test-wav2vec2-darija-marocain-aya-29000

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 54.1290
• Wer: 1.0
• Cer: 1.0

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
27.3086	0.0364	500	54.1290	1.0	1.0
10.8536	0.0728	1000	25.7240	1.0	1.0
5.6544	0.1091	1500	8.7983	1.0	1.0
4.0069	0.1455	2000	4.8197	1.0	1.0
3.6432	0.1819	2500	4.4943	1.0	1.0
3.477	0.2183	3000	4.4012	1.0	1.0
3.4495	0.2546	3500	3.9120	1.0	1.0
3.3688	0.2910	4000	3.7199	1.0	1.0
3.2459	0.3274	4500	3.5998	1.0	1.0
3.2401	0.3638	5000	3.5128	1.0	1.0
3.4504	0.4001	5500	3.3765	1.0	1.0
3.142	0.4365	6000	3.1360	1.0	1.0
3.0648	0.4729	6500	2.8708	1.0	0.9982
2.8215	0.5093	7000	2.5534	1.0	0.9350
2.5345	0.5457	7500	2.2860	0.9995	0.8187
2.5151	0.5820	8000	2.0754	0.9968	0.6671
2.3124	0.6184	8500	1.8855	0.9969	0.6009
2.0244	0.6548	9000	1.7425	0.9939	0.5501
2.1518	0.6912	9500	1.6193	0.9880	0.5155
2.1723	0.7275	10000	1.5097	0.9844	0.4980
1.9636	0.7639	10500	1.4213	0.9727	0.4651
1.7568	0.8003	11000	1.3623	0.9549	0.4281
1.7105	0.8367	11500	1.3058	0.9296	0.4012
1.739	0.8730	12000	1.2470	0.9163	0.3894
1.6762	0.9094	12500	1.2088	0.9095	0.3864
1.5635	0.9458	13000	1.1740	0.9033	0.3727
1.3595	0.9822	13500	1.1522	0.8795	0.3645
1.8905	1.0186	14000	1.1483	0.8725	0.3605
1.5877	1.0549	14500	1.0925	0.8755	0.3568
1.8459	1.0913	15000	1.0628	0.8630	0.3521
1.4473	1.1277	15500	1.1005	0.8560	0.3515
1.7997	1.1641	16000	1.0273	0.8666	0.3528
1.5638	1.2004	16500	1.0178	0.8514	0.3440
1.6806	1.2368	17000	1.0114	0.8373	0.3387
1.5214	1.2732	17500	1.0049	0.8307	0.3361
1.409	1.3096	18000	0.9952	0.8325	0.3355
1.4618	1.3459	18500	0.9835	0.8305	0.3318
1.5613	1.3823	19000	0.9587	0.8306	0.3355
1.8607	1.4187	19500	0.9556	0.8257	0.3328
1.8506	1.4551	20000	0.9594	0.8187	0.3279
1.4933	1.4915	20500	0.9329	0.8134	0.3262
1.7531	1.5278	21000	0.9182	0.8087	0.3232
1.2786	1.5642	21500	0.9241	0.8100	0.3218
1.5976	1.6006	22000	0.9095	0.8075	0.3213
1.5346	1.6370	22500	0.9155	0.8010	0.3188
1.4204	1.6733	23000	0.9010	0.8070	0.3185
1.509	1.7097	23500	0.8933	0.8010	0.3168
1.1191	1.7461	24000	0.8943	0.7944	0.3156
1.7361	1.7825	24500	0.8802	0.8008	0.3178
1.3272	1.8188	25000	0.8761	0.7988	0.3156
1.4057	1.8552	25500	0.8705	0.7906	0.3130
1.5288	1.8916	26000	0.8595	0.7869	0.3116
1.6068	1.9280	26500	0.8633	0.7906	0.3118
1.9591	1.9644	27000	0.8562	0.7847	0.3112
1.6623	2.0007	27500	0.8536	0.7818	0.3094
2.2612	2.0371	28000	0.8462	0.7820	0.3074
1.0876	2.0735	28500	0.8511	0.7757	0.3068
1.3043	2.1099	29000	0.8464	0.7847	0.3102
1.1624	2.1462	29500	0.8520	0.7746	0.3064
2.6973	2.1826	30000	0.8365	0.7752	0.3069
1.3618	2.2190	30500	0.8316	0.7729	0.3053
1.384	2.2554	31000	0.8355	0.7670	0.3029
1.4903	2.2917	31500	0.8334	0.7691	0.3035
1.6345	2.3281	32000	0.8350	0.7715	0.3040
1.0043	2.3645	32500	0.8278	0.7659	0.3021
1.3774	2.4009	33000	0.8278	0.7569	0.2989
1.2622	2.4372	33500	0.8215	0.7606	0.3004
1.4779	2.4736	34000	0.8180	0.7602	0.2988
1.2156	2.5100	34500	0.8157	0.7583	0.2985
1.2055	2.5464	35000	0.8189	0.7565	0.2975
0.9888	2.5828	35500	0.8204	0.7502	0.2957
1.2773	2.6191	36000	0.8126	0.7545	0.2968
1.5654	2.6555	36500	0.8133	0.7544	0.2972
1.1398	2.6919	37000	0.8137	0.7520	0.2959
1.1038	2.7283	37500	0.8075	0.7579	0.2969
1.3804	2.7646	38000	0.8101	0.7542	0.2958
1.281	2.8010	38500	0.8084	0.7536	0.2958
1.2316	2.8374	39000	0.8087	0.7551	0.2959
1.5158	2.8738	39500	0.8071	0.7555	0.2963
1.0682	2.9101	40000	0.8074	0.7542	0.2954
1.5609	2.9465	40500	0.8055	0.7539	0.2953
1.193	2.9829	41000	0.8058	0.7539	0.2956

Framework versions

• Transformers 4.41.0
• Pytorch 2.4.1+cu124
• Datasets 2.18.0
• Tokenizers 0.19.1