asr-africa/mms-1B_all_DigitalUmuganda_Afrivoice_Shona_1hr_v1

facebook/mms-1b-all

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

• Loss: 9.7260
• Model Preparation Time: 0.0108
• Wer: 1.0
• Cer: 1.0263

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 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: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer	Cer
61.2558	0.8696	5	14.2292	0.0108	1.0	2.1395
60.5933	1.8261	10	14.0103	0.0108	1.0	2.0649
59.6126	2.7826	15	13.5539	0.0108	1.0	1.9202
46.7809	3.9130	21	12.4379	0.0108	1.0	1.6586
51.2006	4.8696	26	11.1727	0.0108	1.0	1.3341
45.5066	5.8261	31	9.7087	0.0108	0.9998	0.9584
38.5399	6.7826	36	8.2230	0.0108	1.0	0.8930
26.7322	7.9130	42	6.7125	0.0108	1.0	0.9583
27.0689	8.8696	47	5.8420	0.0108	1.0	0.9882
23.885	9.8261	52	5.3230	0.0108	1.0	0.9931
22.2475	10.7826	57	5.0217	0.0108	1.0	0.9941
17.7159	11.9130	63	4.8187	0.0108	1.0	0.9934
20.6729	12.8696	68	4.7208	0.0108	1.0	0.9943
20.2555	13.8261	73	4.6241	0.0108	1.0	0.9943
19.8974	14.7826	78	4.5231	0.0108	1.0	0.9915
16.2179	15.9130	84	4.3843	0.0108	1.0	0.9906
18.2735	16.8696	89	4.2386	0.0108	1.0	0.9658
18.3992	17.8261	94	4.0855	0.0108	1.0	0.9008
17.9017	18.7826	99	3.9638	0.0108	1.0012	0.8728
14.4746	19.9130	105	3.8507	0.0108	1.0	0.8665
17.0257	20.8696	110	3.7610	0.0108	1.0	0.8752
16.6174	21.8261	115	3.6873	0.0108	1.0058	0.8401
16.2618	22.7826	120	3.6239	0.0108	1.0	0.8587
13.3621	23.9130	126	3.5762	0.0108	1.0	0.8542
15.8358	24.8696	131	3.5434	0.0108	1.0119	0.8190
15.5278	25.8261	136	3.5110	0.0108	1.0010	0.8433
15.2012	26.7826	141	3.4711	0.0108	1.0032	0.8368
12.5298	27.9130	147	3.4341	0.0108	1.0066	0.8298
14.8525	28.8696	152	3.4105	0.0108	1.0085	0.8298
14.5351	29.8261	157	3.3788	0.0108	1.0268	0.8057
14.4403	30.7826	162	3.3549	0.0108	1.0039	0.8339
11.836	31.9130	168	3.3474	0.0108	1.0051	0.8313
14.2697	32.8696	173	3.2991	0.0108	1.0482	0.7892
13.8508	33.8261	178	3.3245	0.0108	1.0029	0.8278
13.6973	34.7826	183	3.2907	0.0108	1.0005	0.8051
11.2833	35.9130	189	3.3046	0.0108	1.0100	0.8228
13.3111	36.8696	194	3.2605	0.0108	1.0153	0.8002
13.2858	37.8261	199	3.2802	0.0108	1.0075	0.8115
13.3125	38.7826	204	3.2877	0.0108	1.0107	0.8076
10.811	39.9130	210	3.2335	0.0108	1.0083	0.7995
12.9063	40.8696	215	3.2435	0.0108	1.0071	0.8072
12.8105	41.8261	220	3.2516	0.0108	1.0054	0.8146
12.6894	42.7826	225	3.2634	0.0108	1.0199	0.7984
10.4284	43.9130	231	3.2434	0.0108	1.0095	0.7988
