Training in progress, epoch 4

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8cd1d9f0935b91513febb715de1ad7e22b50d1f833dbe0e604201469611c3371
+oid sha256:26279f811ebbac8b2395390313fa2137c1780e424aa6fcdf5f9a804f89d8d60a
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38a13535959b62a64d63ac639cb5f27423771ce69e15a7e6f10b8cff1368e2df
+oid sha256:95160f151b32e43da2afe5e62671454eb229b9a5f408f5a63655753fefcb7622
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06b76917d811d382d8b31c2cbffa44ad1a72975431d40a03fb52aa6f319d1847
+oid sha256:e5792050c871b38b8cfddb1cb73f15ebb5b565bcd472dd2eb19fb92e16754eb5
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d45167bd69aa4fa4574ab48c92fbed462378acad8b1a45144a90e29ada4b83fb
+oid sha256:89f5142bc1e1827e6e8e589b739e6dc6cd5daa55deeced4fe94ff594b50ff460
 size 1089994880

trainer_log.jsonl

@@ -129,3 +129,28 @@
 {"current_steps": 129, "total_steps": 155, "loss": 0.1525, "lr": 1.6774642643563955e-06, "epoch": 4.128, "percentage": 83.23, "elapsed_time": "2:34:26", "remaining_time": "0:31:07"}
 {"current_steps": 130, "total_steps": 155, "loss": 0.1645, "lr": 1.5542995220217961e-06, "epoch": 4.16, "percentage": 83.87, "elapsed_time": "2:34:51", "remaining_time": "0:29:46"}
 {"current_steps": 131, "total_steps": 155, "loss": 0.1762, "lr": 1.4354488511294418e-06, "epoch": 4.192, "percentage": 84.52, "elapsed_time": "2:35:20", "remaining_time": "0:28:27"}
+{"current_steps": 132, "total_steps": 155, "loss": 0.1822, "lr": 1.3209729607095022e-06, "epoch": 4.224, "percentage": 85.16, "elapsed_time": "2:35:57", "remaining_time": "0:27:10"}
+{"current_steps": 133, "total_steps": 155, "loss": 0.2034, "lr": 1.2109303251503434e-06, "epoch": 4.256, "percentage": 85.81, "elapsed_time": "2:36:28", "remaining_time": "0:25:53"}
+{"current_steps": 134, "total_steps": 155, "loss": 0.1617, "lr": 1.1053771543297198e-06, "epoch": 4.288, "percentage": 86.45, "elapsed_time": "2:37:05", "remaining_time": "0:24:37"}
+{"current_steps": 135, "total_steps": 155, "loss": 0.1662, "lr": 1.0043673649027519e-06, "epoch": 4.32, "percentage": 87.1, "elapsed_time": "2:37:36", "remaining_time": "0:23:20"}
+{"current_steps": 136, "total_steps": 155, "loss": 0.1981, "lr": 9.079525527612321e-07, "epoch": 4.352, "percentage": 87.74, "elapsed_time": "2:38:03", "remaining_time": "0:22:04"}
+{"current_steps": 137, "total_steps": 155, "loss": 0.1437, "lr": 8.161819666783888e-07, "epoch": 4.384, "percentage": 88.39, "elapsed_time": "2:38:27", "remaining_time": "0:20:49"}
+{"current_steps": 138, "total_steps": 155, "loss": 0.1109, "lr": 7.291024831525961e-07, "epoch": 4.416, "percentage": 89.03, "elapsed_time": "2:38:54", "remaining_time": "0:19:34"}
+{"current_steps": 139, "total_steps": 155, "loss": 0.1161, "lr": 6.467585824627886e-07, "epoch": 4.448, "percentage": 89.68, "elapsed_time": "2:39:29", "remaining_time": "0:18:21"}
