Training in progress, step 360

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:79cc79ea7fd526b12b59ae36824c32634efb31142f984cfb619b01d4886ce1e9
+oid sha256:63341b0a1943750df6d50cf29b005a948966c4f8902064f1a622e791bddb7aee
 size 87368144

logging.jsonl

@@ -40,3 +40,6 @@
 {"loss": 0.63730922, "grad_norm": 0.41076234, "learning_rate": 6.653e-05, "token_acc": 0.8094828, "epoch": 1.27057946, "global_step/max_steps": "320/756", "percentage": "42.33%", "elapsed_time": "2h 12m 56s", "remaining_time": "3h 1m 7s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040118}
 {"loss": 0.64370165, "grad_norm": 0.37233034, "learning_rate": 6.445e-05, "token_acc": 0.80564955, "epoch": 1.31037055, "global_step/max_steps": "330/756", "percentage": "43.65%", "elapsed_time": "2h 16m 41s", "remaining_time": "2h 56m 26s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040238}
 {"loss": 0.62285523, "grad_norm": 0.42613682, "learning_rate": 6.234e-05, "token_acc": 0.81161872, "epoch": 1.35016165, "global_step/max_steps": "340/756", "percentage": "44.97%", "elapsed_time": "2h 20m 25s", "remaining_time": "2h 51m 48s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040353}
+{"loss": 0.63389444, "grad_norm": 0.37893876, "learning_rate": 6.021e-05, "token_acc": 0.80897209, "epoch": 1.38995275, "global_step/max_steps": "350/756", "percentage": "46.30%", "elapsed_time": "2h 24m 12s", "remaining_time": "2h 47m 17s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040448}
+{"eval_loss": 2.5464642, "eval_runtime": 81.8498, "eval_samples_per_second": 2.236, "eval_steps_per_second": 0.562, "eval_token_acc": 0.81189995, "epoch": 1.38995275, "global_step/max_steps": "350/756", "percentage": "46.30%", "elapsed_time": "2h 25m 34s", "remaining_time": "2h 48m 52s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040069}
+{"loss": 0.63284993, "grad_norm": 0.47583783, "learning_rate": 5.806e-05, "token_acc": 0.80906272, "epoch": 1.42974384, "global_step/max_steps": "360/756", "percentage": "47.62%", "elapsed_time": "2h 29m 21s", "remaining_time": "2h 44m 17s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040174}

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 sha256:0dd112536a31f0a332489b83f289387b0b02bea4b4ae16b32256bbf53f616380
-size 19045
+oid sha256:4231a64a960efb38e483651f21df9da07999c12ceed1e93b75bd151302134898
+size 19899