minor fix to train step logic

wandb · Oct 18, 2024 · bb34f6d · bb34f6d
1 parent a1cd349
commit bb34f6d
Showing 1 changed file with 1 addition and 2 deletions.
diff --git a/colabs/system_metrics/WandB_Llama3_1_Training_Colab_TPU.ipynb b/colabs/system_metrics/WandB_Llama3_1_Training_Colab_TPU.ipynb
@@ -1027,7 +1027,6 @@
         "from jax.sharding import NamedSharding\n",
         "from jax.sharding import PartitionSpec as PS\n",
         "\n",
-        "\n",
         "def train(trainer, train_dataloader, eval_dataloader):\n",
         "    total_training_time = 0\n",
         "    total_steps = 0\n",
@@ -1065,7 +1064,7 @@
         "            to_log = {}\n",
         "            to_log.update(\n",
         "                {\n",
-        "                    \"train_step\": step,\n",
+        "                    \"train_step\": total_steps,\n",
         "                    \"train/loss\": metrics[\"loss\"],\n",
         "                    \"train/accuracy\": metrics[\"accuracy\"],\n",
         "                    \"train/step_time\": step_duration,\n",