164714362948279, "percentage": 47.23, "elapsed_time": "12:18:02", "remaining_time": "13:44:31"}
+{"current_steps": 2630, "total_steps": 5547, "loss": 0.4289, "lr": 6.309285836518113e-06, "epoch": 1.4218778158226708, "percentage": 47.41, "elapsed_time": "12:20:52", "remaining_time": "13:41:43"}
+{"current_steps": 2640, "total_steps": 5547, "loss": 0.4312, "lr": 6.2788918013403695e-06, "epoch": 1.4272841953505135, "percentage": 47.59, "elapsed_time": "12:23:36", "remaining_time": "13:38:49"}
+{"current_steps": 2650, "total_steps": 5547, "loss": 0.4327, "lr": 6.248447115753009e-06, "epoch": 1.4326905748783565, "percentage": 47.77, "elapsed_time": "12:26:25", "remaining_time": "13:35:59"}
+{"current_steps": 2660, "total_steps": 5547, "loss": 0.4206, "lr": 6.21795298551542e-06, "epoch": 1.4380969544061992, "percentage": 47.95, "elapsed_time": "12:29:03", "remaining_time": "13:32:58"}
+{"current_steps": 2670, "total_steps": 5547, "loss": 0.4337, "lr": 6.187410618345241e-06, "epoch": 1.4435033339340422, "percentage": 48.13, "elapsed_time": "12:31:51", "remaining_time": "13:30:08"}
+{"current_steps": 2680, "total_steps": 5547, "loss": 0.428, "lr": 6.156821223870533e-06, "epoch": 1.4489097134618851, "percentage": 48.31, "elapsed_time": "12:34:36", "remaining_time": "13:27:15"}
+{"current_steps": 2690, "total_steps": 5547, "loss": 0.4442, "lr": 6.126186013581868e-06, "epoch": 1.4543160929897279, "percentage": 48.49, "elapsed_time": "12:37:25", "remaining_time": "13:24:26"}
+{"current_steps": 2700, "total_steps": 5547, "loss": 0.4313, "lr": 6.095506200784349e-06, "epoch": 1.4597224725175708, "percentage": 48.67, "elapsed_time": "12:40:16", "remaining_time": "13:21:39"}
+{"current_steps": 2710, "total_steps": 5547, "loss": 0.4443, "lr": 6.06478300054956e-06, "epoch": 1.4651288520454135, "percentage": 48.86, "elapsed_time": "12:43:04", "remaining_time": "13:18:50"}
+{"current_steps": 2720, "total_steps": 5547, "loss": 0.4321, "lr": 6.034017629667439e-06, "epoch": 1.4705352315732565, "percentage": 49.04, "elapsed_time": "12:45:51", "remaining_time": "13:15:58"}
+{"current_steps": 2730, "total_steps": 5547, "loss": 0.4302, "lr": 6.003211306598089e-06, "epoch": 1.4759416111010992, "percentage": 49.22, "elapsed_time": "12:48:35", "remaining_time": "13:13:05"}
+{"current_steps": 2740, "total_steps": 5547, "loss": 0.4331, "lr": 5.972365251423521e-06, "epoch": 1.4813479906289422, "percentage": 49.4, "elapsed_time": "12:51:22", "remaining_time": "13:10:14"}
+{"current_steps": 2750, "total_steps": 5547, "loss": 0.433, "lr": 5.941480685799338e-06, "epoch": 1.4867543701567851, "percentage": 49.58, "elapsed_time": "12:54:04", "remaining_time": "13:07:18"}
+{"current_steps": 2760, "total_steps": 5547, "loss": 0.4378, "lr": 5.910558832906341e-06, "epoch": 1.4921607496846279, "percentage": 49.76, "elapsed_time": "12:56:52", "remaining_time": "13:04:28"}
