Saving weights and logs of epoch 1

events.out.tfevents.1626318482.t1v-n-b95d739e-w-0.622701.3.v2 → events.out.tfevents.1626319431.t1v-n-b95d739e-w-0.625591.3.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c0c8833b4f4649f58ab0d01d47c772f8c05080f371d5f9d57e7134d997e944a1
-size 157187
+oid sha256:bb35f1f14d51c1f4e9afb3664ceab7a11e147493b1d3184329d618a7986301dd
+size 314319

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91922c046bb159618da797c8c5076e9684aa45c6ded263ff8c60dab3cb008059
+oid sha256:427f60d6c87c92649b52d6c06f9ac916c80a12d17a2ebb28681f7bd0178fd90a
 size 498796983

nohup.out

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:99c46650710b372548e97ab2d4a123983e2b495c3ceb094847f500b4ac3a64f7
-size 193918
+oid sha256:68e25027079dd5d9cfb4eecc723c5224ffbe7f29a76c231c10439cc7b4e817d4
+size 234600

run.sh

@@ -9,8 +9,8 @@ python3 run_mlm_flax.py \
     --max_seq_length="128" \
     --preprocessing_num_workers="64" \
     --weight_decay="0.0" \
-    --per_device_train_batch_size="128" \
-    --per_device_eval_batch_size="128" \
+    --per_device_train_batch_size="64" \
+    --per_device_eval_batch_size="64" \
     --learning_rate="2e-4" \
     --warmup_steps="1000" \
     --overwrite_output_dir \