ws_whisper_small_distil_att_loss_mozilla_epochs_50_batch_2_try3

This model is a fine-tuned version of rohitp1/kkkh_whisper_small_distillation_att_loss_mozilla_epochs_100_batch_4_concat_dataset on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.8151
• Wer: 19.8454

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 1024
• total_train_batch_size: 2048
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.147	1.46	100	1.6545	16.9156
1.1459	2.93	200	1.6637	16.9591
1.1433	4.39	300	1.6806	16.9069
1.4547	5.85	400	2.3254	20.3259
2.4311	7.31	500	3.9239	19.4764
2.5513	8.78	600	2.7875	19.4555
2.2862	10.24	700	3.7611	19.7410
2.3854	11.7	800	2.8151	19.8454

Framework versions

• Transformers 4.29.2
• Pytorch 1.12.1
• Datasets 2.7.1
• Tokenizers 0.11.0