Saving weights and logs of step 1000

events.out.tfevents.1644152162.t1v-n-ccbf3e94-w-0.742851.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3cc823acc0987cbd5ee0268b216764439077fafaa8a4d0dac6ce85248a66836
+size 40

events.out.tfevents.1644154971.t1v-n-ccbf3e94-w-0.756187.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ab370b15b8387af7b1f61244d453535d8602bc828029d6095739c6fed2bb9fe
+size 40

events.out.tfevents.1644155872.t1v-n-ccbf3e94-w-0.758395.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cea223d240ece0f2901cb26443d866f4a1e5250ad2b2e0244f21b122b8c573f2
+size 147136

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4abb41156cf5e1bcf659c487aa968be2612b5af34c29a3e312dedf77fe42746c
+oid sha256:217ab9d795fd2998dddd71aefa7e009066740f1b26675342e3cae64e294b913b
 size 498796983

run_512.sh

@@ -8,8 +8,8 @@ python run_mlm_flax.py \
     --cache_dir="/mnt/disks/flaxdisk/cache/" \
     --max_seq_length="512" \
     --weight_decay="0.01" \
-    --per_device_train_batch_size="232" \
-    --per_device_eval_batch_size="232" \
+    --per_device_train_batch_size="46" \
+    --per_device_eval_batch_size="46" \
     --pad_to_max_length \
     --learning_rate="0.00015" \
     --warmup_steps="10000" \