Moroccan-Darija-STT-small-v1.6.14

This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4439
• Wer: 85.3414
• Cer: 49.3370

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: 1.25e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Use OptimizerNames.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: 10
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.5007	0.0888	60	0.3851	88.7467	46.0771
0.4537	0.1775	120	0.3661	87.9685	46.5061
0.4063	0.2663	180	0.3480	88.2614	47.5381
0.3917	0.3550	240	0.3455	91.3655	51.6443
0.3729	0.4438	300	0.3408	79.2922	38.5555
0.3568	0.5325	360	0.3354	80.3882	37.6062
0.3487	0.6213	420	0.3397	78.0706	36.3833
0.3394	0.7101	480	0.3361	86.8641	44.8879
0.3184	0.7988	540	0.3305	75.7446	34.0337
0.3282	0.8876	600	0.3382	83.4923	42.3458
0.2858	0.9763	660	0.3339	81.1663	40.9016
0.2897	1.0651	720	0.3523	78.2296	40.7277
0.2931	1.1538	780	0.3436	88.7132	48.3337
0.2846	1.2426	840	0.3533	82.7309	41.5266
0.2561	1.3314	900	0.3497	94.8544	60.7300
0.274	1.4201	960	0.3525	78.3384	38.3122
0.2588	1.5089	1020	0.3725	82.7477	46.0754
0.2599	1.5976	1080	0.3617	87.5920	46.6919
0.2405	1.6864	1140	0.3701	78.4890	40.0470
0.237	1.7751	1200	0.3698	77.4096	38.2835
0.237	1.8639	1260	0.3775	82.5218	45.6835
0.24	1.9527	1320	0.3761	91.6750	51.9197
0.2059	2.0414	1380	0.3858	86.8139	47.1632
0.2045	2.1302	1440	0.3900	80.7062	42.4151
0.1791	2.2189	1500	0.4005	83.7098	48.4232
0.2109	2.3077	1560	0.4003	94.7122	56.3806
0.1796	2.3964	1620	0.4056	82.9903	47.7273
0.192	2.4852	1680	0.3970	80.9153	41.4844
0.1932	2.5740	1740	0.4128	83.9106	46.9436
0.1893	2.6627	1800	0.4105	89.9933	50.2694
0.1924	2.7515	1860	0.4134	88.1944	51.3842
0.1734	2.8402	1920	0.4134	90.2443	57.4734
0.1656	2.9290	1980	0.4139	93.75	52.3403
0.1573	3.0178	2040	0.4240	82.2122	44.0940
0.1534	3.1065	2100	0.4249	81.4592	44.0147
0.1429	3.1953	2160	0.4323	82.9987	47.3844
0.144	3.2840	2220	0.4339	89.8092	52.5058
0.1393	3.3728	2280	0.4334	82.2875	45.4656
0.1532	3.4615	2340	0.4392	83.8939	48.2560
0.1499	3.5503	2400	0.4397	87.4665	49.2897
0.1472	3.6391	2460	0.4429	88.4036	52.5548
0.153	3.7278	2520	0.4427	86.8056	50.2407
0.1473	3.8166	2580	0.4441	85.5254	48.8573
0.1407	3.9053	2640	0.4446	85.3163	49.9992
0.1439	3.9941	2700	0.4439	85.3414	49.3370

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0