Wav2vec2.0-xlsr-53 model is fine-tuned on commonvoice russian dataset
Configs (yaml)
checkpoint: save_interval: 1000 \n save_interval_updates: 1000 \n keep_interval_updates: 1 no_epoch_ckechpoints: true best_checkpoint_metric: wer task: _name: audio_finetuning normalize: true labels: phn dataset: ---num_workers: 6 ---max_tokens: 800000 ---skip_invalid_size_inputs_valid_test: true ---valid_subset: valie -distributed_training: ---ddp_backend: legacy_ddp ---distributed_world_size: 4 -criterion: ---_name: ctc ---zero_infinity: true -optimization: ---max_update: 25000 ---lr: [0.00001] ---sentence_avg: true ---update_freq: [4] -optimizer: ---_name: adam ---adam_betas: (0.9, 0.98) ---adam_eps: 1e-8 -lr_scheduler: ---_name: tri_stage ---phase_ratio: [0.1, 0.4, 0.5] ---final_lr_scale: 0.05 -model: ---_name: wav2vec_ctc ---apply_mask: true ---mask_prob: 0.5 ---mask_channel_prob: 0.1 ---mask_channel_length: 64 ---layerdrop: 0.1 ---activation_dropout: 0.1 ---feature_grad_mult: 0.0 ---freeze_finetune_updates: 0