File size: 2,078 Bytes
8506a48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fd5771e
 
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
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
accelerate launch ./run_parler_tts_training.py \
    --model_name_or_path "parler-tts/parler_tts_mini_v0.1" \
    --feature_extractor_name "parler-tts/dac_44khZ_8kbps" \
    --description_tokenizer_name "parler-tts/parler_tts_mini_v0.1" \
    --prompt_tokenizer_name "parler-tts/parler_tts_mini_v0.1" \
    --report_to "wandb" \
    --overwrite_output_dir true \
    --train_dataset_name "sanchit-gandhi/expresso-concatenated-half-normal" \
    --train_metadata_dataset_name "sanchit-gandhi/expresso-concatenated-half-normal-tags-mistral" \
    --train_dataset_config_name "read" \
    --train_split_name "train[:-35]" \
    --eval_dataset_name "sanchit-gandhi/expresso-concatenated-half-normal" \
    --eval_metadata_dataset_name "sanchit-gandhi/expresso-concatenated-half-normal-tags-mistral" \
    --eval_dataset_config_name "read" \
    --eval_split_name "train" \
    --max_eval_samples 8 \
    --per_device_eval_batch_size 16 \
    --target_audio_column_name "audio" \
    --description_column_name "text_description" \
    --prompt_column_name "text" \
    --max_duration_in_seconds 30.0 \
    --min_duration_in_seconds 2.0 \
    --max_text_length 400 \
    --preprocessing_num_workers 2 \
    --do_train true \
    --num_train_epochs 8 \
    --gradient_accumulation_steps 8 \
    --gradient_checkpointing true \
    --per_device_train_batch_size 16 \
    --learning_rate 0.00008 \
    --adam_beta1 0.9 \
    --adam_beta2 0.99 \
    --weight_decay 0.01 \
    --lr_scheduler_type "cosine" \
    --warmup_steps 250 \
    --logging_steps 2 \
    --freeze_text_encoder true \
    --audio_encoder_per_device_batch_size 4 \
    --dtype "bfloat16" \
    --seed 456 \
    --output_dir "./" \
    --temporary_save_to_disk "../audio_code_tmp_concat/" \
    --save_to_disk "../tmp_dataset_audio_concat/" \
    --dataloader_num_workers 4 \
    --do_eval \
    --predict_with_generate \
    --include_inputs_for_metrics \
    --save_strategy "steps" \
    --save_steps 72 \
    --evaluation_strategy "epoch" \
    --save_total_limit 5 \
    --group_by_length true \
    --push_to_hub