bambara-mms-10-hours-oza75bambara-asr-hf

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: 2.1148
• Wer: 0.5199
• Cer: 0.2465

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
1.5938	0.8313	500	1.2774	0.8294	0.3861
1.5142	1.6625	1000	1.2540	0.7729	0.3702
1.3973	2.4938	1500	1.2161	0.7125	0.3521
1.326	3.3250	2000	1.1719	0.7260	0.3426
1.2698	4.1563	2500	1.1710	0.6738	0.3350
1.218	4.9875	3000	1.0692	0.6507	0.3147
1.1672	5.8188	3500	1.0691	0.6408	0.3071
1.1307	6.6500	4000	1.0518	0.6405	0.3035
1.079	7.4813	4500	1.1091	0.6304	0.2972
1.0474	8.3126	5000	1.0662	0.6240	0.2994
1.0169	9.1438	5500	1.0669	0.6173	0.2937
0.9753	9.9751	6000	1.0039	0.6329	0.3027
0.9303	10.8063	6500	0.9909	0.6101	0.2863
0.8867	11.6376	7000	1.0058	0.5929	0.2888
0.8537	12.4688	7500	1.0321	0.6015	0.2859
0.8268	13.3001	8000	1.0427	0.5960	0.2808
0.7908	14.1313	8500	1.0816	0.5847	0.2791
0.7625	14.9626	9000	1.0817	0.5839	0.2748
0.7254	15.7938	9500	1.1185	0.5764	0.2768
0.6817	16.6251	10000	1.1043	0.5658	0.2749
0.6602	17.4564	10500	1.1736	0.5640	0.2708
0.6244	18.2876	11000	1.1894	0.5725	0.2721
0.6073	19.1189	11500	1.2478	0.5647	0.2737
0.5752	19.9501	12000	1.1698	0.5671	0.2770
0.5457	20.7814	12500	1.1864	0.5587	0.2697
0.5231	21.6126	13000	1.1907	0.5592	0.2693
0.5018	22.4439	13500	1.1874	0.5675	0.2707
0.4756	23.2751	14000	1.2167	0.5626	0.2664
0.458	24.1064	14500	1.2149	0.5592	0.2712
0.4301	24.9377	15000	1.3165	0.5499	0.2660
0.4165	25.7689	15500	1.2436	0.5689	0.2713
0.3897	26.6002	16000	1.3646	0.5470	0.2624
0.3743	27.4314	16500	1.4319	0.5477	0.2623
0.355	28.2627	17000	1.4645	0.5580	0.2683
0.3414	29.0939	17500	1.5373	0.5508	0.2590
0.3251	29.9252	18000	1.4997	0.5473	0.2648
0.3049	30.7564	18500	1.5231	0.5494	0.2617
0.2924	31.5877	19000	1.6212	0.5490	0.2610
0.2807	32.4190	19500	1.5959	0.5571	0.2661
0.2669	33.2502	20000	1.6300	0.5478	0.2628
0.2575	34.0815	20500	1.7042	0.5447	0.2616
0.2426	34.9127	21000	1.6750	0.5430	0.2607
0.2301	35.7440	21500	1.7449	0.5421	0.2596
0.2202	36.5752	22000	1.6587	0.5418	0.2584
0.213	37.4065	22500	1.7982	0.5361	0.2546
0.2092	38.2377	23000	1.7748	0.5271	0.2537
0.1956	39.0690	23500	1.8427	0.5310	0.2562
0.1869	39.9002	24000	1.7940	0.5275	0.2541
0.1799	40.7315	24500	1.7794	0.5275	0.2520
0.1727	41.5628	25000	1.9008	0.5374	0.2540
0.1681	42.3940	25500	1.9119	0.5297	0.2522
0.1596	43.2253	26000	1.9836	0.5258	0.2472
0.1569	44.0565	26500	1.9823	0.5195	0.2472
0.1516	44.8878	27000	1.9638	0.5179	0.2483
0.1482	45.7190	27500	2.0763	0.5146	0.2468
0.1379	46.5503	28000	2.0760	0.5234	0.2483
0.1407	47.3815	28500	2.0269	0.5220	0.2482
0.1331	48.2128	29000	2.0818	0.5221	0.2481
0.1328	49.0441	29500	2.0947	0.5205	0.2467
0.1308	49.8753	30000	2.1148	0.5199	0.2465

Framework versions

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