update

run_step1.sh

@@ -3,16 +3,17 @@
     --model_type="roberta" \
     --config_name="./" \
     --tokenizer_name="./" \
-    --dataset_name="NbAiLab/nbailab_extended" \
+    --train_file /mnt/disks/flaxdisk/corpus/train_1_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="256" \
-    --per_device_eval_batch_size="256" \
-    --learning_rate="6e-4" \
+    --per_device_train_batch_size="200" \
+    --per_device_eval_batch_size="200" \
+    --learning_rate="4e-4" \
     --warmup_steps="10000" \
     --overwrite_output_dir \
-    --num_train_epochs="1000" \
+    --num_train_epochs="2" \
     --adam_beta1="0.9" \
     --adam_beta2="0.98" \
     --logging_steps="10000" \
@@ -21,6 +22,6 @@
     --preprocessing_num_workers="64" \
     --auth_token="True" \
     --static_learning_rate="True" \
-    --dtype="bloat16" \
+    --dtype="bfloat16" \
     --adafactor \
     --push_to_hub