Training in progress, step 780

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:99f5afef6dce3d57b41997d6aae56cca739719f87667fb749df980a95813fa08
+oid sha256:8b568f653304bc315cc5c9fd8441398c88b4abfb024c0b55b1609a64aaa1d25b
 size 4976698672

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f253d23fbb4e033c1291e5ba481eec54044fa728be86c57c548a46fa03d085ce
+oid sha256:c5e13e88ebabfbc3e654cc88086f4db906ca8559ac06de4d6b0f461478a78cc0
 size 4999802720

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8bd835c474434b500901174637d19f51e71525fb4bdfbbc02374341b6f78dce3
+oid sha256:dc02832f1314c19e82cf74a261326c8520c2878274f13b9548b573918aa4513e
 size 4915916176

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b587e71b6fb7127e1cc92f052865b066f6ebddabe0094b3b45ff8aa35415b32a
+oid sha256:af974431558d8f149f8b83556a6a14eb4aced72c2290ca5eeab0de6c7d882728
 size 1168138808

trainer_log.jsonl

@@ -51,3 +51,29 @@
 {"current_steps": 510, "total_steps": 780, "loss": 0.0176, "lr": 3.860398860398861e-06, "epoch": 3.2496, "percentage": 65.38, "elapsed_time": "6:35:22", "remaining_time": "3:29:19"}
 {"current_steps": 520, "total_steps": 780, "loss": 0.0214, "lr": 3.7179487179487184e-06, "epoch": 3.3136, "percentage": 66.67, "elapsed_time": "6:43:01", "remaining_time": "3:21:30"}
 {"current_steps": 530, "total_steps": 780, "loss": 0.0283, "lr": 3.5754985754985762e-06, "epoch": 3.3776, "percentage": 67.95, "elapsed_time": "6:50:41", "remaining_time": "3:13:43"}
+{"current_steps": 540, "total_steps": 780, "loss": 0.0272, "lr": 3.433048433048433e-06, "epoch": 3.4416, "percentage": 69.23, "elapsed_time": "6:58:20", "remaining_time": "3:05:55"}
+{"current_steps": 550, "total_steps": 780, "loss": 0.0191, "lr": 3.290598290598291e-06, "epoch": 3.5056000000000003, "percentage": 70.51, "elapsed_time": "7:05:59", "remaining_time": "2:58:08"}
+{"current_steps": 560, "total_steps": 780, "loss": 0.0267, "lr": 3.1481481481481483e-06, "epoch": 3.5696, "percentage": 71.79, "elapsed_time": "7:13:38", "remaining_time": "2:50:21"}
+{"current_steps": 570, "total_steps": 780, "loss": 0.0397, "lr": 3.005698005698006e-06, "epoch": 3.6336, "percentage": 73.08, "elapsed_time": "7:21:17", "remaining_time": "2:42:34"}
+{"current_steps": 580, "total_steps": 780, "loss": 0.0271, "lr": 2.8632478632478635e-06, "epoch": 3.6976, "percentage": 74.36, "elapsed_time": "7:28:57", "remaining_time": "2:34:48"}
+{"current_steps": 590, "total_steps": 780, "loss": 0.0243, "lr": 2.720797720797721e-06, "epoch": 3.7616, "percentage": 75.64, "elapsed_time": "7:36:37", "remaining_time": "2:27:02"}
+{"current_steps": 600, "total_steps": 780, "loss": 0.0235, "lr": 2.5783475783475787e-06, "epoch": 3.8256, "percentage": 76.92, "elapsed_time": "7:44:17", "remaining_time": "2:19:17"}
+{"current_steps": 610, "total_steps": 780, "loss": 0.0249, "lr": 2.435897435897436e-06, "epoch": 3.8895999999999997, "percentage": 78.21, "elapsed_time": "7:51:57", "remaining_time": "2:11:31"}
+{"current_steps": 620, "total_steps": 780, "loss": 0.0186, "lr": 2.293447293447294e-06, "epoch": 3.9536, "percentage": 79.49, "elapsed_time": "7:59:36", "remaining_time": "2:03:46"}