+{"current_steps": 2770, "total_steps": 5547, "loss": 0.4322, "lr": 5.879600917402089e-06, "epoch": 1.4975671292124706, "percentage": 49.94, "elapsed_time": "12:59:39", "remaining_time": "13:01:38"}
+{"current_steps": 2780, "total_steps": 5547, "loss": 0.425, "lr": 5.848608165372403e-06, "epoch": 1.5029735087403135, "percentage": 50.12, "elapsed_time": "13:02:28", "remaining_time": "12:58:49"}
+{"current_steps": 2790, "total_steps": 5547, "loss": 0.4357, "lr": 5.8175818042828e-06, "epoch": 1.5083798882681565, "percentage": 50.3, "elapsed_time": "13:05:13", "remaining_time": "12:55:55"}
+{"current_steps": 2800, "total_steps": 5547, "loss": 0.4395, "lr": 5.78652306292988e-06, "epoch": 1.5137862677959992, "percentage": 50.48, "elapsed_time": "13:07:54", "remaining_time": "12:52:59"}
+{"current_steps": 2810, "total_steps": 5547, "loss": 0.4426, "lr": 5.75543317139266e-06, "epoch": 1.5191926473238422, "percentage": 50.66, "elapsed_time": "13:10:44", "remaining_time": "12:50:12"}
+{"current_steps": 2820, "total_steps": 5547, "loss": 0.4335, "lr": 5.724313360983859e-06, "epoch": 1.5245990268516851, "percentage": 50.84, "elapsed_time": "13:13:36", "remaining_time": "12:47:26"}
+{"current_steps": 2830, "total_steps": 5547, "loss": 0.4343, "lr": 5.693164864201134e-06, "epoch": 1.5300054063795279, "percentage": 51.02, "elapsed_time": "13:16:29", "remaining_time": "12:44:40"}
+{"current_steps": 2840, "total_steps": 5547, "loss": 0.4201, "lr": 5.661988914678257e-06, "epoch": 1.5354117859073706, "percentage": 51.2, "elapsed_time": "13:19:21", "remaining_time": "12:41:55"}
+{"current_steps": 2850, "total_steps": 5547, "loss": 0.4263, "lr": 5.630786747136269e-06, "epoch": 1.5408181654352135, "percentage": 51.38, "elapsed_time": "13:22:05", "remaining_time": "12:39:02"}
+{"current_steps": 2860, "total_steps": 5547, "loss": 0.4327, "lr": 5.599559597334568e-06, "epoch": 1.5462245449630565, "percentage": 51.56, "elapsed_time": "13:24:56", "remaining_time": "12:36:14"}
+{"current_steps": 2870, "total_steps": 5547, "loss": 0.4284, "lr": 5.56830870202198e-06, "epoch": 1.5516309244908992, "percentage": 51.74, "elapsed_time": "13:27:42", "remaining_time": "12:33:23"}
+{"current_steps": 2880, "total_steps": 5547, "loss": 0.4291, "lr": 5.537035298887764e-06, "epoch": 1.557037304018742, "percentage": 51.92, "elapsed_time": "13:30:30", "remaining_time": "12:30:34"}
+{"current_steps": 2890, "total_steps": 5547, "loss": 0.4333, "lr": 5.505740626512601e-06, "epoch": 1.562443683546585, "percentage": 52.1, "elapsed_time": "13:33:24", "remaining_time": "12:27:49"}
+{"current_steps": 2900, "total_steps": 5547, "loss": 0.4313, "lr": 5.474425924319538e-06, "epoch": 1.5678500630744279, "percentage": 52.28, "elapsed_time": "13:36:12", "remaining_time": "12:25:00"}
+{"current_steps": 2910, "total_steps": 5547, "loss": 0.4446, "lr": 5.443092432524906e-06, "epoch": 1.5732564426022706, "percentage": 52.46, "elapsed_time": "13:39:03", "remaining_time": "12:22:13"}
