Training in progress, epoch 2

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:11a543f3ef469e820b7f5c5e973ce740968f29e9da203c1902e03c0696d5c73f
+oid sha256:e53da8e9270c0c84f099b2f6f265e7b3e4ff83709135674b6a4b318e7c29c00a
 size 4903351912

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7919a5eb31df82f7aa8d12b18d2f09cc6f260c4b302d13a07c91965573d67743
+oid sha256:67222d91ba19923a54dc0185ada5efa8c12d3ff656f412ae0a65b3df6bb75894
 size 4947570872

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:87a1edc48bd1c06e3e660aeb048d7b1a8b13f6a957d2edd711f5153c719ca1aa
+oid sha256:64ea9f31b89be3730de8c2ccdda74ca8e3d0b91f6af9167a68e5964e76403d3f
 size 4962221464

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d86237a6c3db73d5abd99aa7404831f17b78a18b110f923cc146947320269328
+oid sha256:0936a29378e79e04ce0589439ba68897bb3449683fe0d10c94e700edda66ea9a
 size 3670322200

trainer_log.jsonl

@@ -88,3 +88,47 @@
 {"current_steps": 870, "total_steps": 1329, "loss": 0.5438, "learning_rate": 1.815779420709386e-06, "epoch": 1.9633286318758816, "percentage": 65.46, "elapsed_time": "21:32:55", "remaining_time": "11:22:07"}
 {"current_steps": 880, "total_steps": 1329, "loss": 0.5415, "learning_rate": 1.7651195083347916e-06, "epoch": 1.9858956276445698, "percentage": 66.22, "elapsed_time": "21:47:39", "remaining_time": "11:07:11"}
 {"current_steps": 886, "total_steps": 1329, "eval_loss": 0.5835571885108948, "epoch": 1.9994358251057829, "percentage": 66.67, "elapsed_time": "22:08:15", "remaining_time": "11:04:07"}
+{"current_steps": 890, "total_steps": 1329, "loss": 0.5719, "learning_rate": 1.7150698942358537e-06, "epoch": 2.008462623413258, "percentage": 66.97, "elapsed_time": "22:15:08", "remaining_time": "10:58:34"}
+{"current_steps": 900, "total_steps": 1329, "loss": 0.4955, "learning_rate": 1.6656615925238655e-06, "epoch": 2.0310296191819464, "percentage": 67.72, "elapsed_time": "22:29:51", "remaining_time": "10:43:25"}
+{"current_steps": 910, "total_steps": 1329, "loss": 0.498, "learning_rate": 1.616925219909782e-06, "epoch": 2.0535966149506346, "percentage": 68.47, "elapsed_time": "22:44:33", "remaining_time": "10:28:17"}
+{"current_steps": 920, "total_steps": 1329, "loss": 0.4901, "learning_rate": 1.568890976732041e-06, "epoch": 2.076163610719323, "percentage": 69.22, "elapsed_time": "22:59:16", "remaining_time": "10:13:10"}
+{"current_steps": 930, "total_steps": 1329, "loss": 0.497, "learning_rate": 1.5215886282423984e-06, "epoch": 2.098730606488011, "percentage": 69.98, "elapsed_time": "23:13:58", "remaining_time": "9:58:03"}
+{"current_steps": 940, "total_steps": 1329, "loss": 0.5006, "learning_rate": 1.4750474861613729e-06, "epoch": 2.1212976022566994, "percentage": 70.73, "elapsed_time": "23:28:41", "remaining_time": "9:42:57"}
+{"current_steps": 950, "total_steps": 1329, "loss": 0.5016, "learning_rate": 1.4292963905147344e-06, "epoch": 2.143864598025388, "percentage": 71.48, "elapsed_time": "23:43:24", "remaining_time": "9:27:51"}
+{"current_steps": 960, "total_steps": 1329, "loss": 0.4977, "learning_rate": 1.3843636917622792e-06, "epoch": 2.1664315937940763, "percentage": 72.23, "elapsed_time": "23:58:06", "remaining_time": "9:12:46"}
+{"current_steps": 970, "total_steps": 1329, "loss": 0.4945, "learning_rate": 1.3402772332299752e-06, "epoch": 2.1889985895627646, "percentage": 72.99, "elapsed_time": "1 day, 0:12:48", "remaining_time": "8:57:41"}
+{"current_steps": 980, "total_steps": 1329, "loss": 0.4889, "learning_rate": 1.2970643338563623e-06, "epoch": 2.211565585331453, "percentage": 73.74, "elapsed_time": "1 day, 0:27:30", "remaining_time": "8:42:36"}
