speecht5_finetuned_librispeech_polish_epo6_batch2_gas2

This model is a fine-tuned version of dawid511/speecht5_finetuned_librispeech_polish_epo3_batch8_gas4 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3612

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: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 6
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.7252	0.2558	100	0.3732
0.7795	0.5115	200	0.3882
0.7822	0.7673	300	0.3806
0.7795	1.0230	400	0.3819
0.7723	1.2788	500	0.3776
0.7567	1.5345	600	0.3782
0.7719	1.7903	700	0.3794
0.7775	2.0460	800	0.3737
0.7635	2.3018	900	0.3744
0.7613	2.5575	1000	0.3751
0.7519	2.8133	1100	0.3714
0.7514	3.0691	1200	0.3697
0.7461	3.3248	1300	0.3711
0.7462	3.5806	1400	0.3708
0.7407	3.8363	1500	0.3678
0.7309	4.0921	1600	0.3691
0.7243	4.3478	1700	0.3670
0.7218	4.6036	1800	0.3662
0.7312	4.8593	1900	0.3645
0.7086	5.1151	2000	0.3636
0.7339	5.3708	2100	0.3637
0.718	5.6266	2200	0.3623
0.7171	5.8824	2300	0.3612

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0