mms-bambara-5-hours-mali-asr-dataset

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

• Loss: 3.4671
• Wer: 0.5549
• Cer: 0.2722

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: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.7442	1.7241	500	1.5016	0.8074	0.3917
1.2377	3.4483	1000	1.4359	0.7090	0.3303
1.0648	5.1724	1500	1.6144	0.6935	0.3324
0.9677	6.8966	2000	1.5016	0.6696	0.3195
0.8607	8.6207	2500	1.5432	0.6492	0.3165
0.7663	10.3448	3000	1.7123	0.6522	0.3164
0.6906	12.0690	3500	1.7516	0.6208	0.3015
0.6025	13.7931	4000	1.7237	0.6187	0.3121
0.5379	15.5172	4500	1.8363	0.6310	0.3129
0.4772	17.2414	5000	1.8713	0.5894	0.2843
0.4267	18.9655	5500	2.0141	0.5962	0.2915
0.3759	20.6897	6000	2.0988	0.5882	0.2848
0.3404	22.4138	6500	2.2643	0.5826	0.2869
0.3042	24.1379	7000	2.4384	0.5733	0.2812
0.2825	25.8621	7500	2.3103	0.5718	0.2844
0.2543	27.5862	8000	2.1798	0.5724	0.2880
0.23	29.3103	8500	2.5892	0.5714	0.2843
0.2147	31.0345	9000	2.6667	0.5722	0.2822
0.1914	32.7586	9500	2.7395	0.5748	0.2812
0.1794	34.4828	10000	2.8872	0.5802	0.2847
0.1675	36.2069	10500	2.7069	0.5690	0.2827
0.1493	37.9310	11000	2.8134	0.5705	0.2840
0.1386	39.6552	11500	3.0683	0.5615	0.2771
0.1237	41.3793	12000	3.2212	0.5567	0.2753
0.117	43.1034	12500	3.2128	0.5593	0.2703
0.1082	44.8276	13000	3.2066	0.5562	0.2732
0.0978	46.5517	13500	3.4042	0.5551	0.2720
0.0927	48.2759	14000	3.4410	0.5541	0.2723
0.0915	50.0	14500	3.4671	0.5549	0.2722

Framework versions

• Transformers 4.45.1
• Pytorch 2.1.0+cu118
• Datasets 2.17.0
• Tokenizers 0.20.3