patrickvonplaten commited on
Commit
c9a10ac
1 Parent(s): a85a6a3
Files changed (1) hide show
  1. run.sh +37 -0
run.sh ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env bash
2
+ python -m torch.distributed.launch \
3
+ --nproc_per_node 2 run_speech_recognition_ctc.py \
4
+ --dataset_name="common_voice" \
5
+ --model_name_or_path="facebook/wav2vec2-xls-r-300m" \
6
+ --tokenizer_name_or_path="./created_tokenizer_from_cv_sv" \
7
+ --dataset_config_name="sv-SE" \
8
+ --phoneme_language="sv" \
9
+ --output_dir="./wav2vec2-xls-r-300m-phoneme-sv" \
10
+ --overwrite_output_dir \
11
+ --num_train_epochs="100" \
12
+ --per_device_train_batch_size="16" \
13
+ --gradient_accumulation_steps="1" \
14
+ --learning_rate="5e-4" \
15
+ --warmup_steps="500" \
16
+ --evaluation_strategy="steps" \
17
+ --text_column_name="sentence" \
18
+ --pad_token="<pad>" \
19
+ --unk_token="<unk>" \
20
+ --save_steps="100" \
21
+ --eval_steps="100" \
22
+ --logging_steps="1" \
23
+ --layerdrop="0.0" \
24
+ --save_total_limit="3" \
25
+ --freeze_feature_extractor \
26
+ --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
27
+ --mask_time_prob="0.3" \
28
+ --mask_time_length="10" \
29
+ --mask_feature_prob="0.1" \
30
+ --mask_feature_length="64" \
31
+ --fp16 \
32
+ --group_by_length \
33
+ --do_eval \
34
+ --gradient_checkpointing \
35
+ --use_auth_token \
36
+ --do_train --do_eval \
37
+ --push_to_hub \