finetuning_common_voice.json

{
  "model_name_or_path": "facebook/wav2vec2-large-xlsr-53",
  "dataset_name": "dataset/common_voice",
  "dataset_config_name": "lg",
  "dataset_data_dir": "/content/data/cv-corpus-7.0-2021-07-21",
  "dataset_min_filesize": 10000,
  "dataset_min_textlength": 10,
  "output_dir": "output/wav2vec2-common_voice-lg",
  "overwrite_output_dir": true,
  "num_train_epochs": 200,
  "per_device_eval_batch_size": 8,
  "per_device_train_batch_size": 8,
  "evaluation_strategy": "steps",
  "learning_rate": 1e-4,
  "warmup_steps": 300,
  "fp16": true,
  "freeze_feature_extractor": true,
  "save_steps": 2000,
  "eval_steps": 2000,
  "save_total_limit": 1,
  "logging_steps": 2000,
  "group_by_length": true,
  "feat_proj_dropout": 0.04,
  "layerdrop": 0.041,
  "attention_dropout": 0.094,
  "activation_dropout": 0.055,
  "hidden_dropout": 0.047,
  "mask_time_prob": 0.4,
  "do_train": true,
  "do_eval": true,
  "gradient_accumulation_steps": 2,
  "dataloader_num_workers": 8,
  "train_split_name": "train+validation+other+invalidated",
  "chars_to_ignore": [",", "?", ".", "!", "-", ";", ":", "\"\"", "%", "'", "\"", "�", "'", "\u2018", "\u2019"]
}