+{"current_steps": 990, "total_steps": 1329, "loss": 0.4886, "learning_rate": 1.2547517712639013e-06, "epoch": 2.234132581100141, "percentage": 74.49, "elapsed_time": "1 day, 0:42:13", "remaining_time": "8:27:32"}
+{"current_steps": 1000, "total_steps": 1329, "loss": 0.4978, "learning_rate": 1.2133657651657524e-06, "epoch": 2.2566995768688294, "percentage": 75.24, "elapsed_time": "1 day, 0:56:55", "remaining_time": "8:12:29"}
+{"current_steps": 1010, "total_steps": 1329, "loss": 0.5033, "learning_rate": 1.1729319611182772e-06, "epoch": 2.2792665726375176, "percentage": 76.0, "elapsed_time": "1 day, 1:11:37", "remaining_time": "7:57:26"}
+{"current_steps": 1020, "total_steps": 1329, "loss": 0.4972, "learning_rate": 1.1334754146293222e-06, "epoch": 2.301833568406206, "percentage": 76.75, "elapsed_time": "1 day, 1:26:20", "remaining_time": "7:42:23"}
+{"current_steps": 1030, "total_steps": 1329, "loss": 0.5012, "learning_rate": 1.095020575632144e-06, "epoch": 2.324400564174894, "percentage": 77.5, "elapsed_time": "1 day, 1:41:02", "remaining_time": "7:27:21"}
+{"current_steps": 1040, "total_steps": 1329, "loss": 0.4985, "learning_rate": 1.057591273334579e-06, "epoch": 2.3469675599435824, "percentage": 78.25, "elapsed_time": "1 day, 1:55:44", "remaining_time": "7:12:18"}
+{"current_steps": 1050, "total_steps": 1329, "loss": 0.4958, "learning_rate": 1.0212107014528651e-06, "epoch": 2.3695345557122707, "percentage": 79.01, "elapsed_time": "1 day, 2:10:26", "remaining_time": "6:57:17"}
+{"current_steps": 1060, "total_steps": 1329, "loss": 0.4949, "learning_rate": 9.859014038392498e-07, "epoch": 2.392101551480959, "percentage": 79.76, "elapsed_time": "1 day, 2:25:09", "remaining_time": "6:42:16"}
+{"current_steps": 1070, "total_steps": 1329, "loss": 0.4951, "learning_rate": 9.516852605123054e-07, "epoch": 2.414668547249647, "percentage": 80.51, "elapsed_time": "1 day, 2:39:51", "remaining_time": "6:27:15"}
+{"current_steps": 1080, "total_steps": 1329, "loss": 0.4987, "learning_rate": 9.185834740985922e-07, "epoch": 2.4372355430183354, "percentage": 81.26, "elapsed_time": "1 day, 2:54:34", "remaining_time": "6:12:14"}
+{"current_steps": 1090, "total_steps": 1329, "loss": 0.4967, "learning_rate": 8.866165566940863e-07, "epoch": 2.459802538787024, "percentage": 82.02, "elapsed_time": "1 day, 3:09:17", "remaining_time": "5:57:14"}
+{"current_steps": 1100, "total_steps": 1329, "loss": 0.4992, "learning_rate": 8.55804317153501e-07, "epoch": 2.4823695345557124, "percentage": 82.77, "elapsed_time": "1 day, 3:24:00", "remaining_time": "5:42:15"}
+{"current_steps": 1110, "total_steps": 1329, "loss": 0.5022, "learning_rate": 8.26165848815389e-07, "epoch": 2.5049365303244007, "percentage": 83.52, "elapsed_time": "1 day, 3:38:42", "remaining_time": "5:27:15"}
+{"current_steps": 1120, "total_steps": 1329, "loss": 0.4921, "learning_rate": 7.977195176706218e-07, "epoch": 2.527503526093089, "percentage": 84.27, "elapsed_time": "1 day, 3:53:25", "remaining_time": "5:12:16"}
+{"current_steps": 1130, "total_steps": 1329, "loss": 0.497, "learning_rate": 7.704829509815869e-07, "epoch": 2.550070521861777, "percentage": 85.03, "elapsed_time": "1 day, 4:08:07", "remaining_time": "4:57:17"}
+{"current_steps": 1140, "total_steps": 1329, "loss": 0.4951, "learning_rate": 7.444730263591466e-07, "epoch": 2.5726375176304654, "percentage": 85.78, "elapsed_time": "1 day, 4:22:50", "remaining_time": "4:42:18"}
+{"current_steps": 1150, "total_steps": 1329, "loss": 0.4955, "learning_rate": 7.197058613041364e-07, "epoch": 2.5952045133991537, "percentage": 86.53, "elapsed_time": "1 day, 4:37:32", "remaining_time": "4:27:20"}