+{"current_steps": 630, "total_steps": 780, "loss": 0.0167, "lr": 2.150997150997151e-06, "epoch": 4.0128, "percentage": 80.77, "elapsed_time": "8:06:42", "remaining_time": "1:55:53"}
+{"current_steps": 640, "total_steps": 780, "loss": 0.0159, "lr": 2.008547008547009e-06, "epoch": 4.0768, "percentage": 82.05, "elapsed_time": "8:14:22", "remaining_time": "1:48:08"}
+{"current_steps": 650, "total_steps": 780, "loss": 0.0183, "lr": 1.8660968660968661e-06, "epoch": 4.1408, "percentage": 83.33, "elapsed_time": "8:22:02", "remaining_time": "1:40:24"}
+{"current_steps": 660, "total_steps": 780, "loss": 0.0137, "lr": 1.723646723646724e-06, "epoch": 4.2048, "percentage": 84.62, "elapsed_time": "8:29:42", "remaining_time": "1:32:40"}
+{"current_steps": 670, "total_steps": 780, "loss": 0.0176, "lr": 1.5811965811965813e-06, "epoch": 4.2688, "percentage": 85.9, "elapsed_time": "8:37:22", "remaining_time": "1:24:56"}
+{"current_steps": 680, "total_steps": 780, "loss": 0.0142, "lr": 1.4387464387464389e-06, "epoch": 4.3328, "percentage": 87.18, "elapsed_time": "8:45:00", "remaining_time": "1:17:12"}
+{"current_steps": 690, "total_steps": 780, "loss": 0.0185, "lr": 1.2962962962962962e-06, "epoch": 4.3968, "percentage": 88.46, "elapsed_time": "8:52:40", "remaining_time": "1:09:28"}
+{"current_steps": 700, "total_steps": 780, "loss": 0.0097, "lr": 1.153846153846154e-06, "epoch": 4.4608, "percentage": 89.74, "elapsed_time": "9:00:20", "remaining_time": "1:01:45"}
+{"current_steps": 710, "total_steps": 780, "loss": 0.0104, "lr": 1.0113960113960116e-06, "epoch": 4.5248, "percentage": 91.03, "elapsed_time": "9:07:59", "remaining_time": "0:54:01"}
+{"current_steps": 720, "total_steps": 780, "loss": 0.0163, "lr": 8.689458689458691e-07, "epoch": 4.5888, "percentage": 92.31, "elapsed_time": "9:15:38", "remaining_time": "0:46:18"}
+{"current_steps": 730, "total_steps": 780, "loss": 0.0151, "lr": 7.264957264957266e-07, "epoch": 4.6528, "percentage": 93.59, "elapsed_time": "9:23:17", "remaining_time": "0:38:34"}
+{"current_steps": 740, "total_steps": 780, "loss": 0.0132, "lr": 5.84045584045584e-07, "epoch": 4.7168, "percentage": 94.87, "elapsed_time": "9:30:56", "remaining_time": "0:30:51"}
+{"current_steps": 750, "total_steps": 780, "loss": 0.012, "lr": 4.415954415954416e-07, "epoch": 4.7808, "percentage": 96.15, "elapsed_time": "9:38:35", "remaining_time": "0:23:08"}
+{"current_steps": 760, "total_steps": 780, "loss": 0.0152, "lr": 2.991452991452992e-07, "epoch": 4.8448, "percentage": 97.44, "elapsed_time": "9:46:13", "remaining_time": "0:15:25"}
+{"current_steps": 770, "total_steps": 780, "loss": 0.0109, "lr": 1.566951566951567e-07, "epoch": 4.9088, "percentage": 98.72, "elapsed_time": "9:53:53", "remaining_time": "0:07:42"}
+{"current_steps": 780, "total_steps": 780, "loss": 0.0113, "lr": 1.4245014245014247e-08, "epoch": 4.9728, "percentage": 100.0, "elapsed_time": "10:01:33", "remaining_time": "0:00:00"}
+{"current_steps": 780, "total_steps": 780, "epoch": 4.9728, "percentage": 100.0, "elapsed_time": "10:08:34", "remaining_time": "0:00:00"}