Training in progress, step 1500

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef6059da8352e52905470e3b9e7300b13a23926e02ba6acd33645b3f26b2d62c
+oid sha256:ee38e5c4c8458a9a062c756c5a001d3e73ec8606794dd586949e87c044656335
 size 42002584

trainer_log.jsonl

@@ -98,3 +98,53 @@
 {"current_steps": 980, "total_steps": 1875, "loss": 0.3462, "lr": 4.6442536626561675e-05, "epoch": 1.568, "percentage": 52.27, "elapsed_time": "2:19:09", "remaining_time": "2:07:05"}
 {"current_steps": 990, "total_steps": 1875, "loss": 0.3486, "lr": 4.560744017246284e-05, "epoch": 1.584, "percentage": 52.8, "elapsed_time": "2:20:34", "remaining_time": "2:05:39"}
 {"current_steps": 1000, "total_steps": 1875, "loss": 0.3384, "lr": 4.477357683661734e-05, "epoch": 1.6, "percentage": 53.33, "elapsed_time": "2:21:58", "remaining_time": "2:04:13"}
+{"current_steps": 1010, "total_steps": 1875, "loss": 0.3976, "lr": 4.394118070851749e-05, "epoch": 1.616, "percentage": 53.87, "elapsed_time": "2:23:32", "remaining_time": "2:02:55"}
+{"current_steps": 1020, "total_steps": 1875, "loss": 0.3607, "lr": 4.31104854657681e-05, "epoch": 1.6320000000000001, "percentage": 54.4, "elapsed_time": "2:24:55", "remaining_time": "2:01:29"}
+{"current_steps": 1030, "total_steps": 1875, "loss": 0.371, "lr": 4.228172430848644e-05, "epoch": 1.6480000000000001, "percentage": 54.93, "elapsed_time": "2:26:19", "remaining_time": "2:00:02"}
+{"current_steps": 1040, "total_steps": 1875, "loss": 0.3627, "lr": 4.1455129893836174e-05, "epoch": 1.6640000000000001, "percentage": 55.47, "elapsed_time": "2:27:42", "remaining_time": "1:58:35"}
+{"current_steps": 1050, "total_steps": 1875, "loss": 0.3725, "lr": 4.063093427071376e-05, "epoch": 1.6800000000000002, "percentage": 56.0, "elapsed_time": "2:29:06", "remaining_time": "1:57:09"}
+{"current_steps": 1060, "total_steps": 1875, "loss": 0.3466, "lr": 3.9809368814605766e-05, "epoch": 1.696, "percentage": 56.53, "elapsed_time": "2:30:30", "remaining_time": "1:55:42"}
+{"current_steps": 1070, "total_steps": 1875, "loss": 0.3629, "lr": 3.899066416263493e-05, "epoch": 1.712, "percentage": 57.07, "elapsed_time": "2:31:53", "remaining_time": "1:54:16"}
+{"current_steps": 1080, "total_steps": 1875, "loss": 0.3941, "lr": 3.817505014881378e-05, "epoch": 1.728, "percentage": 57.6, "elapsed_time": "2:33:17", "remaining_time": "1:52:50"}
+{"current_steps": 1090, "total_steps": 1875, "loss": 0.3362, "lr": 3.736275573952354e-05, "epoch": 1.744, "percentage": 58.13, "elapsed_time": "2:34:41", "remaining_time": "1:51:24"}
+{"current_steps": 1100, "total_steps": 1875, "loss": 0.3197, "lr": 3.655400896923672e-05, "epoch": 1.76, "percentage": 58.67, "elapsed_time": "2:36:05", "remaining_time": "1:49:58"}
+{"current_steps": 1110, "total_steps": 1875, "loss": 0.3447, "lr": 3.5749036876501194e-05, "epoch": 1.776, "percentage": 59.2, "elapsed_time": "2:37:29", "remaining_time": "1:48:32"}
+{"current_steps": 1120, "total_steps": 1875, "loss": 0.3535, "lr": 3.494806544020398e-05, "epoch": 1.792, "percentage": 59.73, "elapsed_time": "2:38:53", "remaining_time": "1:47:06"}
+{"current_steps": 1130, "total_steps": 1875, "loss": 0.3447, "lr": 3.4151319516132416e-05, "epoch": 1.808, "percentage": 60.27, "elapsed_time": "2:40:16", "remaining_time": "1:45:40"}
+{"current_steps": 1140, "total_steps": 1875, "loss": 0.338, "lr": 3.335902277385067e-05, "epoch": 1.8239999999999998, "percentage": 60.8, "elapsed_time": "2:41:40", "remaining_time": "1:44:14"}
