Saving weights and logs of step 2500

events.out.tfevents.1734083129.t1v-n-53cd541d-w-35.1088772.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:814f063da08f57c0ad2c30e14e1e94d1ab74b8accf7de1f9932ab3e97efde6d3
+size 25422

events.out.tfevents.1734083608.t1v-n-53cd541d-w-35.1090638.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcb5f593aeba3784103714133c5a0f3ce020d535c73fd0e5d66e1f6ab0811e28
+size 63038

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c4c381d4203e6b03a768f5cd7066942ed7e02f54c5534d09efd039be2c86d2e
+oid sha256:33b60ef116cd82e5520339f5cbc2e21252393aa9fc04137b712e77711865066e
 size 1421658229

run_train.sh

@@ -10,7 +10,7 @@ python run_mlm_flax.py \
     --weight_decay="0.01" \
     --per_device_train_batch_size="64" \
     --per_device_eval_batch_size="64" \
-    --learning_rate="5e-4" \
+    --learning_rate="2e-3" \
     --warmup_steps="1000" \
     --overwrite_output_dir \
     --num_train_epochs="100" \