Training in progress, step 500

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b522bd160268df61bb273c69c57f76e6f030a6342559450a47e85d9628b28a31
+oid sha256:6c88a7b59d9130243467a296aa23a20e3e4e1cdd749abbdae2fee261e594abba
 size 87368144

logging.jsonl

@@ -57,3 +57,6 @@
 {"loss": 0.6332561, "grad_norm": 0.42475322, "learning_rate": 3.639e-05, "token_acc": 0.80931666, "epoch": 1.82765481, "global_step/max_steps": "460/756", "percentage": "60.85%", "elapsed_time": "3h 10m 12s", "remaining_time": "2h 2m 23s", "memory(GiB)": 31.47, "train_speed(iter/s)": 0.040307}
 {"loss": 0.62653365, "grad_norm": 0.40236804, "learning_rate": 3.43e-05, "token_acc": 0.81024547, "epoch": 1.86744591, "global_step/max_steps": "470/756", "percentage": "62.17%", "elapsed_time": "3h 13m 54s", "remaining_time": "1h 57m 59s", "memory(GiB)": 31.47, "train_speed(iter/s)": 0.040398}
 {"loss": 0.63048306, "grad_norm": 0.42044917, "learning_rate": 3.224e-05, "token_acc": 0.80941748, "epoch": 1.90723701, "global_step/max_steps": "480/756", "percentage": "63.49%", "elapsed_time": "3h 17m 38s", "remaining_time": "1h 53m 38s", "memory(GiB)": 31.47, "train_speed(iter/s)": 0.040478}
+{"loss": 0.63408885, "grad_norm": 0.46262404, "learning_rate": 3.021e-05, "token_acc": 0.80842041, "epoch": 1.9470281, "global_step/max_steps": "490/756", "percentage": "64.81%", "elapsed_time": "3h 21m 23s", "remaining_time": "1h 49m 19s", "memory(GiB)": 31.66, "train_speed(iter/s)": 0.040551}
+{"loss": 0.62491407, "grad_norm": 0.40705633, "learning_rate": 2.822e-05, "token_acc": 0.8104753, "epoch": 1.9868192, "global_step/max_steps": "500/756", "percentage": "66.14%", "elapsed_time": "3h 25m 14s", "remaining_time": "1h 45m 4s", "memory(GiB)": 31.67, "train_speed(iter/s)": 0.040604}
+{"eval_loss": 2.46459222, "eval_runtime": 81.9508, "eval_samples_per_second": 2.233, "eval_steps_per_second": 0.561, "eval_token_acc": 0.81585048, "epoch": 1.9868192, "global_step/max_steps": "500/756", "percentage": "66.14%", "elapsed_time": "3h 26m 36s", "remaining_time": "1h 45m 46s", "memory(GiB)": 31.67, "train_speed(iter/s)": 0.040335}

runs/events.out.tfevents.1770820661.h004.gautschi.rcac.purdue.edu.2476766.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:528df304222133c7f16f126ae68cd97e4ee27f2b2b56ec19abbf0710c1661f9c
-size 23727
+oid sha256:b7109b8662908c35369f0f31686de72ef3bca52e424da32dde664804fda0cfbb
+size 24581