+{"current_steps": 2920, "total_steps": 5547, "loss": 0.4276, "lr": 5.411741392089192e-06, "epoch": 1.5786628221301136, "percentage": 52.64, "elapsed_time": "13:42:01", "remaining_time": "12:19:32"}
+{"current_steps": 2930, "total_steps": 5547, "loss": 0.4363, "lr": 5.380374044667896e-06, "epoch": 1.5840692016579565, "percentage": 52.82, "elapsed_time": "13:44:48", "remaining_time": "12:16:41"}
+{"current_steps": 2940, "total_steps": 5547, "loss": 0.4347, "lr": 5.348991632562355e-06, "epoch": 1.5894755811857992, "percentage": 53.0, "elapsed_time": "13:47:40", "remaining_time": "12:13:55"}
+{"current_steps": 2950, "total_steps": 5547, "loss": 0.4203, "lr": 5.317595398670543e-06, "epoch": 1.594881960713642, "percentage": 53.18, "elapsed_time": "13:50:25", "remaining_time": "12:11:03"}
+{"current_steps": 2960, "total_steps": 5547, "loss": 0.4394, "lr": 5.286186586437845e-06, "epoch": 1.600288340241485, "percentage": 53.36, "elapsed_time": "13:53:16", "remaining_time": "12:08:16"}
+{"current_steps": 2970, "total_steps": 5547, "loss": 0.4224, "lr": 5.254766439807807e-06, "epoch": 1.6056947197693279, "percentage": 53.54, "elapsed_time": "13:56:02", "remaining_time": "12:05:25"}
+{"current_steps": 2980, "total_steps": 5547, "loss": 0.4289, "lr": 5.223336203172874e-06, "epoch": 1.6111010992971706, "percentage": 53.72, "elapsed_time": "13:58:49", "remaining_time": "12:02:33"}
+{"current_steps": 2990, "total_steps": 5547, "loss": 0.43, "lr": 5.191897121325111e-06, "epoch": 1.6165074788250136, "percentage": 53.9, "elapsed_time": "14:01:28", "remaining_time": "11:59:36"}
+{"current_steps": 3000, "total_steps": 5547, "loss": 0.4344, "lr": 5.16045043940689e-06, "epoch": 1.6219138583528565, "percentage": 54.08, "elapsed_time": "14:04:18", "remaining_time": "11:56:48"}
+{"current_steps": 3010, "total_steps": 5547, "loss": 0.4306, "lr": 5.128997402861584e-06, "epoch": 1.6273202378806992, "percentage": 54.26, "elapsed_time": "14:07:05", "remaining_time": "11:53:58"}
+{"current_steps": 3020, "total_steps": 5547, "loss": 0.4154, "lr": 5.09753925738424e-06, "epoch": 1.632726617408542, "percentage": 54.44, "elapsed_time": "14:09:52", "remaining_time": "11:51:07"}
+{"current_steps": 3030, "total_steps": 5547, "loss": 0.4314, "lr": 5.06607724887225e-06, "epoch": 1.638132996936385, "percentage": 54.62, "elapsed_time": "14:12:46", "remaining_time": "11:48:23"}
+{"current_steps": 3040, "total_steps": 5547, "loss": 0.4412, "lr": 5.034612623375993e-06, "epoch": 1.6435393764642279, "percentage": 54.8, "elapsed_time": "14:15:40", "remaining_time": "11:45:39"}
+{"current_steps": 3050, "total_steps": 5547, "loss": 0.4295, "lr": 5.003146627049499e-06, "epoch": 1.6489457559920706, "percentage": 54.98, "elapsed_time": "14:18:26", "remaining_time": "11:42:48"}
+{"current_steps": 3060, "total_steps": 5547, "loss": 0.4259, "lr": 4.971680506101086e-06, "epoch": 1.6543521355199133, "percentage": 55.16, "elapsed_time": "14:21:19", "remaining_time": "11:40:01"}
