File size: 938 Bytes
1a10833
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
python run_speech_recognition_ctc.py \
    --dataset_name="mozilla-foundation/common_voice_8_0" \
    --dataset_config_name="nl" \
    --text_column_name="sentence" \
    --length_column_name="input_length" \
    --use_auth_token \
    --model_name_or_path="/workspace/hug_data/xls-r-2b-22-to-16.encoder.ctc48" \
    --tokenizer_name_or_path="./" \
    --output_dir="./output.step1" \
    --overwrite_output_dir \
    --num_train_epochs="5" \
    --per_device_train_batch_size="8" \
    --gradient_accumulation_steps="4" \
    --learning_rate="5e-5" \
    --warmup_steps="2000" \
    --evaluation_strategy="steps" \
    --save_steps="500" \
    --eval_steps="500" \
    --save_total_limit="4" \
    --layerdrop="0.0" \
    --freeze_feature_encoder \
    --gradient_checkpointing \
    --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � _ \
    --fp16 \
    --group_by_length \
    --do_train --do_eval
    
#	--overwrite_output_dir \