Update run.sh
Browse files
run.sh
CHANGED
@@ -2,14 +2,14 @@
|
|
2 |
|
3 |
python run_speech_recognition_ctc.py \
|
4 |
--dataset_name="mozilla-foundation/common_voice_7_0" \
|
5 |
-
--model_name_or_path="facebook/wav2vec2-xls-r-
|
6 |
--dataset_config_name="uk" \
|
7 |
--output_dir="./cv-uk-ft" \
|
8 |
--num_train_epochs="100" \
|
9 |
--per_device_train_batch_size="8" \
|
10 |
--per_device_eval_batch_size="8" \
|
11 |
--gradient_accumulation_steps="20" \
|
12 |
-
--learning_rate="
|
13 |
--warmup_steps="500" \
|
14 |
--length_column_name="input_length" \
|
15 |
--evaluation_strategy="steps" \
|
@@ -27,7 +27,7 @@ python run_speech_recognition_ctc.py \
|
|
27 |
--mask_feature_prob="0.25" \
|
28 |
--mask_feature_length="64" \
|
29 |
--freeze_feature_encoder \
|
30 |
-
--chars_to_ignore
|
31 |
--fp16 \
|
32 |
--group_by_length \
|
33 |
--use_auth_token \
|
2 |
|
3 |
python run_speech_recognition_ctc.py \
|
4 |
--dataset_name="mozilla-foundation/common_voice_7_0" \
|
5 |
+
--model_name_or_path="facebook/wav2vec2-xls-r-300m" \
|
6 |
--dataset_config_name="uk" \
|
7 |
--output_dir="./cv-uk-ft" \
|
8 |
--num_train_epochs="100" \
|
9 |
--per_device_train_batch_size="8" \
|
10 |
--per_device_eval_batch_size="8" \
|
11 |
--gradient_accumulation_steps="20" \
|
12 |
+
--learning_rate="3e-4" \
|
13 |
--warmup_steps="500" \
|
14 |
--length_column_name="input_length" \
|
15 |
--evaluation_strategy="steps" \
|
27 |
--mask_feature_prob="0.25" \
|
28 |
--mask_feature_length="64" \
|
29 |
--freeze_feature_encoder \
|
30 |
+
--chars_to_ignore — , ? . ! - \; \: \" “ % ‘ ” � \
|
31 |
--fp16 \
|
32 |
--group_by_length \
|
33 |
--use_auth_token \
|