+{"current_steps": 3070, "total_steps": 5547, "loss": 0.4254, "lr": 4.940215506744011e-06, "epoch": 1.6597585150477565, "percentage": 55.35, "elapsed_time": "14:24:06", "remaining_time": "11:37:11"}
+{"current_steps": 3080, "total_steps": 5547, "loss": 0.4286, "lr": 4.90875287514711e-06, "epoch": 1.6651648945755992, "percentage": 55.53, "elapsed_time": "14:26:57", "remaining_time": "11:34:24"}
+{"current_steps": 3090, "total_steps": 5547, "loss": 0.426, "lr": 4.87729385738544e-06, "epoch": 1.670571274103442, "percentage": 55.71, "elapsed_time": "14:29:44", "remaining_time": "11:31:34"}
+{"current_steps": 3100, "total_steps": 5547, "loss": 0.4229, "lr": 4.845839699390936e-06, "epoch": 1.675977653631285, "percentage": 55.89, "elapsed_time": "14:32:34", "remaining_time": "11:28:46"}
+{"current_steps": 3110, "total_steps": 5547, "loss": 0.4296, "lr": 4.814391646903063e-06, "epoch": 1.6813840331591279, "percentage": 56.07, "elapsed_time": "14:35:26", "remaining_time": "11:25:59"}
+{"current_steps": 3120, "total_steps": 5547, "loss": 0.4304, "lr": 4.782950945419475e-06, "epoch": 1.6867904126869706, "percentage": 56.25, "elapsed_time": "14:38:16", "remaining_time": "11:23:11"}
+{"current_steps": 3130, "total_steps": 5547, "loss": 0.4329, "lr": 4.751518840146695e-06, "epoch": 1.6921967922148133, "percentage": 56.43, "elapsed_time": "14:41:08", "remaining_time": "11:20:25"}
+{"current_steps": 3140, "total_steps": 5547, "loss": 0.4319, "lr": 4.720096575950784e-06, "epoch": 1.6976031717426563, "percentage": 56.61, "elapsed_time": "14:43:57", "remaining_time": "11:17:36"}
+{"current_steps": 3150, "total_steps": 5547, "loss": 0.42, "lr": 4.688685397308061e-06, "epoch": 1.7030095512704992, "percentage": 56.79, "elapsed_time": "14:46:46", "remaining_time": "11:14:47"}
+{"current_steps": 3160, "total_steps": 5547, "loss": 0.4369, "lr": 4.657286548255789e-06, "epoch": 1.708415930798342, "percentage": 56.97, "elapsed_time": "14:49:36", "remaining_time": "11:11:59"}
+{"current_steps": 3170, "total_steps": 5547, "loss": 0.4274, "lr": 4.6259012723429285e-06, "epoch": 1.713822310326185, "percentage": 57.15, "elapsed_time": "14:52:26", "remaining_time": "11:09:11"}
+{"current_steps": 3180, "total_steps": 5547, "loss": 0.4216, "lr": 4.594530812580876e-06, "epoch": 1.7192286898540279, "percentage": 57.33, "elapsed_time": "14:55:10", "remaining_time": "11:06:18"}
+{"current_steps": 3190, "total_steps": 5547, "loss": 0.4238, "lr": 4.563176411394229e-06, "epoch": 1.7246350693818706, "percentage": 57.51, "elapsed_time": "14:58:03", "remaining_time": "11:03:32"}
+{"current_steps": 3200, "total_steps": 5547, "loss": 0.4291, "lr": 4.531839310571595e-06, "epoch": 1.7300414489097133, "percentage": 57.69, "elapsed_time": "15:00:52", "remaining_time": "11:00:43"}
+{"current_steps": 3210, "total_steps": 5547, "loss": 0.4388, "lr": 4.5005207512163914e-06, "epoch": 1.7354478284375563, "percentage": 57.87, "elapsed_time": "15:03:36", "remaining_time": "10:57:51"}