12.5526	44.8696	236	3.2996	0.0108	1.0039	0.8245
12.4532	45.8261	241	3.2234	0.0108	1.0231	0.7907
12.4143	46.7826	246	3.2599	0.0108	1.0029	0.8155
10.2517	47.9130	252	3.2585	0.0108	1.0063	0.8053
12.4357	48.8696	257	3.2407	0.0108	1.0236	0.7921
12.2622	49.8261	262	3.2640	0.0108	1.0039	0.8084
12.1975	50.7826	267	3.2645	0.0108	1.0117	0.8033
10.1193	51.9130	273	3.2395	0.0108	1.0285	0.7824
12.0931	52.8696	278	3.2823	0.0108	1.0034	0.8013
12.2157	53.8261	283	3.2564	0.0108	1.0124	0.7909
12.0174	54.7826	288	3.2492	0.0108	1.0294	0.7742
9.9986	55.9130	294	3.2344	0.0108	1.0032	0.7960
11.9855	56.8696	299	3.2709	0.0108	1.0214	0.7785
11.9514	57.8261	304	3.2556	0.0108	1.0095	0.7893
11.9152	58.7826	309	3.2716	0.0108	1.0090	0.7857
9.8905	59.9130	315	3.3192	0.0108	1.0165	0.7794
11.9491	60.8696	320	3.2668	0.0108	1.0212	0.7717
11.8561	61.8261	325	3.2821	0.0108	1.0151	0.7769
11.8206	62.7826	330	3.2512	0.0108	1.0370	0.7625
9.8339	63.9130	336	3.3038	0.0108	1.0114	0.7772
11.7774	64.8696	341	3.2527	0.0108	1.0365	0.7566
11.7058	65.8261	346	3.3124	0.0108	1.0190	0.7691
11.7123	66.7826	351	3.2726	0.0108	1.0246	0.7655
9.7587	67.9130	357	3.2848	0.0108	1.0190	0.7630
11.7171	68.8696	362	3.2727	0.0108	1.0187	0.7622
11.6455	69.8261	367	3.2705	0.0108	1.0148	0.7609
11.6714	70.7826	372	3.2773	0.0108	1.0190	0.7566
9.6664	71.9130	378	3.2537	0.0108	1.0314	0.7496
11.6148	72.8696	383	3.2945	0.0108	1.0151	0.7616
11.5807	73.8261	388	3.3068	0.0108	1.0202	0.7563
11.5939	74.7826	393	3.2805	0.0108	1.0258	0.7471
9.63	75.9130	399	3.3203	0.0108	1.0124	0.7556
11.5625	76.8696	404	3.2776	0.0108	1.0158	0.7530
11.5198	77.8261	409	3.3043	0.0108	1.0209	0.7513
11.5147	78.7826	414	3.3050	0.0108	1.0195	0.7519
9.58	79.9130	420	3.2798	0.0108	1.0207	0.7472
11.4839	80.8696	425	3.2974	0.0108	1.0268	0.7435
11.4907	81.8261	430	3.2751	0.0108	1.0243	0.7434
11.4485	82.7826	435	3.3051	0.0108	1.0097	0.7520
9.5367	83.9130	441	3.2787	0.0108	1.0248	0.7407
11.4168	84.8696	446	3.3047	0.0108	1.0158	0.7476
11.4169	85.8261	451	3.2901	0.0108	1.0250	0.7394
11.3964	86.7826	456	3.3415	0.0108	1.0168	0.7476
9.4842	87.9130	462	3.2788	0.0108	1.0270	0.7371
11.3778	88.8696	467	3.3244	0.0108	1.0243	0.7421
11.3442	89.8261	472	3.3127	0.0108	1.0340	0.7390
11.3428	90.7826	477	3.3120	0.0108	1.0209	0.7412
9.4492	91.9130	483	3.3177	0.0108	1.0195	0.7418
11.3562	92.8696	488	3.2997	0.0108	1.0224	0.7367
11.3161	93.8261	493	3.3567	0.0108	1.0180	0.7448
11.2836	94.7826	498	3.2956	0.0108	1.0263	0.7358
