Training in progress, step 15

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3301075aa1f1b7e190353e81cb8efe7860f1018f4a34d7aee5a645ae97ff9b19
+oid sha256:f5490f71ea5f5266bb67e208f8cd13b7ea0968cd56aabc4ad52572b9d83a1eca
 size 1478569432

logging.jsonl

@@ -11,3 +11,8 @@
 {"eval_loss": 0.18784808, "eval_token_acc": 0.95191512, "eval_runtime": 5.2458, "eval_samples_per_second": 5.719, "eval_steps_per_second": 0.763, "epoch": 0.6557377, "global_step/max_steps": "10/15", "percentage": "66.67%", "elapsed_time": "8m 12s", "remaining_time": "4m 6s"}
 {"loss": 0.1912355, "token_acc": 0.95364897, "grad_norm": 4.21334696, "learning_rate": 3.08e-06, "memory(GiB)": 127.02, "train_speed(iter/s)": 0.019415, "epoch": 0.72131148, "global_step/max_steps": "11/15", "percentage": "73.33%", "elapsed_time": "9m 17s", "remaining_time": "3m 22s"}
 {"loss": 0.17585593, "token_acc": 0.93333333, "grad_norm": 4.19101381, "learning_rate": 2.31e-06, "memory(GiB)": 127.02, "train_speed(iter/s)": 0.019755, "epoch": 0.78688525, "global_step/max_steps": "12/15", "percentage": "80.00%", "elapsed_time": "9m 58s", "remaining_time": "2m 29s"}
+{"loss": 0.15041173, "token_acc": 0.94953342, "grad_norm": 4.02991343, "learning_rate": 1.54e-06, "memory(GiB)": 128.91, "train_speed(iter/s)": 0.019883, "epoch": 0.85245902, "global_step/max_steps": "13/15", "percentage": "86.67%", "elapsed_time": "10m 44s", "remaining_time": "1m 39s"}
+{"loss": 0.1629689, "token_acc": 0.96349914, "grad_norm": 3.61858225, "learning_rate": 7.7e-07, "memory(GiB)": 128.91, "train_speed(iter/s)": 0.020077, "epoch": 0.91803279, "global_step/max_steps": "14/15", "percentage": "93.33%", "elapsed_time": "11m 28s", "remaining_time": "49s"}
+{"loss": 0.11973689, "token_acc": 0.96463922, "grad_norm": 3.35976338, "learning_rate": 0.0, "memory(GiB)": 128.91, "train_speed(iter/s)": 0.01962, "epoch": 0.98360656, "global_step/max_steps": "15/15", "percentage": "100.00%", "elapsed_time": "12m 35s", "remaining_time": "0s"}
+{"eval_loss": 0.15088241, "eval_token_acc": 0.96354056, "eval_runtime": 5.2152, "eval_samples_per_second": 5.752, "eval_steps_per_second": 0.767, "epoch": 0.98360656, "global_step/max_steps": "15/15", "percentage": "100.00%", "elapsed_time": "12m 40s", "remaining_time": "0s"}
+{"train_runtime": 766.5904, "train_samples_per_second": 1.269, "train_steps_per_second": 0.02, "total_flos": 17121736916992.0, "train_loss": 0.26244897, "epoch": 0.98360656, "global_step/max_steps": "15/15", "percentage": "100.00%", "elapsed_time": "12m 46s", "remaining_time": "0s"}

runs/events.out.tfevents.1743101387.notebook-67532d59-da58-11ef-a92b-d22a634ff48d-0.226850.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c4a930881e631352cd4456f9cea7b56009ec271aab9804f70d971a7df141b099
-size 10951
+oid sha256:a7f6fc0bcd3943a425b53305a601ab088d06e03fd54b1af6074936f40bda905a
+size 13507