File size: 1,035 Bytes
0eb8a8d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
python run_speech_recognition_ctc_bnb.py \
	--dataset_name="mozilla-foundation/common_voice_8_0" \
	--model_name_or_path="facebook/wav2vec2-xls-r-1b" \
	--dataset_config_name="et" \
	--output_dir="./" \
	--overwrite_output_dir \
	--max_steps="25_000" \
	--per_device_train_batch_size="32" \
	--per_device_eval_batch_size="32" \
	--gradient_accumulation_steps="1" \
	--learning_rate="7e-5" \
	--warmup_steps="500" \
	--length_column_name="input_length" \
	--evaluation_strategy="steps" \
	--text_column_name="sentence" \
	--chars_to_ignore , ? . ! \- \; \: \" “ % ‘ ” � — ’ … „ – \
	--save_steps="500" \
	--eval_steps="500" \
	--logging_steps="100" \
	--layerdrop="0.0" \
	--activation_dropout="7e-2" \
	--save_total_limit="3" \
	--freeze_feature_encoder \
	--feat_proj_dropout="0.0" \
	--mask_time_prob="9e-2" \
	--mask_time_length="10" \
	--mask_feature_prob="0.25" \
	--mask_feature_length="64" \
	--gradient_checkpointing \
	--use_auth_token \
	--fp16 \
	--group_by_length \
	--do_train --do_eval \
	--push_to_hub