Update run.sh
Browse files
run.sh
CHANGED
@@ -4,13 +4,13 @@ python run_phoneme_recognition_ctc.py \
|
|
4 |
--dataset_config_name "clean" "clean" \
|
5 |
--train_split_name "train[:15%]" "train.100" \
|
6 |
--eval_split_name "validation[:8%]" "validation" \
|
7 |
-
|
8 |
--text_column_name "text" "text" \
|
9 |
-
|
10 |
--output_dir="./saved/wav2vec2-base-en_kr_200h-pr" \
|
11 |
--cache_dir=".cache" \
|
12 |
--overwrite_output_dir \
|
13 |
-
|
14 |
--num_train_epochs="15" \
|
15 |
--eval_metrics "wer" "cer" \
|
16 |
--per_device_train_batch_size="32" \
|
@@ -26,7 +26,7 @@ python run_phoneme_recognition_ctc.py \
|
|
26 |
--layerdrop="0.0" \
|
27 |
--save_total_limit="3" \
|
28 |
--freeze_feature_encoder \
|
29 |
-
|
30 |
--gradient_checkpointing \
|
31 |
--chars_to_ignore , ? . ! - \; \: \" \` “ % ‘ ” � "\b" \
|
32 |
--group_by_length \
|
|
|
4 |
--dataset_config_name "clean" "clean" \
|
5 |
--train_split_name "train[:15%]" "train.100" \
|
6 |
--eval_split_name "validation[:8%]" "validation" \
|
7 |
+
--audio_column_name "audio" "audio" \
|
8 |
--text_column_name "text" "text" \
|
9 |
+
--phoneme_language "ko" "en-us" \
|
10 |
--output_dir="./saved/wav2vec2-base-en_kr_200h-pr" \
|
11 |
--cache_dir=".cache" \
|
12 |
--overwrite_output_dir \
|
13 |
+
--preprocessing_num_workers "64" \
|
14 |
--num_train_epochs="15" \
|
15 |
--eval_metrics "wer" "cer" \
|
16 |
--per_device_train_batch_size="32" \
|
|
|
26 |
--layerdrop="0.0" \
|
27 |
--save_total_limit="3" \
|
28 |
--freeze_feature_encoder \
|
29 |
+
--ctc_zero_infinity \
|
30 |
--gradient_checkpointing \
|
31 |
--chars_to_ignore , ? . ! - \; \: \" \` “ % ‘ ” � "\b" \
|
32 |
--group_by_length \
|