+{"current_steps": 1160, "total_steps": 1329, "loss": 0.4989, "learning_rate": 6.961968032198748e-07, "epoch": 2.617771509167842, "percentage": 87.28, "elapsed_time": "1 day, 4:52:14", "remaining_time": "4:12:22"}
+{"current_steps": 1170, "total_steps": 1329, "loss": 0.4875, "learning_rate": 6.739604199018791e-07, "epoch": 2.64033850493653, "percentage": 88.04, "elapsed_time": "1 day, 5:06:57", "remaining_time": "3:57:24"}
+{"current_steps": 1180, "total_steps": 1329, "loss": 0.4985, "learning_rate": 6.530104905106762e-07, "epoch": 2.6629055007052185, "percentage": 88.79, "elapsed_time": "1 day, 5:21:39", "remaining_time": "3:42:26"}
+{"current_steps": 1190, "total_steps": 1329, "loss": 0.4991, "learning_rate": 6.333599970333109e-07, "epoch": 2.685472496473907, "percentage": 89.54, "elapsed_time": "1 day, 5:36:22", "remaining_time": "3:27:29"}
+{"current_steps": 1200, "total_steps": 1329, "loss": 0.4974, "learning_rate": 6.150211162388293e-07, "epoch": 2.7080394922425954, "percentage": 90.29, "elapsed_time": "1 day, 5:51:03", "remaining_time": "3:12:32"}
+{"current_steps": 1210, "total_steps": 1329, "loss": 0.498, "learning_rate": 5.980052121327338e-07, "epoch": 2.7306064880112837, "percentage": 91.05, "elapsed_time": "1 day, 6:05:46", "remaining_time": "2:57:35"}
+{"current_steps": 1220, "total_steps": 1329, "loss": 0.4945, "learning_rate": 5.823228289150787e-07, "epoch": 2.753173483779972, "percentage": 91.8, "elapsed_time": "1 day, 6:20:29", "remaining_time": "2:42:38"}
+{"current_steps": 1230, "total_steps": 1329, "loss": 0.5013, "learning_rate": 5.679836844465767e-07, "epoch": 2.77574047954866, "percentage": 92.55, "elapsed_time": "1 day, 6:35:11", "remaining_time": "2:27:42"}
+{"current_steps": 1240, "total_steps": 1329, "loss": 0.4996, "learning_rate": 5.549966642267552e-07, "epoch": 2.7983074753173485, "percentage": 93.3, "elapsed_time": "1 day, 6:49:55", "remaining_time": "2:12:46"}
+{"current_steps": 1250, "total_steps": 1329, "loss": 0.4992, "learning_rate": 5.433698158879043e-07, "epoch": 2.8208744710860367, "percentage": 94.06, "elapsed_time": "1 day, 7:04:37", "remaining_time": "1:57:50"}
+{"current_steps": 1260, "total_steps": 1329, "loss": 0.5009, "learning_rate": 5.331103442082198e-07, "epoch": 2.843441466854725, "percentage": 94.81, "elapsed_time": "1 day, 7:19:20", "remaining_time": "1:42:54"}
+{"current_steps": 1270, "total_steps": 1329, "loss": 0.498, "learning_rate": 5.242246066472395e-07, "epoch": 2.8660084626234132, "percentage": 95.56, "elapsed_time": "1 day, 7:34:02", "remaining_time": "1:27:59"}
+{"current_steps": 1280, "total_steps": 1329, "loss": 0.4955, "learning_rate": 5.167181094063323e-07, "epoch": 2.8885754583921015, "percentage": 96.31, "elapsed_time": "1 day, 7:48:44", "remaining_time": "1:13:04"}
+{"current_steps": 1290, "total_steps": 1329, "loss": 0.4994, "learning_rate": 5.105955040166854e-07, "epoch": 2.9111424541607898, "percentage": 97.07, "elapsed_time": "1 day, 8:03:26", "remaining_time": "0:58:09"}
+{"current_steps": 1300, "total_steps": 1329, "loss": 0.4945, "learning_rate": 5.058605844569015e-07, "epoch": 2.933709449929478, "percentage": 97.82, "elapsed_time": "1 day, 8:18:09", "remaining_time": "0:43:14"}
+{"current_steps": 1310, "total_steps": 1329, "loss": 0.4985, "learning_rate": 5.025162848019948e-07, "epoch": 2.9562764456981663, "percentage": 98.57, "elapsed_time": "1 day, 8:32:51", "remaining_time": "0:28:19"}
+{"current_steps": 1320, "total_steps": 1329, "loss": 0.4941, "learning_rate": 5.00564677405241e-07, "epoch": 2.9788434414668545, "percentage": 99.32, "elapsed_time": "1 day, 8:47:33", "remaining_time": "0:13:24"}