9.404	95.9130	504	3.3472	0.0108	1.0131	0.7449
11.2805	96.8696	509	3.3002	0.0108	1.0297	0.7322
11.2702	97.8261	514	3.3282	0.0108	1.0136	0.7417
11.2554	98.7826	519	3.3361	0.0108	1.0161	0.7388
9.3581	99.9130	525	3.3132	0.0108	1.0328	0.7329
11.2385	100.8696	530	3.3154	0.0108	1.0224	0.7349
11.2548	101.8261	535	3.3215	0.0108	1.0272	0.7323
11.2565	102.7826	540	3.3022	0.0108	1.0248	0.7312
9.3435	103.9130	546	3.3193	0.0108	1.0173	0.7353
11.2047	104.8696	551	3.3239	0.0108	1.0250	0.7335
11.2158	105.8261	556	3.3206	0.0108	1.0306	0.7310
11.1969	106.7826	561	3.3349	0.0108	1.0231	0.7335
9.2975	107.9130	567	3.3312	0.0108	1.0158	0.7346
11.1944	108.8696	572	3.3321	0.0108	1.0214	0.7330
11.1897	109.8261	577	3.3160	0.0108	1.0399	0.7269
11.1698	110.7826	582	3.3406	0.0108	1.0175	0.7347
9.2901	111.9130	588	3.3186	0.0108	1.0178	0.7316
11.1881	112.8696	593	3.3042	0.0108	1.0340	0.7262
11.1726	113.8261	598	3.3469	0.0108	1.0231	0.7342
11.0822	114.7826	603	3.3216	0.0108	1.0275	0.7289
9.2644	115.9130	609	3.3320	0.0108	1.0224	0.7311
11.1101	116.8696	614	3.3357	0.0108	1.0263	0.7300
11.1348	117.8261	619	3.3371	0.0108	1.0231	0.7314
11.0795	118.7826	624	3.3338	0.0108	1.0248	0.7298
9.242	119.9130	630	3.3327	0.0108	1.0260	0.7292
11.1011	120.8696	635	3.3183	0.0108	1.0255	0.7283
11.0889	121.8261	640	3.3466	0.0108	1.0204	0.7338
11.0732	122.7826	645	3.3447	0.0108	1.0287	0.7314
9.2586	123.9130	651	3.3184	0.0108	1.0348	0.7265
11.0636	124.8696	656	3.3406	0.0108	1.0302	0.7296
11.0775	125.8261	661	3.3429	0.0108	1.0302	0.7299
11.064	126.7826	666	3.3512	0.0108	1.0236	0.7310
9.2299	127.9130	672	3.3380	0.0108	1.0292	0.7269
11.0375	128.8696	677	3.3346	0.0108	1.0340	0.7267
11.0343	129.8261	682	3.3309	0.0108	1.0316	0.7256
11.0733	130.7826	687	3.3285	0.0108	1.0292	0.7275
9.2115	131.9130	693	3.3336	0.0108	1.0255	0.7283
11.0232	132.8696	698	3.3391	0.0108	1.0243	0.7294
11.0263	133.8261	703	3.3352	0.0108	1.0258	0.7280
11.0658	134.7826	708	3.3274	0.0108	1.0292	0.7272
9.1985	135.9130	714	3.3292	0.0108	1.0297	0.7284
11.0514	136.8696	719	3.3325	0.0108	1.0277	0.7287
11.0376	137.8261	724	3.3337	0.0108	1.0270	0.7287
11.0298	138.7826	729	3.3398	0.0108	1.0260	0.7294
9.1757	139.9130	735	3.3446	0.0108	1.0265	0.7294
11.0135	140.8696	740	3.3472	0.0108	1.0248	0.7301
11.0265	141.8261	745	3.3432	0.0108	1.0258	0.7286
9.4511	142.7826	750	3.3405	0.0108	1.0268	0.7286

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.1.0+cu118
• Datasets 3.0.2
• Tokenizers 0.20.1