asr-baatonou

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4710
• Wer: 47.1082

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9171	0.1265	100	0.7232	56.8144
0.705	0.2530	200	0.6174	51.8920
0.6686	0.3795	300	0.5594	56.5408
0.647	0.5060	400	0.5373	50.7973
0.5975	0.6325	500	0.5198	46.9513
0.5446	0.7590	600	0.5040	47.5862
0.5714	0.8855	700	0.4922	46.8345
0.4901	1.0114	800	0.4785	52.0416
0.4949	1.1379	900	0.4715	46.1595
0.4987	1.2644	1000	0.4640	42.8535
0.489	1.3909	1100	0.4582	44.4919
0.4498	1.5174	1200	0.4508	46.3018
0.5076	1.6439	1300	0.4522	44.3386
0.4795	1.7704	1400	0.4449	48.8706
0.4703	1.8969	1500	0.4419	43.8059
0.4461	2.0228	1600	0.4423	52.4685
0.4418	2.1493	1700	0.4375	46.2580
0.4383	2.2758	1800	0.4327	44.8093
0.4169	2.4023	1900	0.4317	45.1925
0.4143	2.5288	2000	0.4299	43.4081
0.4187	2.6553	2100	0.4295	43.3534
0.4213	2.7818	2200	0.4278	44.1160
0.401	2.9083	2300	0.4217	45.0502
0.44	3.0342	2400	0.4242	44.6999
0.3967	3.1607	2500	0.4256	48.4109
0.3607	3.2872	2600	0.4245	42.7221
0.3863	3.4137	2700	0.4211	46.0427
0.3773	3.5402	2800	0.4187	44.4663
0.402	3.6667	2900	0.4180	43.9445
0.3634	3.7932	3000	0.4205	43.2403
0.401	3.9197	3100	0.4183	44.9589
0.3289	4.0455	3200	0.4233	45.3713
0.3503	4.1720	3300	0.4210	45.4406
0.3712	4.2985	3400	0.4195	44.6086
0.3443	4.4250	3500	0.4222	45.0246
0.3641	4.5515	3600	0.4193	42.2149
0.3789	4.6781	3700	0.4200	43.9628
0.3561	4.8046	3800	0.4187	42.9119
0.3744	4.9311	3900	0.4140	46.7615
0.3287	5.0569	4000	0.4231	43.1272
0.3278	5.1834	4100	0.4221	48.1664
0.3425	5.3099	4200	0.4183	45.0392
0.3353	5.4364	4300	0.4233	46.9878
0.3497	5.5629	4400	0.4170	46.4477
0.3045	5.6894	4500	0.4190	45.1487
0.3205	5.8159	4600	0.4206	46.0573
0.3386	5.9424	4700	0.4182	43.0505
0.2765	6.0683	4800	0.4286	44.7035
0.2935	6.1948	4900	0.4269	45.5574
0.3174	6.3213	5000	0.4305	47.0425
0.3003	6.4478	5100	0.4300	44.6962
0.287	6.5743	5200	0.4292	45.0867
0.2898	6.7008	5300	0.4280	45.3311
0.2914	6.8273	5400	0.4296	46.5061
0.2807	6.9538	5500	0.4289	45.4662
0.2535	7.0797	5600	0.4376	44.7874
0.2348	7.2062	5700	0.4428	46.4441
0.2756	7.3327	5800	0.4407	45.8055
0.2521	7.4592	5900	0.4443	45.5720
0.2815	7.5857	6000	0.4410	46.8564
0.2501	7.7122	6100	0.4440	45.9369
0.256	7.8387	6200	0.4435	45.8128
0.283	7.9652	6300	0.4450	44.5685
0.2254	8.0911	6400	0.4547	45.5756
0.2392	8.2176	6500	0.4541	46.1485
0.2278	8.3441	6600	0.4567	47.8964
0.2319	8.4706	6700	0.4576	46.9987
0.2427	8.5971	6800	0.4581	47.2031
0.2307	8.7236	6900	0.4604	45.8858
0.2229	8.8501	7000	0.4596	46.8418
0.2361	8.9766	7100	0.4596	47.8453
0.1967	9.1025	7200	0.4677	47.6555
0.2068	9.2290	7300	0.4688	47.1775
0.2138	9.3555	7400	0.4697	47.5753
0.2036	9.4820	7500	0.4697	47.0826
0.205	9.6085	7600	0.4714	46.5061
0.2192	9.7350	7700	0.4712	47.3381
0.2054	9.8615	7800	0.4709	47.0316
0.2176	9.9880	7900	0.4710	47.1082

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.1