Training in progress, step 3360

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5fbaac5838169f22c3aafcec9038dafe1f658040bb3bbe5d99832852842b2274
+oid sha256:4812f0ffbffc63a681b3c2080ab9e13ec8587a70d7fbd30e8f2e917b91a5ef83
 size 4943274328

trainer_log.jsonl

@@ -305,3 +305,39 @@
 {"current_steps": 3000, "total_steps": 3360, "loss": 0.0055, "lr": 1.036893770336938e-05, "epoch": 4.463283138129764, "percentage": 89.29, "elapsed_time": "7:52:40", "remaining_time": "0:56:43"}
 {"current_steps": 3000, "total_steps": 3360, "eval_loss": 0.03200867399573326, "epoch": 4.463283138129764, "percentage": 89.29, "elapsed_time": "7:56:13", "remaining_time": "0:57:08"}
 {"current_steps": 3010, "total_steps": 3360, "loss": 0.0056, "lr": 9.807161197651742e-06, "epoch": 4.478155791039226, "percentage": 89.58, "elapsed_time": "7:58:41", "remaining_time": "0:55:39"}
+{"current_steps": 3020, "total_steps": 3360, "loss": 0.0055, "lr": 9.260515359712517e-06, "epoch": 4.493028443948689, "percentage": 89.88, "elapsed_time": "8:00:11", "remaining_time": "0:54:03"}
+{"current_steps": 3030, "total_steps": 3360, "loss": 0.0054, "lr": 8.729059187690479e-06, "epoch": 4.507901096858152, "percentage": 90.18, "elapsed_time": "8:01:41", "remaining_time": "0:52:27"}
+{"current_steps": 3040, "total_steps": 3360, "loss": 0.0055, "lr": 8.212850040341273e-06, "epoch": 4.522773749767615, "percentage": 90.48, "elapsed_time": "8:03:11", "remaining_time": "0:50:51"}
+{"current_steps": 3050, "total_steps": 3360, "loss": 0.0053, "lr": 7.711943630846684e-06, "epoch": 4.537646402677078, "percentage": 90.77, "elapsed_time": "8:04:41", "remaining_time": "0:49:15"}
+{"current_steps": 3060, "total_steps": 3360, "loss": 0.0054, "lr": 7.226394020801645e-06, "epoch": 4.55251905558654, "percentage": 91.07, "elapsed_time": "8:06:11", "remaining_time": "0:47:39"}
+{"current_steps": 3070, "total_steps": 3360, "loss": 0.0056, "lr": 6.7562536143796254e-06, "epoch": 4.567391708496003, "percentage": 91.37, "elapsed_time": "8:07:41", "remaining_time": "0:46:04"}
+{"current_steps": 3080, "total_steps": 3360, "loss": 0.0054, "lr": 6.301573152676664e-06, "epoch": 4.582264361405466, "percentage": 91.67, "elapsed_time": "8:09:11", "remaining_time": "0:44:28"}
+{"current_steps": 3090, "total_steps": 3360, "loss": 0.0052, "lr": 5.862401708235076e-06, "epoch": 4.597137014314928, "percentage": 91.96, "elapsed_time": "8:10:41", "remaining_time": "0:42:52"}
+{"current_steps": 3100, "total_steps": 3360, "loss": 0.0055, "lr": 5.438786679747081e-06, "epoch": 4.612009667224391, "percentage": 92.26, "elapsed_time": "8:12:11", "remaining_time": "0:41:16"}
+{"current_steps": 3110, "total_steps": 3360, "loss": 0.0055, "lr": 5.030773786939319e-06, "epoch": 4.626882320133854, "percentage": 92.56, "elapsed_time": "8:13:41", "remaining_time": "0:39:41"}
+{"current_steps": 3120, "total_steps": 3360, "loss": 0.0054, "lr": 4.638407065638322e-06, "epoch": 4.641754973043317, "percentage": 92.86, "elapsed_time": "8:15:11", "remaining_time": "0:38:05"}
+{"current_steps": 3130, "total_steps": 3360, "loss": 0.0054, "lr": 4.261728863017827e-06, "epoch": 4.65662762595278, "percentage": 93.15, "elapsed_time": "8:16:41", "remaining_time": "0:36:29"}
+{"current_steps": 3140, "total_steps": 3360, "loss": 0.0054, "lr": 3.900779833028472e-06, "epoch": 4.671500278862242, "percentage": 93.45, "elapsed_time": "8:18:11", "remaining_time": "0:34:54"}
+{"current_steps": 3150, "total_steps": 3360, "loss": 0.0053, "lr": 3.5555989320099952e-06, "epoch": 4.686372931771705, "percentage": 93.75, "elapsed_time": "8:19:41", "remaining_time": "0:33:18"}
+{"current_steps": 3160, "total_steps": 3360, "loss": 0.0054, "lr": 3.2262234144868116e-06, "epoch": 4.701245584681168, "percentage": 94.05, "elapsed_time": "8:21:11", "remaining_time": "0:31:43"}
+{"current_steps": 3170, "total_steps": 3360, "loss": 0.0052, "lr": 2.912688829147214e-06, "epoch": 4.7161182375906305, "percentage": 94.35, "elapsed_time": "8:22:41", "remaining_time": "0:30:07"}
