Training in progress, step 560

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:df2593f2d763b5d64e9c06a92c8425bca84c718b6c32e248743f518a1199625e
+oid sha256:78bd0c986ec13b294fa4a56d3d9ac4cb5333208e3c97450845a22914a2441deb
 size 87368144

logging.jsonl

@@ -64,3 +64,6 @@
 {"loss": 0.68433537, "grad_norm": 0.41219425, "learning_rate": 4.924e-05, "token_acc": 0.79433732, "epoch": 1.59134431, "global_step/max_steps": "520/981", "percentage": "53.01%", "elapsed_time": "3h 45m 42s", "remaining_time": "3h 20m 6s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038397}
 {"loss": 0.66399922, "grad_norm": 0.37705389, "learning_rate": 4.755e-05, "token_acc": 0.80079552, "epoch": 1.62198391, "global_step/max_steps": "530/981", "percentage": "54.03%", "elapsed_time": "3h 49m 35s", "remaining_time": "3h 15m 22s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038474}
 {"loss": 0.67126565, "grad_norm": 0.45150718, "learning_rate": 4.587e-05, "token_acc": 0.79797129, "epoch": 1.65262352, "global_step/max_steps": "540/981", "percentage": "55.05%", "elapsed_time": "3h 53m 29s", "remaining_time": "3h 10m 41s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038544}
+{"loss": 0.66462779, "grad_norm": 0.38564503, "learning_rate": 4.419e-05, "token_acc": 0.8000114, "epoch": 1.68326312, "global_step/max_steps": "550/981", "percentage": "56.07%", "elapsed_time": "3h 57m 21s", "remaining_time": "3h 6m 0s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038618}
+{"eval_loss": 2.65649629, "eval_runtime": 125.0271, "eval_samples_per_second": 2.168, "eval_steps_per_second": 0.544, "eval_token_acc": 0.79980905, "epoch": 1.68326312, "global_step/max_steps": "550/981", "percentage": "56.07%", "elapsed_time": "3h 59m 26s", "remaining_time": "3h 7m 38s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038282}
+{"loss": 0.66296554, "grad_norm": 0.51134419, "learning_rate": 4.252e-05, "token_acc": 0.80055125, "epoch": 1.71390272, "global_step/max_steps": "560/981", "percentage": "57.08%", "elapsed_time": "4h 3m 22s", "remaining_time": "3h 2m 57s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038351}

runs/events.out.tfevents.1772968616.h016.gautschi.rcac.purdue.edu.2790912.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d9de3e449946b8923fba2c0750cf4c2c7e7726d6ac193417c9a2143032583ae5
-size 25645
+oid sha256:b9f7cb2b99a34814c7ef9ebe44f9e3e61c9bb58fd160340b48edf5c4f771103f
+size 26499