+{"current_steps": 3220, "total_steps": 5547, "loss": 0.4373, "lr": 4.469221973697714e-06, "epoch": 1.7408542079653992, "percentage": 58.05, "elapsed_time": "15:06:27", "remaining_time": "10:55:04"}
+{"current_steps": 3230, "total_steps": 5547, "loss": 0.4291, "lr": 4.43794421760119e-06, "epoch": 1.746260587493242, "percentage": 58.23, "elapsed_time": "15:09:25", "remaining_time": "10:52:22"}
+{"current_steps": 3240, "total_steps": 5547, "loss": 0.4219, "lr": 4.4066887216799055e-06, "epoch": 1.751666967021085, "percentage": 58.41, "elapsed_time": "15:12:11", "remaining_time": "10:49:30"}
+{"current_steps": 3250, "total_steps": 5547, "loss": 0.4308, "lr": 4.375456723805321e-06, "epoch": 1.7570733465489279, "percentage": 58.59, "elapsed_time": "15:14:58", "remaining_time": "10:46:40"}
+{"current_steps": 3260, "total_steps": 5547, "loss": 0.4213, "lr": 4.344249460918271e-06, "epoch": 1.7624797260767706, "percentage": 58.77, "elapsed_time": "15:17:43", "remaining_time": "10:43:48"}
+{"current_steps": 3270, "total_steps": 5547, "loss": 0.4364, "lr": 4.313068168979957e-06, "epoch": 1.7678861056046133, "percentage": 58.95, "elapsed_time": "15:20:37", "remaining_time": "10:41:03"}
+{"current_steps": 3280, "total_steps": 5547, "loss": 0.4165, "lr": 4.281914082923002e-06, "epoch": 1.7732924851324563, "percentage": 59.13, "elapsed_time": "15:23:29", "remaining_time": "10:38:16"}
+{"current_steps": 3290, "total_steps": 5547, "loss": 0.4269, "lr": 4.250788436602548e-06, "epoch": 1.7786988646602993, "percentage": 59.31, "elapsed_time": "15:26:10", "remaining_time": "10:35:22"}
+{"current_steps": 3300, "total_steps": 5547, "loss": 0.4304, "lr": 4.2196924627473715e-06, "epoch": 1.784105244188142, "percentage": 59.49, "elapsed_time": "15:29:04", "remaining_time": "10:32:36"}
+{"current_steps": 3310, "total_steps": 5547, "loss": 0.4281, "lr": 4.188627392911091e-06, "epoch": 1.7895116237159847, "percentage": 59.67, "elapsed_time": "15:31:52", "remaining_time": "10:29:47"}
+{"current_steps": 3320, "total_steps": 5547, "loss": 0.432, "lr": 4.157594457423357e-06, "epoch": 1.7949180032438277, "percentage": 59.85, "elapsed_time": "15:34:45", "remaining_time": "10:27:01"}
+{"current_steps": 3330, "total_steps": 5547, "loss": 0.427, "lr": 4.1265948853411506e-06, "epoch": 1.8003243827716706, "percentage": 60.03, "elapsed_time": "15:37:32", "remaining_time": "10:24:11"}
+{"current_steps": 3340, "total_steps": 5547, "loss": 0.4268, "lr": 4.095629904400097e-06, "epoch": 1.8057307622995133, "percentage": 60.21, "elapsed_time": "15:40:21", "remaining_time": "10:21:21"}
+{"current_steps": 3350, "total_steps": 5547, "loss": 0.4334, "lr": 4.06470074096584e-06, "epoch": 1.8111371418273563, "percentage": 60.39, "elapsed_time": "15:43:10", "remaining_time": "10:18:33"}