+{"current_steps": 3180, "total_steps": 3360, "loss": 0.0055, "lr": 2.6150290150067588e-06, "epoch": 4.730990890500093, "percentage": 94.64, "elapsed_time": "8:24:11", "remaining_time": "0:28:32"}
+{"current_steps": 3190, "total_steps": 3360, "loss": 0.0053, "lr": 2.3332760977559873e-06, "epoch": 4.7458635434095555, "percentage": 94.94, "elapsed_time": "8:25:41", "remaining_time": "0:26:56"}
+{"current_steps": 3200, "total_steps": 3360, "loss": 0.0055, "lr": 2.0674604862932654e-06, "epoch": 4.7607361963190185, "percentage": 95.24, "elapsed_time": "8:27:11", "remaining_time": "0:25:21"}
+{"current_steps": 3210, "total_steps": 3360, "loss": 0.0052, "lr": 1.8176108694427927e-06, "epoch": 4.775608849228481, "percentage": 95.54, "elapsed_time": "8:28:41", "remaining_time": "0:23:46"}
+{"current_steps": 3220, "total_steps": 3360, "loss": 0.0054, "lr": 1.583754212858329e-06, "epoch": 4.790481502137943, "percentage": 95.83, "elapsed_time": "8:30:11", "remaining_time": "0:22:10"}
+{"current_steps": 3230, "total_steps": 3360, "loss": 0.0057, "lr": 1.3659157561127732e-06, "epoch": 4.805354155047406, "percentage": 96.13, "elapsed_time": "8:31:41", "remaining_time": "0:20:35"}
+{"current_steps": 3240, "total_steps": 3360, "loss": 0.0053, "lr": 1.1641190099741904e-06, "epoch": 4.820226807956869, "percentage": 96.43, "elapsed_time": "8:33:11", "remaining_time": "0:19:00"}
+{"current_steps": 3250, "total_steps": 3360, "loss": 0.0053, "lr": 9.783857538683603e-07, "epoch": 4.835099460866332, "percentage": 96.73, "elapsed_time": "8:34:41", "remaining_time": "0:17:25"}
+{"current_steps": 3260, "total_steps": 3360, "loss": 0.0055, "lr": 8.087360335281235e-07, "epoch": 4.849972113775795, "percentage": 97.02, "elapsed_time": "8:36:11", "remaining_time": "0:15:50"}
+{"current_steps": 3270, "total_steps": 3360, "loss": 0.0054, "lr": 6.551881588299279e-07, "epoch": 4.864844766685257, "percentage": 97.32, "elapsed_time": "8:37:41", "remaining_time": "0:14:14"}
+{"current_steps": 3280, "total_steps": 3360, "loss": 0.0054, "lr": 5.177587018176777e-07, "epoch": 4.87971741959472, "percentage": 97.62, "elapsed_time": "8:39:11", "remaining_time": "0:12:39"}
+{"current_steps": 3290, "total_steps": 3360, "loss": 0.0054, "lr": 3.964624949141626e-07, "epoch": 4.894590072504183, "percentage": 97.92, "elapsed_time": "8:40:41", "remaining_time": "0:11:04"}
+{"current_steps": 3300, "total_steps": 3360, "loss": 0.0053, "lr": 2.913126293202228e-07, "epoch": 4.909462725413646, "percentage": 98.21, "elapsed_time": "8:42:11", "remaining_time": "0:09:29"}
+{"current_steps": 3310, "total_steps": 3360, "loss": 0.0051, "lr": 2.0232045360184523e-07, "epoch": 4.924335378323108, "percentage": 98.51, "elapsed_time": "8:43:41", "remaining_time": "0:07:54"}
+{"current_steps": 3320, "total_steps": 3360, "loss": 0.0053, "lr": 1.2949557246537678e-07, "epoch": 4.939208031232571, "percentage": 98.81, "elapsed_time": "8:45:11", "remaining_time": "0:06:19"}
+{"current_steps": 3330, "total_steps": 3360, "loss": 0.0052, "lr": 7.284584572085361e-08, "epoch": 4.954080684142034, "percentage": 99.11, "elapsed_time": "8:46:41", "remaining_time": "0:04:44"}
+{"current_steps": 3340, "total_steps": 3360, "loss": 0.0052, "lr": 3.237738743372964e-08, "epoch": 4.968953337051497, "percentage": 99.4, "elapsed_time": "8:48:11", "remaining_time": "0:03:09"}
+{"current_steps": 3350, "total_steps": 3360, "loss": 0.0054, "lr": 8.094565265054365e-09, "epoch": 4.983825989960959, "percentage": 99.7, "elapsed_time": "8:49:41", "remaining_time": "0:01:34"}
+{"current_steps": 3360, "total_steps": 3360, "loss": 0.0054, "lr": 0.0, "epoch": 4.998698642870422, "percentage": 100.0, "elapsed_time": "8:51:11", "remaining_time": "0:00:00"}
+{"current_steps": 3360, "total_steps": 3360, "epoch": 4.998698642870422, "percentage": 100.0, "elapsed_time": "8:52:10", "remaining_time": "0:00:00"}