Saving weights and logs of step 10000

events.out.tfevents.1641652701.t1v-n-ccbf3e94-w-0.758060.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26a36c7a2f4536b164580c1516c21cdac735756710877e0288c87542b164c7ef
+size 1470136

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:913188ed6c49a926dcb19e8579c89c51e2ee791ba74ea64c5b044b2414d1c30f
+oid sha256:e2cd902501be2cee2d166624671c545d2626cf6948c9cba950a8f2c709e5a7e2
 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="512" \
+    --weight_decay="0.01" \
+    --per_device_train_batch_size="40" \
+    --per_device_eval_batch_size="40" \
+    --learning_rate="1e-4" \
+    --end_learning_rate="5e-3" \
+    --warmup_steps="0" \
+    --overwrite_output_dir \
+    --num_train_epochs="2" \
+    --adam_beta1="0.9" \
+    --adam_beta2="0.98" \
+    --adam_epsilon="1e-6" \
+    --logging_steps="10000" \
+    --save_steps="10000" \
+    --eval_steps="10000" \
+    --preprocessing_num_workers="64" \
+    --auth_token="True" \
+    --dtype="bfloat16" \
+    --push_to_hub