+{"current_steps": 3360, "total_steps": 5547, "loss": 0.4248, "lr": 4.0338086199854765e-06, "epoch": 1.8165435213551993, "percentage": 60.57, "elapsed_time": "15:46:00", "remaining_time": "10:15:44"}
+{"current_steps": 3370, "total_steps": 5547, "loss": 0.4307, "lr": 4.0029547649390346e-06, "epoch": 1.821949900883042, "percentage": 60.75, "elapsed_time": "15:48:51", "remaining_time": "10:12:57"}
+{"current_steps": 3380, "total_steps": 5547, "loss": 0.435, "lr": 3.97214039779103e-06, "epoch": 1.8273562804108847, "percentage": 60.93, "elapsed_time": "15:51:36", "remaining_time": "10:10:05"}
+{"current_steps": 3390, "total_steps": 5547, "loss": 0.4246, "lr": 3.941366738942058e-06, "epoch": 1.8327626599387277, "percentage": 61.11, "elapsed_time": "15:54:30", "remaining_time": "10:07:20"}
+{"current_steps": 3400, "total_steps": 5547, "loss": 0.4394, "lr": 3.910635007180468e-06, "epoch": 1.8381690394665706, "percentage": 61.29, "elapsed_time": "15:57:21", "remaining_time": "10:04:32"}
+{"current_steps": 3410, "total_steps": 5547, "loss": 0.4268, "lr": 3.879946419634087e-06, "epoch": 1.8435754189944134, "percentage": 61.47, "elapsed_time": "16:00:05", "remaining_time": "10:01:40"}
+{"current_steps": 3420, "total_steps": 5547, "loss": 0.4289, "lr": 3.8493021917220225e-06, "epoch": 1.8489817985222563, "percentage": 61.65, "elapsed_time": "16:02:58", "remaining_time": "9:58:54"}
+{"current_steps": 3430, "total_steps": 5547, "loss": 0.427, "lr": 3.818703537106522e-06, "epoch": 1.8543881780500993, "percentage": 61.84, "elapsed_time": "16:05:51", "remaining_time": "9:56:07"}
+{"current_steps": 3440, "total_steps": 5547, "loss": 0.4334, "lr": 3.7881516676449014e-06, "epoch": 1.859794557577942, "percentage": 62.02, "elapsed_time": "16:08:35", "remaining_time": "9:53:15"}
+{"current_steps": 3450, "total_steps": 5547, "loss": 0.4358, "lr": 3.7576477933415612e-06, "epoch": 1.8652009371057847, "percentage": 62.2, "elapsed_time": "16:11:23", "remaining_time": "9:50:26"}
+{"current_steps": 3460, "total_steps": 5547, "loss": 0.4294, "lr": 3.7271931223000507e-06, "epoch": 1.8706073166336277, "percentage": 62.38, "elapsed_time": "16:14:07", "remaining_time": "9:47:34"}
+{"current_steps": 3470, "total_steps": 5547, "loss": 0.433, "lr": 3.6967888606752345e-06, "epoch": 1.8760136961614706, "percentage": 62.56, "elapsed_time": "16:16:56", "remaining_time": "9:44:45"}
+{"current_steps": 3480, "total_steps": 5547, "loss": 0.4283, "lr": 3.6664362126255087e-06, "epoch": 1.8814200756893134, "percentage": 62.74, "elapsed_time": "16:19:44", "remaining_time": "9:41:55"}
+{"current_steps": 3490, "total_steps": 5547, "loss": 0.4189, "lr": 3.636136380265124e-06, "epoch": 1.886826455217156, "percentage": 62.92, "elapsed_time": "16:22:27", "remaining_time": "9:39:03"}
+{"current_steps": 3500, "total_steps": 5547, "loss": 0.4309, "lr": 3.6058905636165674e-06, "epoch": 1.8922328347449993, "percentage": 63.1, "elapsed_time": "16:25:13", "remaining_time": "9:36:12"}
