mms-zeroshot-300m-bembaspeech-model

This model is a fine-tuned version of mms-meta/mms-zeroshot-300m on the BEMBASPEECH - BEM dataset. It achieves the following results on the evaluation set:

• Loss: 0.2038
• Wer: 0.4007

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.1405	200	2.2803	1.0
No log	0.2811	400	0.2945	0.5101
2.741	0.4216	600	0.2641	0.4834
2.741	0.5622	800	0.2611	0.4743
0.5962	0.7027	1000	0.2608	0.4831
0.5962	0.8433	1200	0.2477	0.4563
0.5962	0.9838	1400	0.2407	0.4567
0.536	1.1244	1600	0.2343	0.4412
0.536	1.2649	1800	0.2307	0.4423
0.5221	1.4055	2000	0.2252	0.4348
0.5221	1.5460	2200	0.2228	0.4326
0.5221	1.6866	2400	0.2162	0.4253
0.5027	1.8271	2600	0.2200	0.4188
0.5027	1.9677	2800	0.2131	0.4142
0.4818	2.1082	3000	0.2281	0.4281
0.4818	2.2488	3200	0.2178	0.4147
0.4818	2.3893	3400	0.2123	0.4155
0.4619	2.5299	3600	0.2142	0.4079
0.4619	2.6704	3800	0.2156	0.4011
0.464	2.8110	4000	0.2072	0.4009
0.464	2.9515	4200	0.2128	0.4013
0.464	3.0921	4400	0.2056	0.3982
0.4464	3.2326	4600	0.2038	0.4007
0.4464	3.3732	4800	0.2089	0.3987
0.4418	3.5137	5000	0.2043	0.4009
0.4418	3.6543	5200	0.2050	0.3964

Framework versions

• Transformers 4.46.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0