Training in progress, step 120

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1eb8fe3e843c91b07fc4f81e323dad7b66eb9efac2d6801b75e629750f136a29
+oid sha256:243c7f679d88d6205c931792cc7d274a72db37a42ba8b6112d77b5efdfd14e46
 size 87368144

logging.jsonl

@@ -12,3 +12,5 @@
 {"loss": 0.8164465, "grad_norm": 0.26674172, "learning_rate": 9.955e-05, "token_acc": 0.76918452, "epoch": 0.27575642, "global_step/max_steps": "90/981", "percentage": "9.17%", "elapsed_time": "39m 33s", "remaining_time": "6h 31m 37s", "memory(GiB)": 31.44, "train_speed(iter/s)": 0.037919}
 {"loss": 0.79769683, "grad_norm": 0.27703497, "learning_rate": 9.929e-05, "token_acc": 0.77386085, "epoch": 0.30639602, "global_step/max_steps": "100/981", "percentage": "10.19%", "elapsed_time": "43m 29s", "remaining_time": "6h 23m 10s", "memory(GiB)": 31.44, "train_speed(iter/s)": 0.03832}
 {"eval_loss": 3.061028, "eval_runtime": 124.5655, "eval_samples_per_second": 2.176, "eval_steps_per_second": 0.546, "eval_token_acc": 0.78014999, "epoch": 0.30639602, "global_step/max_steps": "100/981", "percentage": "10.19%", "elapsed_time": "45m 34s", "remaining_time": "6h 41m 28s", "memory(GiB)": 31.44, "train_speed(iter/s)": 0.036574}
+{"loss": 0.77646089, "grad_norm": 0.28990954, "learning_rate": 9.898e-05, "token_acc": 0.77767867, "epoch": 0.33703562, "global_step/max_steps": "110/981", "percentage": "11.21%", "elapsed_time": "49m 28s", "remaining_time": "6h 31m 49s", "memory(GiB)": 31.44, "train_speed(iter/s)": 0.03705}
+{"loss": 0.7550271, "grad_norm": 0.31104651, "learning_rate": 9.861e-05, "token_acc": 0.78237806, "epoch": 0.36767522, "global_step/max_steps": "120/981", "percentage": "12.23%", "elapsed_time": "53m 27s", "remaining_time": "6h 23m 33s", "memory(GiB)": 31.44, "train_speed(iter/s)": 0.037412}

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:a33e67fa14ccaf6de6897a35291a55a9a92153bc117a267a9a0343af5ba93f85
-size 11403
+oid sha256:b31c0524d4a9430a9f1b4f74a0b6acec54fbe6d464bbd528f12543c42dfe4080
+size 11923