+{"current_steps": 3510, "total_steps": 5547, "loss": 0.4184, "lr": 3.575699960563038e-06, "epoch": 1.897639214272842, "percentage": 63.28, "elapsed_time": "16:27:58", "remaining_time": "9:33:21"}
+{"current_steps": 3520, "total_steps": 5547, "loss": 0.4253, "lr": 3.5455657668010057e-06, "epoch": 1.9030455938006847, "percentage": 63.46, "elapsed_time": "16:30:50", "remaining_time": "9:30:34"}
+{"current_steps": 3530, "total_steps": 5547, "loss": 0.4257, "lr": 3.5154891757928523e-06, "epoch": 1.9084519733285277, "percentage": 63.64, "elapsed_time": "16:33:42", "remaining_time": "9:27:47"}
+{"current_steps": 3540, "total_steps": 5547, "loss": 0.4324, "lr": 3.4854713787196105e-06, "epoch": 1.9138583528563706, "percentage": 63.82, "elapsed_time": "16:36:34", "remaining_time": "9:25:00"}
+{"current_steps": 3550, "total_steps": 5547, "loss": 0.4262, "lr": 3.4555135644337803e-06, "epoch": 1.9192647323842134, "percentage": 64.0, "elapsed_time": "16:39:23", "remaining_time": "9:22:11"}
+{"current_steps": 3560, "total_steps": 5547, "loss": 0.4344, "lr": 3.42561691941225e-06, "epoch": 1.924671111912056, "percentage": 64.18, "elapsed_time": "16:42:15", "remaining_time": "9:19:24"}
+{"current_steps": 3570, "total_steps": 5547, "loss": 0.4278, "lr": 3.3957826277093074e-06, "epoch": 1.930077491439899, "percentage": 64.36, "elapsed_time": "16:45:04", "remaining_time": "9:16:35"}
+{"current_steps": 3580, "total_steps": 5547, "loss": 0.4242, "lr": 3.3660118709097347e-06, "epoch": 1.935483870967742, "percentage": 64.54, "elapsed_time": "16:47:51", "remaining_time": "9:13:45"}
+{"current_steps": 3590, "total_steps": 5547, "loss": 0.4319, "lr": 3.336305828082024e-06, "epoch": 1.9408902504955847, "percentage": 64.72, "elapsed_time": "16:50:42", "remaining_time": "9:10:57"}
+{"current_steps": 3600, "total_steps": 5547, "loss": 0.4324, "lr": 3.306665675731674e-06, "epoch": 1.9462966300234277, "percentage": 64.9, "elapsed_time": "16:53:33", "remaining_time": "9:08:09"}
+{"current_steps": 3610, "total_steps": 5547, "loss": 0.4283, "lr": 3.277092587754598e-06, "epoch": 1.9517030095512706, "percentage": 65.08, "elapsed_time": "16:56:18", "remaining_time": "9:05:19"}
+{"current_steps": 3620, "total_steps": 5547, "loss": 0.4285, "lr": 3.247587735390628e-06, "epoch": 1.9571093890791134, "percentage": 65.26, "elapsed_time": "16:59:07", "remaining_time": "9:02:30"}
+{"current_steps": 3630, "total_steps": 5547, "loss": 0.4233, "lr": 3.218152287177133e-06, "epoch": 1.962515768606956, "percentage": 65.44, "elapsed_time": "17:01:57", "remaining_time": "8:59:41"}
+{"current_steps": 3640, "total_steps": 5547, "loss": 0.4275, "lr": 3.1887874089027304e-06, "epoch": 1.967922148134799, "percentage": 65.62, "elapsed_time": "17:04:43", "remaining_time": "8:56:51"}
+{"current_steps": 3650, "total_steps": 5547, "loss": 0.429, "lr": 3.159494263561126e-06, "epoch": 1.973328527662642, "percentage": 65.8, "elapsed_time": "17:07:30", "remaining_time": "8:54:01"}
