Training in progress, step 400

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a69ee03b64c8cbdd48c01fef90c7d673be28d6e06f91f4f9888a6047712c9fcb
+oid sha256:aabb88eda7208825a69335e47610428682efc921845bb28e6df346ef44b472df
 size 87368144

logging.jsonl

@@ -45,3 +45,6 @@
 {"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}
 {"loss": 0.63130598, "grad_norm": 0.40477979, "learning_rate": 5.589e-05, "token_acc": 0.80867286, "epoch": 1.46953494, "global_step/max_steps": "370/756", "percentage": "48.94%", "elapsed_time": "2h 33m 18s", "remaining_time": "2h 39m 56s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040222}
 {"loss": 0.64159236, "grad_norm": 0.42596149, "learning_rate": 5.372e-05, "token_acc": 0.80728027, "epoch": 1.50932604, "global_step/max_steps": "380/756", "percentage": "50.26%", "elapsed_time": "2h 37m 12s", "remaining_time": "2h 35m 32s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040287}
+{"loss": 0.64312091, "grad_norm": 0.42579779, "learning_rate": 5.153e-05, "token_acc": 0.80699175, "epoch": 1.54911714, "global_step/max_steps": "390/756", "percentage": "51.59%", "elapsed_time": "2h 41m 0s", "remaining_time": "2h 31m 6s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040369}
+{"loss": 0.63438926, "grad_norm": 0.39897516, "learning_rate": 4.934e-05, "token_acc": 0.80871207, "epoch": 1.58890823, "global_step/max_steps": "400/756", "percentage": "52.91%", "elapsed_time": "2h 44m 42s", "remaining_time": "2h 26m 35s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040477}
+{"eval_loss": 2.51271057, "eval_runtime": 82.0433, "eval_samples_per_second": 2.231, "eval_steps_per_second": 0.561, "eval_token_acc": 0.81330806, "epoch": 1.58890823, "global_step/max_steps": "400/756", "percentage": "52.91%", "elapsed_time": "2h 46m 4s", "remaining_time": "2h 27m 48s", "memory(GiB)": 31.41, "train_speed(iter/s)": 0.040143}

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 sha256:312afb309fecee3bcc6acb3e19003575c25734839c1034789c9e892fcf4d56a2
-size 20429
+oid sha256:b9a3f97a5f226e1dfb062d6347c9b0e4dbac6889d7dd49070a675062a281534a
+size 21283