+{"current_steps": 140, "total_steps": 155, "loss": 0.1948, "lr": 5.691923259479093e-07, "epoch": 4.48, "percentage": 90.32, "elapsed_time": "2:40:07", "remaining_time": "0:17:09"}
+{"current_steps": 141, "total_steps": 155, "loss": 0.193, "lr": 4.964433345219354e-07, "epoch": 4.5120000000000005, "percentage": 90.97, "elapsed_time": "2:40:37", "remaining_time": "0:15:56"}
+{"current_steps": 142, "total_steps": 155, "loss": 0.1182, "lr": 4.285487684354772e-07, "epoch": 4.5440000000000005, "percentage": 91.61, "elapsed_time": "2:41:00", "remaining_time": "0:14:44"}
+{"current_steps": 143, "total_steps": 155, "loss": 0.1588, "lr": 3.6554330829429716e-07, "epoch": 4.576, "percentage": 92.26, "elapsed_time": "2:41:33", "remaining_time": "0:13:33"}
+{"current_steps": 144, "total_steps": 155, "loss": 0.1395, "lr": 3.0745913734441357e-07, "epoch": 4.608, "percentage": 92.9, "elapsed_time": "2:42:02", "remaining_time": "0:12:22"}
+{"current_steps": 145, "total_steps": 155, "loss": 0.1719, "lr": 2.5432592503288e-07, "epoch": 4.64, "percentage": 93.55, "elapsed_time": "2:42:34", "remaining_time": "0:11:12"}
+{"current_steps": 146, "total_steps": 155, "loss": 0.1248, "lr": 2.0617081185259512e-07, "epoch": 4.672, "percentage": 94.19, "elapsed_time": "2:42:59", "remaining_time": "0:10:02"}
+{"current_steps": 147, "total_steps": 155, "loss": 0.1756, "lr": 1.630183954789233e-07, "epoch": 4.704, "percentage": 94.84, "elapsed_time": "2:43:30", "remaining_time": "0:08:53"}
+{"current_steps": 148, "total_steps": 155, "loss": 0.1914, "lr": 1.2489071820517394e-07, "epoch": 4.736, "percentage": 95.48, "elapsed_time": "2:43:58", "remaining_time": "0:07:45"}
+{"current_steps": 149, "total_steps": 155, "loss": 0.1829, "lr": 9.180725568338045e-08, "epoch": 4.768, "percentage": 96.13, "elapsed_time": "2:44:26", "remaining_time": "0:06:37"}
+{"current_steps": 150, "total_steps": 155, "loss": 0.1593, "lr": 6.378490697611761e-08, "epoch": 4.8, "percentage": 96.77, "elapsed_time": "2:45:01", "remaining_time": "0:05:30"}
+{"current_steps": 151, "total_steps": 155, "loss": 0.188, "lr": 4.083798592444899e-08, "epoch": 4.832, "percentage": 97.42, "elapsed_time": "2:45:43", "remaining_time": "0:04:23"}
+{"current_steps": 152, "total_steps": 155, "loss": 0.1141, "lr": 2.2978213836400974e-08, "epoch": 4.864, "percentage": 98.06, "elapsed_time": "2:46:05", "remaining_time": "0:03:16"}
+{"current_steps": 153, "total_steps": 155, "loss": 0.1586, "lr": 1.0214713499706596e-08, "epoch": 4.896, "percentage": 98.71, "elapsed_time": "2:46:43", "remaining_time": "0:02:10"}
+{"current_steps": 154, "total_steps": 155, "loss": 0.1662, "lr": 2.5540045218819256e-09, "epoch": 4.928, "percentage": 99.35, "elapsed_time": "2:47:11", "remaining_time": "0:01:05"}
+{"current_steps": 155, "total_steps": 155, "loss": 0.2068, "lr": 0.0, "epoch": 4.96, "percentage": 100.0, "elapsed_time": "2:47:43", "remaining_time": "0:00:00"}
+{"current_steps": 155, "total_steps": 155, "epoch": 4.96, "percentage": 100.0, "elapsed_time": "3:58:09", "remaining_time": "0:00:00"}