+{"current_steps": 1150, "total_steps": 1875, "loss": 0.3915, "lr": 3.257139763390925e-05, "epoch": 1.8399999999999999, "percentage": 61.33, "elapsed_time": "2:43:04", "remaining_time": "1:42:48"}
+{"current_steps": 1160, "total_steps": 1875, "loss": 0.332, "lr": 3.178866520540509e-05, "epoch": 1.8559999999999999, "percentage": 61.87, "elapsed_time": "2:44:28", "remaining_time": "1:41:22"}
+{"current_steps": 1170, "total_steps": 1875, "loss": 0.3734, "lr": 3.101104522390995e-05, "epoch": 1.8719999999999999, "percentage": 62.4, "elapsed_time": "2:45:52", "remaining_time": "1:39:56"}
+{"current_steps": 1180, "total_steps": 1875, "loss": 0.3409, "lr": 3.023875598978419e-05, "epoch": 1.888, "percentage": 62.93, "elapsed_time": "2:47:16", "remaining_time": "1:38:31"}
+{"current_steps": 1190, "total_steps": 1875, "loss": 0.3333, "lr": 2.9472014306893603e-05, "epoch": 1.904, "percentage": 63.47, "elapsed_time": "2:48:40", "remaining_time": "1:37:05"}
+{"current_steps": 1200, "total_steps": 1875, "loss": 0.369, "lr": 2.8711035421746367e-05, "epoch": 1.92, "percentage": 64.0, "elapsed_time": "2:50:04", "remaining_time": "1:35:39"}
+{"current_steps": 1210, "total_steps": 1875, "loss": 0.3449, "lr": 2.795603296306708e-05, "epoch": 1.936, "percentage": 64.53, "elapsed_time": "2:51:28", "remaining_time": "1:34:14"}
+{"current_steps": 1220, "total_steps": 1875, "loss": 0.3662, "lr": 2.7207218881825014e-05, "epoch": 1.952, "percentage": 65.07, "elapsed_time": "2:52:52", "remaining_time": "1:32:48"}
+{"current_steps": 1230, "total_steps": 1875, "loss": 0.3773, "lr": 2.6464803391733374e-05, "epoch": 1.968, "percentage": 65.6, "elapsed_time": "2:54:16", "remaining_time": "1:31:23"}
+{"current_steps": 1240, "total_steps": 1875, "loss": 0.3245, "lr": 2.5728994910236304e-05, "epoch": 1.984, "percentage": 66.13, "elapsed_time": "2:55:40", "remaining_time": "1:29:57"}
+{"current_steps": 1250, "total_steps": 1875, "loss": 0.3412, "lr": 2.500000000000001e-05, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "2:57:04", "remaining_time": "1:28:32"}
+{"current_steps": 1260, "total_steps": 1875, "loss": 0.3027, "lr": 2.4278023310924673e-05, "epoch": 2.016, "percentage": 67.2, "elapsed_time": "2:58:28", "remaining_time": "1:27:06"}
+{"current_steps": 1270, "total_steps": 1875, "loss": 0.2745, "lr": 2.3563267522693415e-05, "epoch": 2.032, "percentage": 67.73, "elapsed_time": "2:59:52", "remaining_time": "1:25:41"}
+{"current_steps": 1280, "total_steps": 1875, "loss": 0.2686, "lr": 2.2855933287874138e-05, "epoch": 2.048, "percentage": 68.27, "elapsed_time": "3:01:16", "remaining_time": "1:24:15"}
+{"current_steps": 1290, "total_steps": 1875, "loss": 0.2628, "lr": 2.215621917559062e-05, "epoch": 2.064, "percentage": 68.8, "elapsed_time": "3:02:40", "remaining_time": "1:22:50"}
+{"current_steps": 1300, "total_steps": 1875, "loss": 0.274, "lr": 2.1464321615778422e-05, "epoch": 2.08, "percentage": 69.33, "elapsed_time": "3:04:04", "remaining_time": "1:21:25"}
+{"current_steps": 1310, "total_steps": 1875, "loss": 0.2711, "lr": 2.07804348440414e-05, "epoch": 2.096, "percentage": 69.87, "elapsed_time": "3:05:28", "remaining_time": "1:19:59"}
+{"current_steps": 1320, "total_steps": 1875, "loss": 0.2662, "lr": 2.0104750847124075e-05, "epoch": 2.112, "percentage": 70.4, "elapsed_time": "3:06:52", "remaining_time": "1:18:34"}
