fix run.sh
Browse files
run.sh
CHANGED
@@ -9,17 +9,17 @@ python3 run_mlm_flax.py \
|
|
9 |
--max_seq_length="128" \
|
10 |
--weight_decay="0.01" \
|
11 |
--preprocessing_num_workers="64" \
|
12 |
-
--per_device_train_batch_size="
|
13 |
-
--per_device_eval_batch_size="
|
14 |
--learning_rate="3e-5" \
|
15 |
-
--warmup_steps="
|
16 |
--overwrite_output_dir \
|
17 |
--seed="19" \
|
18 |
--num_train_epochs="8" \
|
19 |
--adam_beta1="0.9" \
|
20 |
--adam_beta2="0.98" \
|
21 |
-
--logging_steps="
|
22 |
-
--save_steps="
|
23 |
-
--eval_steps="
|
24 |
--dtype="bfloat16" \
|
25 |
--push_to_hub
|
|
|
9 |
--max_seq_length="128" \
|
10 |
--weight_decay="0.01" \
|
11 |
--preprocessing_num_workers="64" \
|
12 |
+
--per_device_train_batch_size="512" \
|
13 |
+
--per_device_eval_batch_size="512" \
|
14 |
--learning_rate="3e-5" \
|
15 |
+
--warmup_steps="312" \
|
16 |
--overwrite_output_dir \
|
17 |
--seed="19" \
|
18 |
--num_train_epochs="8" \
|
19 |
--adam_beta1="0.9" \
|
20 |
--adam_beta2="0.98" \
|
21 |
+
--logging_steps="31" \
|
22 |
+
--save_steps="312" \
|
23 |
+
--eval_steps="1250" \
|
24 |
--dtype="bfloat16" \
|
25 |
--push_to_hub
|