Saving weights and logs of step 10000

events.out.tfevents.1641683356.t1v-n-6f5efcd5-w-0.162700.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0163b0c99b1ed78e8f550e99c635edd775ef6a2bef63b3193060a8ce4ad072b
+size 1470136

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a52a246c98980ecccb6b4d5286f493cced8595c0b764843c8b64f81af08a5742
+oid sha256:6718a309635a20277a4667d1714c69179726612c375f76fdf73ca19206cd4354
 size 498796983

run_step3.sh

@@ -0,0 +1,28 @@
+./run_mlm_flax.py \
+    --output_dir="./" \
+    --model_type="roberta" \
+    --model_name_or_path="./" \
+    --config_name="./" \
+    --tokenizer_name="./" \
+    --train_file /mnt/disks/flaxdisk/corpus/train_3_4.json \
+    --validation_file /mnt/disks/flaxdisk/corpus/validation.json \
+    --cache_dir="/mnt/disks/flaxdisk/cache/" \
+    --max_seq_length="128" \
+    --weight_decay="0.01" \
+    --per_device_train_batch_size="200" \
+    --per_device_eval_batch_size="200" \
+    --learning_rate="6e-4" \
+    --warmup_steps="0" \
+    --overwrite_output_dir \
+    --num_train_epochs="2" \
+    --adam_beta1="0.9" \
+    --adam_beta2="0.98" \
+    --logging_steps="10000" \
+    --save_steps="10000" \
+    --eval_steps="10000" \
+    --preprocessing_num_workers="64" \
+    --auth_token="True" \
+    --static_learning_rate="True" \
+    --dtype="bfloat16" \
+    --adafactor \
+    --push_to_hub