Saving weights and logs of step 10000

events.out.tfevents.1629913928.t1v-n-1ae8dadb-w-0.220173.0.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a528c747428e735054fa02c2fdec7ae85ebd170864e9cdf03bb98e932478871
-size 2948827
+oid sha256:ba48449025afa6e28d08421224bf053c3a32ec691cfc7318dec5bd9a48fb65c3
+size 3246967

events.out.tfevents.1629925854.t1v-n-1ae8dadb-w-0.258038.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88f987eeac166770be4352f4826a9e8e92c3054646df228be98b0dbe1b4d4841
+size 1470757

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:edb900664dd5635d727c9c07b98da3d965ec5cd6951723d052ffa3635914e123
+oid sha256:b26c895455dd16c0fe4b651602c1dbb07f341ea81b73ab24ffb20a9fcc473b6e
 size 711588089

start_train.sh

@@ -16,7 +16,7 @@ python3 run_mlm_flax.py \
     --adam_beta2="0.98" \
     --adam_epsilon="1e-6" \
     --learning_rate="4e-4" \
-    --warmup_steps="20000" \
+    --warmup_steps="15000" \
     --overwrite_output_dir \
     --num_train_epochs="3" \
     --save_strategy="steps" \