Training in progress, step 960

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:847b4a6191b4b70443c1533d54595528b86139918c4d0a478943e09fabbcf544
+oid sha256:8b0d75d735fb2e2aa9243b1cfeac814dd8a312034a25db7ca3b847c50a2ac36a
 size 87368144

logging.jsonl

@@ -112,3 +112,6 @@
 {"loss": 0.64668331, "grad_norm": 0.40319866, "learning_rate": 1.06e-06, "token_acc": 0.80315308, "epoch": 2.8150134, "global_step/max_steps": "920/981", "percentage": "93.78%", "elapsed_time": "6h 37m 42s", "remaining_time": "26m 22s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038554}
 {"loss": 0.65769014, "grad_norm": 0.40031841, "learning_rate": 7.4e-07, "token_acc": 0.80058395, "epoch": 2.84565301, "global_step/max_steps": "930/981", "percentage": "94.80%", "elapsed_time": "6h 41m 38s", "remaining_time": "22m 1s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038592}
 {"loss": 0.63896413, "grad_norm": 0.40605664, "learning_rate": 4.8e-07, "token_acc": 0.80590886, "epoch": 2.87629261, "global_step/max_steps": "940/981", "percentage": "95.82%", "elapsed_time": "6h 45m 34s", "remaining_time": "17m 41s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038628}
+{"loss": 0.64728446, "grad_norm": 0.48287588, "learning_rate": 2.7e-07, "token_acc": 0.80227015, "epoch": 2.90693221, "global_step/max_steps": "950/981", "percentage": "96.84%", "elapsed_time": "6h 49m 31s", "remaining_time": "13m 21s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038662}
+{"eval_loss": 2.61313844, "eval_runtime": 125.0544, "eval_samples_per_second": 2.167, "eval_steps_per_second": 0.544, "eval_token_acc": 0.80230966, "epoch": 2.90693221, "global_step/max_steps": "950/981", "percentage": "96.84%", "elapsed_time": "6h 51m 36s", "remaining_time": "13m 25s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038466}
+{"loss": 0.64517288, "grad_norm": 0.44848776, "learning_rate": 1.3e-07, "token_acc": 0.80361007, "epoch": 2.93757181, "global_step/max_steps": "960/981", "percentage": "97.86%", "elapsed_time": "6h 55m 29s", "remaining_time": "9m 5s", "memory(GiB)": 32.39, "train_speed(iter/s)": 0.038509}

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 sha256:158bcbd68f3f6781be6608d1cbd211ff86dee8c0c74c40addbdbbbd0d231e03d
-size 38837
+oid sha256:7e2310d609e43362e09fb8d7e58ad76f58c65c598882691abcbe8850d86285d0
+size 39691