new run
Browse files
run.sh
CHANGED
@@ -6,12 +6,12 @@ python run_speech_recognition_ctc.py \
|
|
6 |
--tokenizer_name_or_path="./" \
|
7 |
--output_dir="./" \
|
8 |
--overwrite_output_dir \
|
9 |
-
--num_train_epochs="
|
10 |
--per_device_train_batch_size="64" \
|
11 |
--per_device_eval_batch_size="64" \
|
12 |
--gradient_accumulation_steps="1" \
|
13 |
-
--learning_rate="
|
14 |
-
--warmup_steps="
|
15 |
--length_column_name="input_length" \
|
16 |
--evaluation_strategy="steps" \
|
17 |
--text_column_name="sentence" \
|
@@ -23,7 +23,7 @@ python run_speech_recognition_ctc.py \
|
|
23 |
--save_total_limit="2" \
|
24 |
--freeze_feature_encoder \
|
25 |
--feat_proj_dropout="0.0" \
|
26 |
-
--mask_time_prob="0.
|
27 |
--mask_time_length="10" \
|
28 |
--mask_feature_prob="0.25" \
|
29 |
--mask_feature_length="10" \
|
@@ -31,7 +31,7 @@ python run_speech_recognition_ctc.py \
|
|
31 |
--report_to="wandb" \
|
32 |
--run_name="xls-r-300m-fr" \
|
33 |
--max_eval_samples="6000" \
|
34 |
-
--max_duration_in_seconds="
|
35 |
--use_auth_token \
|
36 |
--fp16 \
|
37 |
--group_by_length \
|
|
|
6 |
--tokenizer_name_or_path="./" \
|
7 |
--output_dir="./" \
|
8 |
--overwrite_output_dir \
|
9 |
+
--num_train_epochs="4" \
|
10 |
--per_device_train_batch_size="64" \
|
11 |
--per_device_eval_batch_size="64" \
|
12 |
--gradient_accumulation_steps="1" \
|
13 |
+
--learning_rate="2e-4" \
|
14 |
+
--warmup_steps="2500" \
|
15 |
--length_column_name="input_length" \
|
16 |
--evaluation_strategy="steps" \
|
17 |
--text_column_name="sentence" \
|
|
|
23 |
--save_total_limit="2" \
|
24 |
--freeze_feature_encoder \
|
25 |
--feat_proj_dropout="0.0" \
|
26 |
+
--mask_time_prob="0.6" \
|
27 |
--mask_time_length="10" \
|
28 |
--mask_feature_prob="0.25" \
|
29 |
--mask_feature_length="10" \
|
|
|
31 |
--report_to="wandb" \
|
32 |
--run_name="xls-r-300m-fr" \
|
33 |
--max_eval_samples="6000" \
|
34 |
+
--max_duration_in_seconds="10" \
|
35 |
--use_auth_token \
|
36 |
--fp16 \
|
37 |
--group_by_length \
|