Training in progress, step 160

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0879952a39b1a08f2c2ca9d7031b14fcd5793233df62cd197b77ec3d6036b5f8
+oid sha256:1d4dd1db688211a7226677261cccef133c48f89a56a61f893998749aeb1ca91c
 size 87368144

logging.jsonl

@@ -16,3 +16,6 @@
 {"loss": 0.71300731, "grad_norm": 0.27155069, "learning_rate": 9.682e-05, "token_acc": 0.79392519, "epoch": 0.47749316, "global_step/max_steps": "120/756", "percentage": "15.87%", "elapsed_time": "51m 4s", "remaining_time": "4h 30m 42s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039158}
 {"loss": 0.71306467, "grad_norm": 0.29438251, "learning_rate": 9.6e-05, "token_acc": 0.79304077, "epoch": 0.51728426, "global_step/max_steps": "130/756", "percentage": "17.20%", "elapsed_time": "54m 55s", "remaining_time": "4h 24m 27s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039453}
 {"loss": 0.70194082, "grad_norm": 0.32997555, "learning_rate": 9.51e-05, "token_acc": 0.79521872, "epoch": 0.55707535, "global_step/max_steps": "140/756", "percentage": "18.52%", "elapsed_time": "58m 43s", "remaining_time": "4h 18m 24s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039731}
+{"loss": 0.70811157, "grad_norm": 0.29847297, "learning_rate": 9.412e-05, "token_acc": 0.79354036, "epoch": 0.59686645, "global_step/max_steps": "150/756", "percentage": "19.84%", "elapsed_time": "1h 2m 32s", "remaining_time": "4h 12m 39s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039976}
+{"eval_loss": 2.7713089, "eval_runtime": 82.4754, "eval_samples_per_second": 2.219, "eval_steps_per_second": 0.558, "eval_token_acc": 0.80116134, "epoch": 0.59686645, "global_step/max_steps": "150/756", "percentage": "19.84%", "elapsed_time": "1h 3m 54s", "remaining_time": "4h 18m 12s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.039116}
+{"loss": 0.70022655, "grad_norm": 0.34647727, "learning_rate": 9.304e-05, "token_acc": 0.796757, "epoch": 0.63665755, "global_step/max_steps": "160/756", "percentage": "21.16%", "elapsed_time": "1h 7m 44s", "remaining_time": "4h 12m 18s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.03937}

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:d4dc414eab83a3c62ed3c678074e9f48519ad2e141ab4c9f36367aa399f06769
-size 12449
+oid sha256:2254fbe6f02c2113264a141facff71bd117b037fa0ca7e8e964781226932cbb0
+size 13303