helldivers2-jarvis-asrV10

This model is a fine-tuned version of facebook/wav2vec2-base-960h on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 6.5706
• Wer: 0.1743
• Cer: 0.7734

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: 2e-05
• train_batch_size: 15
• eval_batch_size: 15
• seed: 42
• optimizer: Use OptimizerNames.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: 50
• num_epochs: 80

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
522.7094	1.0	77	118.2986	0.3201	0.7856
322.2706	2.0	154	75.9560	0.2855	0.7808
263.1068	3.0	231	57.7799	0.2608	0.7790
228.0345	4.0	308	49.6626	0.2349	0.7777
205.1745	5.0	385	40.2569	0.2188	0.7765
175.9148	6.0	462	28.7668	0.2027	0.7752
166.6824	7.0	539	24.4211	0.2015	0.7751
153.7703	8.0	616	22.5456	0.2002	0.7751
155.0527	9.0	693	29.8641	0.2002	0.7756
131.393	10.0	770	20.1614	0.1904	0.7746
118.7684	11.0	847	15.5499	0.1842	0.7741
108.6234	12.0	924	13.7057	0.1805	0.7739
114.8896	13.0	1001	13.1512	0.1817	0.7740
115.3319	14.0	1078	11.9990	0.1842	0.7743
103.4016	15.0	1155	14.2569	0.1829	0.7740
101.8496	16.0	1232	12.9236	0.1780	0.7740
86.0507	17.0	1309	8.3784	0.1792	0.7737
82.5003	18.0	1386	10.2976	0.1768	0.7736
97.0317	19.0	1463	8.8326	0.1743	0.7735
89.896	20.0	1540	8.7586	0.1743	0.7735
81.7516	21.0	1617	6.4406	0.1731	0.7735
79.7907	22.0	1694	6.6042	0.1743	0.7734
80.7192	23.0	1771	7.7004	0.1731	0.7734
81.5526	24.0	1848	8.4363	0.1743	0.7734
74.399	25.0	1925	6.8051	0.1731	0.7734
75.8625	26.0	2002	6.5706	0.1743	0.7734

Framework versions

• Transformers 4.51.3
• Pytorch 2.5.1+cu121
• Datasets 3.6.0
• Tokenizers 0.21.1