Saving weights and logs of step 10000

events.out.tfevents.1641896043.t1v-n-e1a08808-w-0.1346783.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bee8663b0eb9cbb10e319fd1ca5b113af27cf8ae7cdcb5208afa84ac274bbf60
+size 1470136

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e74c89bc95480c41642de0c02ce549eeb7d4d96ee7d88aad54c0625a29a551df
+oid sha256:11958977b1f537c3ffe3d18da06f5d2db044284ebe998c2cb7a15698a9f6ff52
 size 498796983

run_step1.sh

@@ -8,9 +8,9 @@
     --cache_dir="/mnt/disks/flaxdisk/cache/" \
     --max_seq_length="128" \
     --weight_decay="0.01" \
-    --per_device_train_batch_size="192" \
-    --per_device_eval_batch_size="192" \
-    --learning_rate="4e-4" \
+    --per_device_train_batch_size="200" \
+    --per_device_eval_batch_size="200" \
+    --learning_rate="3e-4" \
     --warmup_steps="10000" \
     --overwrite_output_dir \
     --num_train_epochs="1000" \