Helldivers2ASR_V5

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

• Loss: 124.4297
• Wer: 0.0946
• Cer: 0.0385

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: 0.0005
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2695.3033	0.9655	14	1193.2163	0.5150	0.2548
1690.871	2.0	29	1009.7814	0.4355	0.2077
1449.3326	2.9655	43	866.1747	0.3603	0.1693
1157.594	4.0	58	744.7312	0.3274	0.1547
1123.9697	4.9655	72	656.5463	0.3073	0.1384
967.6211	6.0	87	561.6840	0.2915	0.1309
926.493	6.9655	101	552.4729	0.2851	0.1334
810.3585	8.0	116	489.7581	0.2328	0.1091
802.5302	8.9655	130	477.9092	0.2235	0.0998
675.7667	10.0	145	429.9390	0.2034	0.0897
688.4148	10.9655	159	412.7467	0.2350	0.1008
593.3937	12.0	174	355.4245	0.1991	0.0865
594.2962	12.9655	188	329.0578	0.1819	0.0774
508.8992	14.0	203	350.7224	0.1891	0.0800
503.765	14.9655	217	300.9240	0.1705	0.0714
452.241	16.0	232	326.3430	0.1605	0.0698
474.4419	16.9655	246	292.7792	0.1691	0.0723
404.4718	18.0	261	284.8404	0.1633	0.0706
403.7702	18.9655	275	261.7959	0.1476	0.0650
381.9254	20.0	290	224.7344	0.1490	0.0630
367.5338	20.9655	304	280.7520	0.1540	0.0646
336.0608	22.0	319	226.6506	0.1554	0.0650
343.7045	22.9655	333	231.6033	0.1590	0.0697
279.5685	24.0	348	246.9725	0.1590	0.0686
311.3864	24.9655	362	204.4507	0.1511	0.0635
255.0848	26.0	377	194.7101	0.1289	0.0565
271.3048	26.9655	391	190.2275	0.1254	0.0533
258.5749	28.0	406	193.0812	0.1189	0.0483
252.6549	28.9655	420	182.3211	0.1232	0.0473
198.5987	30.0	435	180.7149	0.1153	0.0480
193.8807	30.9655	449	169.1754	0.1196	0.0480
184.148	32.0	464	155.5871	0.1139	0.0474
185.8568	32.9655	478	160.6432	0.1103	0.0474
182.5008	34.0	493	153.4942	0.1125	0.0486
182.8014	34.9655	507	148.8764	0.1096	0.0466
152.4978	36.0	522	141.8686	0.1125	0.0484
159.9153	36.9655	536	130.8442	0.1117	0.0463
153.8542	38.0	551	142.0020	0.1053	0.0435
194.8882	38.9655	565	128.6881	0.1032	0.0410
159.338	40.0	580	125.0831	0.0981	0.0408
168.303	40.9655	594	131.1471	0.0953	0.0391
149.8229	42.0	609	127.8085	0.0946	0.0394
161.5126	42.9655	623	124.2411	0.0960	0.0393
163.1556	44.0	638	123.0042	0.0967	0.0395
158.075	44.9655	652	124.5911	0.0938	0.0387
151.277	46.0	667	124.5658	0.0938	0.0382
145.2503	46.9655	681	124.3450	0.0938	0.0387
142.6398	48.0	696	124.3733	0.0938	0.0385
145.2239	48.2759	700	124.4297	0.0946	0.0385

Framework versions

• Transformers 4.44.0
• Pytorch 2.5.1+cu121
• Datasets 3.6.0
• Tokenizers 0.19.1