Training in progress, step 260

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:219a0af00b2474a421764053bd1545f1b28a961f1695ce436cf439e07c58b74a
+oid sha256:7c36e07cb0480e03f59ffa1562ba1320f02963a25ae624a6a7b7e81032dffc2e
 size 87368144

logging.jsonl

@@ -28,3 +28,6 @@
 {"loss": 0.6761816, "grad_norm": 0.3342804, "learning_rate": 8.497e-05, "token_acc": 0.80166258, "epoch": 0.87540413, "global_step/max_steps": "220/756", "percentage": "29.10%", "elapsed_time": "1h 32m 18s", "remaining_time": "3h 44m 53s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039722}
 {"loss": 0.67615294, "grad_norm": 0.37265298, "learning_rate": 8.337e-05, "token_acc": 0.80118919, "epoch": 0.91519523, "global_step/max_steps": "230/756", "percentage": "30.42%", "elapsed_time": "1h 36m 8s", "remaining_time": "3h 39m 51s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039873}
 {"loss": 0.67424359, "grad_norm": 0.37488362, "learning_rate": 8.171e-05, "token_acc": 0.7999509, "epoch": 0.95498632, "global_step/max_steps": "240/756", "percentage": "31.75%", "elapsed_time": "1h 39m 52s", "remaining_time": "3h 34m 43s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040051}
+{"loss": 0.66487207, "grad_norm": 0.33123785, "learning_rate": 7.999e-05, "token_acc": 0.80308648, "epoch": 0.99477742, "global_step/max_steps": "250/756", "percentage": "33.07%", "elapsed_time": "1h 43m 41s", "remaining_time": "3h 29m 52s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040183}
+{"eval_loss": 2.62072086, "eval_runtime": 81.3597, "eval_samples_per_second": 2.249, "eval_steps_per_second": 0.565, "eval_token_acc": 0.80856458, "epoch": 0.99477742, "global_step/max_steps": "250/756", "percentage": "33.07%", "elapsed_time": "1h 45m 2s", "remaining_time": "3h 32m 36s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039665}
+{"loss": 0.64975939, "grad_norm": 0.47815546, "learning_rate": 7.821e-05, "token_acc": 0.80659412, "epoch": 1.03183288, "global_step/max_steps": "260/756", "percentage": "34.39%", "elapsed_time": "1h 48m 30s", "remaining_time": "3h 27m 0s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039934}

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 sha256:2ecbc8b480d8597a0be55a8eadbf69fb9839d9a49d43e4f9b820c9d7186fe32d
-size 15747
+oid sha256:09d918d912bfb7a79122cf402dd4ef17b11ba5c9d8dc8265033b7af18526547d
+size 16601