Training in progress, step 900

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a14ca1ab7be1d63b0a8335320615cd44af1308f295fb07ae1d7aed9eea2967db
+oid sha256:8877f100a4c8fb3469e37af9ec91c7c5e5291b6c26dd6d27e8d846985c5bc416
 size 87368144

logging.jsonl

@@ -105,3 +105,6 @@
 {"loss": 0.63968625, "grad_norm": 0.43007135, "learning_rate": 4.11e-06, "token_acc": 0.80489425, "epoch": 2.63117579, "global_step/max_steps": "860/981", "percentage": "87.67%", "elapsed_time": "6h 12m 18s", "remaining_time": "52m 23s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038498}
 {"loss": 0.63850884, "grad_norm": 0.45915908, "learning_rate": 3.47e-06, "token_acc": 0.80574739, "epoch": 2.6618154, "global_step/max_steps": "870/981", "percentage": "88.69%", "elapsed_time": "6h 16m 15s", "remaining_time": "48m 0s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038538}
 {"loss": 0.64804406, "grad_norm": 0.44728887, "learning_rate": 2.88e-06, "token_acc": 0.80335091, "epoch": 2.692455, "global_step/max_steps": "880/981", "percentage": "89.70%", "elapsed_time": "6h 20m 9s", "remaining_time": "43m 37s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.03858}
+{"loss": 0.64023747, "grad_norm": 0.4411357, "learning_rate": 2.34e-06, "token_acc": 0.80492148, "epoch": 2.7230946, "global_step/max_steps": "890/981", "percentage": "90.72%", "elapsed_time": "6h 24m 5s", "remaining_time": "39m 16s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.03862}
+{"loss": 0.65595984, "grad_norm": 0.44405082, "learning_rate": 1.86e-06, "token_acc": 0.80046947, "epoch": 2.7537342, "global_step/max_steps": "900/981", "percentage": "91.74%", "elapsed_time": "6h 27m 59s", "remaining_time": "34m 55s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.03866}
+{"eval_loss": 2.61341238, "eval_runtime": 124.9163, "eval_samples_per_second": 2.169, "eval_steps_per_second": 0.544, "eval_token_acc": 0.8022985, "epoch": 2.7537342, "global_step/max_steps": "900/981", "percentage": "91.74%", "elapsed_time": "6h 30m 4s", "remaining_time": "35m 6s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038454}

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 sha256:7cd6098d4c92834ec80e76dd46a2e8f18ecfcee01451e556810f1448d76d1ddc
-size 36923
+oid sha256:58ccff55379881d12ebd94e82e1df233d8a569ea0e6ba1d6de8f41f9d94d8060
+size 37777