+{"current_steps": 3660, "total_steps": 5547, "loss": 0.4261, "lr": 3.130274011305047e-06, "epoch": 1.9787349071904847, "percentage": 65.98, "elapsed_time": "17:10:19", "remaining_time": "8:51:12"}
+{"current_steps": 3670, "total_steps": 5547, "loss": 0.4352, "lr": 3.1011278094002928e-06, "epoch": 1.9841412867183277, "percentage": 66.16, "elapsed_time": "17:13:10", "remaining_time": "8:48:24"}
+{"current_steps": 3680, "total_steps": 5547, "loss": 0.4302, "lr": 3.0720568121799105e-06, "epoch": 1.9895476662461706, "percentage": 66.34, "elapsed_time": "17:16:02", "remaining_time": "8:45:37"}
+{"current_steps": 3690, "total_steps": 5547, "loss": 0.4274, "lr": 3.043062170998464e-06, "epoch": 1.9949540457740134, "percentage": 66.52, "elapsed_time": "17:18:50", "remaining_time": "8:42:47"}
+{"current_steps": 3700, "total_steps": 5547, "loss": 0.4368, "lr": 3.0141450341864486e-06, "epoch": 2.000360425301856, "percentage": 66.7, "elapsed_time": "17:22:57", "remaining_time": "8:40:37"}
+{"current_steps": 3710, "total_steps": 5547, "loss": 0.4084, "lr": 2.9853065470048016e-06, "epoch": 2.005766804829699, "percentage": 66.88, "elapsed_time": "17:25:50", "remaining_time": "8:37:50"}
+{"current_steps": 3720, "total_steps": 5547, "loss": 0.3899, "lr": 2.956547851599548e-06, "epoch": 2.011173184357542, "percentage": 67.06, "elapsed_time": "17:28:36", "remaining_time": "8:35:00"}
+{"current_steps": 3730, "total_steps": 5547, "loss": 0.406, "lr": 2.9278700869565713e-06, "epoch": 2.0165795638853847, "percentage": 67.24, "elapsed_time": "17:31:25", "remaining_time": "8:32:10"}
+{"current_steps": 3740, "total_steps": 5547, "loss": 0.4107, "lr": 2.8992743888564886e-06, "epoch": 2.0219859434132275, "percentage": 67.42, "elapsed_time": "17:34:14", "remaining_time": "8:29:21"}
+{"current_steps": 3750, "total_steps": 5547, "loss": 0.4052, "lr": 2.8707618898296864e-06, "epoch": 2.0273923229410706, "percentage": 67.6, "elapsed_time": "17:37:02", "remaining_time": "8:26:32"}
+{"current_steps": 3760, "total_steps": 5547, "loss": 0.402, "lr": 2.8423337191114495e-06, "epoch": 2.0327987024689134, "percentage": 67.78, "elapsed_time": "17:39:53", "remaining_time": "8:23:43"}
+{"current_steps": 3770, "total_steps": 5547, "loss": 0.4134, "lr": 2.8139910025972622e-06, "epoch": 2.038205081996756, "percentage": 67.96, "elapsed_time": "17:42:46", "remaining_time": "8:20:56"}
+{"current_steps": 3780, "total_steps": 5547, "loss": 0.4086, "lr": 2.785734862798184e-06, "epoch": 2.043611461524599, "percentage": 68.14, "elapsed_time": "17:45:38", "remaining_time": "8:18:08"}
+{"current_steps": 3790, "total_steps": 5547, "loss": 0.4007, "lr": 2.7575664187964236e-06, "epoch": 2.049017841052442, "percentage": 68.33, "elapsed_time": "17:48:30", "remaining_time": "8:15:20"}
+{"current_steps": 3800, "total_steps": 5547, "loss": 0.4053, "lr": 2.7294867862009937e-06, "epoch": 2.0544242205802847, "percentage": 68.51, "elapsed_time": "17:51:14", "remaining_time": "8:12:29"}