Saving weights and logs of epoch 1

events.out.tfevents.1625848981.t1v-n-b95d739e-w-0.195447.3.v2 → events.out.tfevents.1625926631.t1v-n-b95d739e-w-0.271163.3.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4a0b06bb4c0c8bcf1babf235c79c7b2fcfbd0a8b3874f799b016c5882bc8126e
-size 6822986
+oid sha256:058091f2aa628751ffce3a9b123432d04c86420a4355fec8763decfd8a17d9be
+size 3581817

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ad0a0dab0d811ae79a350baae4a566468756c89b37c7847013a7d71b4d9515a
+oid sha256:8678c0f46f5db25de579fa7beaba06e1c3b2cf73dfa0635f07e5822b760dce16
 size 711588089

nohup.out

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7c6e9b48135ecd060b322a351bcf8d088d66545864cff1ce1225776af641215c
-size 6077478
+oid sha256:355b4eaf707e807efbb0fc5c51bfc2a4a766c6e846b5f6c6e8484f7016d03a01
+size 2037187

run.sh

@@ -8,17 +8,13 @@ python3 run_mlm_flax.py \
     --dataset_name="oscar" \
     --dataset_config_name="unshuffled_deduplicated_id" \
     --max_seq_length="128" \
-    --weight_decay="0.01" \
     --preprocessing_num_workers="64" \
     --per_device_train_batch_size="128" \
     --per_device_eval_batch_size="128" \
-    --learning_rate="2e-5" \
+    --learning_rate="7e-5" \
     --warmup_steps="1000" \
     --overwrite_output_dir \
     --num_train_epochs="10" \
-    --logging_steps="500" \
-    --save_steps="10000" \
-    --eval_steps="10000" \
     --dtype="bfloat16" \
     --push_to_hub