Helldivers2ASR_V3

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: 121.7825
• Wer: 0.0701

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: 7e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
989.1626	1.0	488	609.3144	0.5319
659.4745	2.0	976	441.8672	0.3931
514.4119	3.0	1464	382.5584	0.3219
419.7142	4.0	1952	288.8653	0.2560
341.4243	5.0	2440	281.8234	0.2205
288.763	6.0	2928	226.5483	0.1841
253.7379	7.0	3416	226.4153	0.1735
215.8471	8.0	3904	225.1913	0.1611
193.4694	9.0	4392	261.5921	0.1593
171.5698	10.0	4880	162.4843	0.1218
153.7561	11.0	5368	168.0577	0.1202
143.2963	12.0	5856	188.0740	0.1173
128.2614	13.0	6344	146.3703	0.0960
125.1001	14.0	6832	163.2122	0.0936
112.3151	15.0	7320	121.7357	0.0763
105.5073	16.0	7808	133.4259	0.0808
91.294	17.0	8296	134.9446	0.0827
94.5096	18.0	8784	109.5260	0.0794
83.6972	19.0	9272	163.8145	0.0909
78.8643	20.0	9760	102.4994	0.0665
72.3876	21.0	10248	102.5806	0.0676
71.9532	22.0	10736	109.1589	0.0666
73.799	23.0	11224	114.1257	0.0703
71.033	24.0	11712	121.2289	0.0781
66.3356	25.0	12200	109.2525	0.0666
64.8495	26.0	12688	108.1266	0.0657
60.262	27.0	13176	120.6045	0.0699
55.4432	28.0	13664	124.6207	0.0716
62.6622	29.0	14152	122.5060	0.0683
58.4588	30.0	14640	121.7825	0.0701

Framework versions

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