malagasy-asr

This model is a fine-tuned version of facebook/mms-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1440
• Wer: 0.1571

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use 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: 0.1
• training_steps: 2500

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
28.2487	0.1128	100	3.2587	1.0
11.1227	0.2255	200	2.6702	1.0
8.0281	0.3383	300	1.1044	0.9421
2.3961	0.4511	400	0.5071	0.5568
1.3243	0.5639	500	0.3686	0.4129
0.9699	0.6766	600	0.2877	0.3233
0.9166	0.7894	700	0.2549	0.2900
0.7583	0.9022	800	0.2285	0.2677
0.7533	1.0147	900	0.2171	0.2392
0.6738	1.1274	1000	0.2087	0.2246
0.5797	1.2402	1100	0.2014	0.2275
0.5635	1.3530	1200	0.1934	0.2154
0.5428	1.4657	1300	0.1845	0.2064
0.5114	1.5785	1400	0.1777	0.2050
0.4973	1.6913	1500	0.1743	0.2050
0.5309	1.8041	1600	0.1703	0.1920
0.4796	1.9168	1700	0.1624	0.1861
0.4433	2.0293	1800	0.1628	0.1746
0.3442	2.1421	1900	0.1568	0.1697
0.3531	2.2549	2000	0.1548	0.1707
0.4047	2.3676	2100	0.1524	0.1660
0.3403	2.4804	2200	0.1510	0.1626
0.3958	2.5932	2300	0.1474	0.1612
0.3272	2.7059	2400	0.1451	0.1576
0.3197	2.8187	2500	0.1440	0.1571

Framework versions

• Transformers 5.0.0.dev0
• Pytorch 2.9.0+cu126
• Datasets 4.4.2
• Tokenizers 0.22.2