+{"current_steps": 1330, "total_steps": 1875, "loss": 0.2722, "lr": 1.9437459309015427e-05, "epoch": 2.128, "percentage": 70.93, "elapsed_time": "3:08:16", "remaining_time": "1:17:08"}
+{"current_steps": 1340, "total_steps": 1875, "loss": 0.2965, "lr": 1.8778747557699224e-05, "epoch": 2.144, "percentage": 71.47, "elapsed_time": "3:09:40", "remaining_time": "1:15:43"}
+{"current_steps": 1350, "total_steps": 1875, "loss": 0.2789, "lr": 1.8128800512565513e-05, "epoch": 2.16, "percentage": 72.0, "elapsed_time": "3:11:04", "remaining_time": "1:14:18"}
+{"current_steps": 1360, "total_steps": 1875, "loss": 0.2727, "lr": 1.7487800632498545e-05, "epoch": 2.176, "percentage": 72.53, "elapsed_time": "3:12:27", "remaining_time": "1:12:52"}
+{"current_steps": 1370, "total_steps": 1875, "loss": 0.2579, "lr": 1.685592786465524e-05, "epoch": 2.192, "percentage": 73.07, "elapsed_time": "3:13:51", "remaining_time": "1:11:27"}
+{"current_steps": 1380, "total_steps": 1875, "loss": 0.2639, "lr": 1.6233359593948777e-05, "epoch": 2.208, "percentage": 73.6, "elapsed_time": "3:15:15", "remaining_time": "1:10:02"}
+{"current_steps": 1390, "total_steps": 1875, "loss": 0.2909, "lr": 1.5620270593251635e-05, "epoch": 2.224, "percentage": 74.13, "elapsed_time": "3:16:39", "remaining_time": "1:08:37"}
+{"current_steps": 1400, "total_steps": 1875, "loss": 0.2675, "lr": 1.5016832974331724e-05, "epoch": 2.24, "percentage": 74.67, "elapsed_time": "3:18:03", "remaining_time": "1:07:11"}
+{"current_steps": 1410, "total_steps": 1875, "loss": 0.2501, "lr": 1.4423216139535734e-05, "epoch": 2.2560000000000002, "percentage": 75.2, "elapsed_time": "3:19:26", "remaining_time": "1:05:46"}
+{"current_steps": 1420, "total_steps": 1875, "loss": 0.28, "lr": 1.3839586734232906e-05, "epoch": 2.2720000000000002, "percentage": 75.73, "elapsed_time": "3:20:50", "remaining_time": "1:04:21"}
+{"current_steps": 1430, "total_steps": 1875, "loss": 0.2946, "lr": 1.3266108600032929e-05, "epoch": 2.288, "percentage": 76.27, "elapsed_time": "3:22:14", "remaining_time": "1:02:56"}
+{"current_steps": 1440, "total_steps": 1875, "loss": 0.2801, "lr": 1.2702942728790895e-05, "epoch": 2.304, "percentage": 76.8, "elapsed_time": "3:23:38", "remaining_time": "1:01:30"}
+{"current_steps": 1450, "total_steps": 1875, "loss": 0.2686, "lr": 1.2150247217412186e-05, "epoch": 2.32, "percentage": 77.33, "elapsed_time": "3:25:02", "remaining_time": "1:00:05"}
+{"current_steps": 1460, "total_steps": 1875, "loss": 0.2623, "lr": 1.160817722347014e-05, "epoch": 2.336, "percentage": 77.87, "elapsed_time": "3:26:26", "remaining_time": "0:58:40"}
+{"current_steps": 1470, "total_steps": 1875, "loss": 0.2865, "lr": 1.1076884921648834e-05, "epoch": 2.352, "percentage": 78.4, "elapsed_time": "3:27:50", "remaining_time": "0:57:15"}
+{"current_steps": 1480, "total_steps": 1875, "loss": 0.2677, "lr": 1.0556519461023301e-05, "epoch": 2.368, "percentage": 78.93, "elapsed_time": "3:29:14", "remaining_time": "0:55:50"}
+{"current_steps": 1490, "total_steps": 1875, "loss": 0.2821, "lr": 1.0047226923189024e-05, "epoch": 2.384, "percentage": 79.47, "elapsed_time": "3:30:38", "remaining_time": "0:54:25"}
+{"current_steps": 1500, "total_steps": 1875, "loss": 0.2628, "lr": 9.549150281252633e-06, "epoch": 2.4, "percentage": 80.0, "elapsed_time": "3:32:02", "remaining_time": "0:53:00"}