Saving weights and logs of step 10000

events.out.tfevents.1641905881.t1v-n-6f5efcd5-w-0.388508.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d99cfb1cef2e51ffe101591a801345cab183ad97a1c9e5b3a8f74dbd63222a5c
+size 1470136

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f5531c0f640ed4d6cbc5730949c73f792eb674404dd0b79da3d2a0cdc1942583
+oid sha256:ae417426a414e194db93c6a7bc68590aa518fe5c36a13f22cb27cf42b46316a9
 size 498796983

run_step4.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_4_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