Updating run script
Browse files
run.sh
CHANGED
@@ -1,22 +1,30 @@
|
|
|
|
1 |
HUB_TOKEN=`cat $HOME/.huggingface/token`
|
2 |
./run_mlm_flax.py \
|
3 |
--output_dir="./" \
|
4 |
--model_type="roberta" \
|
5 |
--config_name="./" \
|
6 |
--tokenizer_name="./" \
|
7 |
-
--dataset_name="
|
8 |
-
--dataset_config_name="
|
|
|
9 |
--max_seq_length="128" \
|
10 |
--pad_to_max_length \
|
11 |
--per_device_train_batch_size="128" \
|
12 |
--per_device_eval_batch_size="128" \
|
13 |
-
--
|
|
|
|
|
|
|
|
|
14 |
--save_strategy="steps" \
|
15 |
--save_steps="10000" \
|
16 |
--save_total_limit="5" \
|
17 |
-
--warmup_steps="
|
18 |
--overwrite_output_dir \
|
19 |
-
--
|
|
|
|
|
20 |
--dtype="bfloat16" \
|
21 |
--push_to_hub_model_id="flax-community/bertin-roberta-large-spanish" \
|
22 |
--push_to_hub_token="$HUB_TOKEN"
|
|
|
1 |
+
# From https://arxiv.org/pdf/1907.11692.pdf
|
2 |
HUB_TOKEN=`cat $HOME/.huggingface/token`
|
3 |
./run_mlm_flax.py \
|
4 |
--output_dir="./" \
|
5 |
--model_type="roberta" \
|
6 |
--config_name="./" \
|
7 |
--tokenizer_name="./" \
|
8 |
+
--dataset_name="mc4" \
|
9 |
+
--dataset_config_name="es" \
|
10 |
+
--dataset_streamnig \
|
11 |
--max_seq_length="128" \
|
12 |
--pad_to_max_length \
|
13 |
--per_device_train_batch_size="128" \
|
14 |
--per_device_eval_batch_size="128" \
|
15 |
+
--adam_beta1="0.9" \
|
16 |
+
--adam_beta2="0.98" \
|
17 |
+
--adam_epsilon="1e-6" \
|
18 |
+
--learning_rate="4e-4" \
|
19 |
+
--weight_decay="0.01" \
|
20 |
--save_strategy="steps" \
|
21 |
--save_steps="10000" \
|
22 |
--save_total_limit="5" \
|
23 |
+
--warmup_steps="30000" \
|
24 |
--overwrite_output_dir \
|
25 |
+
--num_train_steps="500000" \
|
26 |
+
--eval_steps="10000" \
|
27 |
+
--logging_steps="500" \
|
28 |
--dtype="bfloat16" \
|
29 |
--push_to_hub_model_id="flax-community/bertin-roberta-large-spanish" \
|
30 |
--push_to_hub_token="$HUB_TOKEN"
|