wav2vec2-large-mms-1b-aft-koo

This model is a fine-tuned version of facebook/mms-1b-all on the audiofolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.8230
• Wer: 0.7010

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: 0.001
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
10.465	0.1153	100	1.2384	0.9140
1.9435	0.2307	200	1.0650	0.8178
5.045	0.3460	300	1.0001	0.8112
1.0003	0.4614	400	1.0103	0.7845
0.9484	0.5767	500	0.9683	0.7849
1.0083	0.6920	600	0.9821	0.7719
0.8022	0.8074	700	1.0747	0.7729
1.323	0.9227	800	0.9499	0.7742
2.3287	1.0381	900	0.9352	0.7529
1.1491	1.1534	1000	0.9849	0.7607
0.9295	1.2687	1100	0.8864	0.7457
0.8009	1.3841	1200	0.8516	0.7473
0.7171	1.4994	1300	0.9102	0.7548
0.9572	1.6148	1400	0.8405	0.7802
1.0648	1.7301	1500	0.8903	0.7568
0.8783	1.8454	1600	0.9100	0.7448
0.7615	1.9608	1700	0.9209	0.7355
0.9199	2.0761	1800	0.9363	0.7306
0.8082	2.1915	1900	0.8896	0.7399
0.9272	2.3068	2000	0.8735	0.7405
0.7382	2.4221	2100	0.8315	0.7397
1.1308	2.5375	2200	0.8300	0.7436
0.7261	2.6528	2300	0.8903	0.7401
0.7848	2.7682	2400	0.8585	0.7335
0.7409	2.8835	2500	0.8192	0.7343
0.791	2.9988	2600	0.8765	0.7324
0.8917	3.1142	2700	0.8125	0.7361
0.8859	3.2295	2800	0.8479	0.7148
1.0343	3.3449	2900	0.8907	0.7027
0.8133	3.4602	3000	0.8621	0.7115
0.6825	3.5755	3100	0.8508	0.7091
0.6756	3.6909	3200	0.8234	0.7117
0.746	3.8062	3300	0.8396	0.7186
0.6916	3.9216	3400	0.8761	0.7204
0.8504	4.0369	3500	0.8581	0.7068
0.7924	4.1522	3600	0.8291	0.7101
0.8869	4.2676	3700	0.8200	0.7051
0.9182	4.3829	3800	0.8582	0.6944
0.6489	4.4983	3900	0.8426	0.7041
0.6127	4.6136	4000	0.8397	0.6995
0.6857	4.7290	4100	0.8232	0.7043
0.7728	4.8443	4200	0.8095	0.7000
0.704	4.9596	4300	0.8230	0.7010

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.1.1
• Tokenizers 0.22.1