wav2vec2-xls-r-300m-Fleurs_AMMI_AFRIVOICE_LRSC-ln-1hrs-v1

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.8424
• Wer: 0.7133
• Cer: 0.2385

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: 4
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• 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
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
7.9233	1.0	39	2.9633	1.0	1.0
2.8667	2.0	78	2.8788	1.0	1.0
2.9406	3.0	117	2.8635	1.0	1.0
2.8363	4.0	156	2.8408	1.0	1.0
2.8262	5.0	195	2.8456	1.0	1.0
2.84	6.0	234	2.8458	1.0	1.0
2.8195	7.0	273	2.8430	1.0	1.0
2.8204	8.0	312	2.8129	1.0	1.0
2.804	9.0	351	2.8006	1.0	1.0
2.7869	10.0	390	2.8017	1.0	1.0
2.7856	11.0	429	2.7912	1.0	1.0
2.765	12.0	468	2.7441	1.0	1.0
2.7333	13.0	507	2.6953	1.0	1.0
2.6093	14.0	546	2.5490	1.0	1.0
2.4857	15.0	585	2.3470	1.0	1.0
2.2251	16.0	624	2.0507	1.0	0.6812
1.9855	17.0	663	1.8561	1.0	0.6490
1.7317	18.0	702	1.6075	1.0	0.5509
1.5049	19.0	741	1.5265	0.9993	0.5068
1.4006	20.0	780	1.4114	0.9426	0.4247
1.2416	21.0	819	1.3780	0.9512	0.4098
1.1441	22.0	858	1.3123	0.8908	0.3474
1.0138	23.0	897	1.2678	0.8808	0.3405
0.9221	24.0	936	1.2555	0.8739	0.3365
0.8665	25.0	975	1.2717	0.8634	0.3380
0.8189	26.0	1014	1.1927	0.8557	0.3150
0.712	27.0	1053	1.2620	0.8260	0.3013
0.6586	28.0	1092	1.2498	0.8221	0.3015
0.6092	29.0	1131	1.1750	0.8127	0.2934
0.5484	30.0	1170	1.2835	0.8137	0.2887
0.504	31.0	1209	1.2157	0.8175	0.2967
0.5021	32.0	1248	1.2737	0.8099	0.2937
0.4412	33.0	1287	1.2819	0.7946	0.2831
0.3965	34.0	1326	1.3634	0.8043	0.2827
0.4001	35.0	1365	1.3748	0.8146	0.2922
0.3519	36.0	1404	1.3540	0.8053	0.2777
0.36	37.0	1443	1.3906	0.7995	0.2891
0.3035	38.0	1482	1.4942	0.8291	0.2811
0.311	39.0	1521	1.4371	0.7835	0.2798
0.2896	40.0	1560	1.4310	0.7827	0.2698
0.2588	41.0	1599	1.5232	0.7824	0.2760
0.2495	42.0	1638	1.4698	0.7726	0.2663
0.2527	43.0	1677	1.4726	0.7628	0.2705
0.221	44.0	1716	1.5491	0.7570	0.2659
0.2203	45.0	1755	1.5355	0.7699	0.2687
0.2133	46.0	1794	1.5278	0.8099	0.2722
0.2062	47.0	1833	1.5152	0.7555	0.2605
0.1936	48.0	1872	1.5899	0.7680	0.2680
0.1865	49.0	1911	1.5933	0.7626	0.2610
0.2022	50.0	1950	1.6041	0.7534	0.2579
0.1913	51.0	1989	1.5845	0.7476	0.2597
0.1784	52.0	2028	1.6421	0.7620	0.2628
0.1696	53.0	2067	1.7008	0.7569	0.2619
0.156	54.0	2106	1.6896	0.7616	0.2593
0.1424	55.0	2145	1.6266	0.7510	0.2597
0.1474	56.0	2184	1.7046	0.7394	0.2545
0.147	57.0	2223	1.7005	0.7419	0.2566
0.1752	58.0	2262	1.6940	0.7476	0.2533
0.1338	59.0	2301	1.7020	0.7447	0.2529
0.1352	60.0	2340	1.6386	0.7406	0.2528
0.1164	61.0	2379	1.7665	0.7637	0.2544
0.1307	62.0	2418	1.7711	0.7396	0.2513
0.117	63.0	2457	1.7694	0.7296	0.2479
0.1196	64.0	2496	1.7590	0.7368	0.2506
0.1125	65.0	2535	1.7932	0.7405	0.2493
0.1155	66.0	2574	1.7361	0.7390	0.2490
0.123	67.0	2613	1.7788	0.7343	0.2498
0.1163	68.0	2652	1.7992	0.7275	0.2476
0.1075	69.0	2691	1.7817	0.7366	0.2483
0.1007	70.0	2730	1.7929	0.7397	0.2504
0.0987	71.0	2769	1.8354	0.7320	0.2491
0.1026	72.0	2808	1.8047	0.7266	0.2463
0.1038	73.0	2847	1.8150	0.7319	0.2482
0.1004	74.0	2886	1.8303	0.7244	0.2446
0.0983	75.0	2925	1.7693	0.7256	0.2456
0.0961	76.0	2964	1.7943	0.7177	0.2432
0.1034	77.0	3003	1.8095	0.7270	0.2450
0.1	78.0	3042	1.8075	0.7348	0.2444
0.0988	79.0	3081	1.7991	0.7226	0.2431
0.0867	80.0	3120	1.8438	0.7316	0.2441
0.0951	81.0	3159	1.8312	0.7284	0.2430
0.103	82.0	3198	1.8002	0.7208	0.2419
0.104	83.0	3237	1.8522	0.7229	0.2444
0.0874	84.0	3276	1.8207	0.7222	0.2416
0.087	85.0	3315	1.8349	0.7176	0.2400
0.0861	86.0	3354	1.8445	0.7229	0.2405
0.0845	87.0	3393	1.8139	0.7173	0.2393
0.0841	88.0	3432	1.8183	0.7160	0.2406
0.0723	89.0	3471	1.8424	0.7151	0.2406
0.0791	90.0	3510	1.8458	0.7187	0.2410
0.0802	91.0	3549	1.8563	0.7205	0.2403
0.0871	92.0	3588	1.8370	0.7124	0.2391
0.0852	93.0	3627	1.8476	0.7163	0.2389
0.0821	94.0	3666	1.8523	0.7130	0.2387
0.0841	95.0	3705	1.8473	0.7139	0.2385
0.0746	96.0	3744	1.8415	0.7130	0.2384
0.0888	97.0	3783	1.8423	0.7133	0.2383
0.124	97.4416	3800	1.8424	0.7133	0.2385

Framework versions

• Transformers 4.48.2
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0