whisper_med_alex.en

This model is a fine-tuned version of crossdelenna/whisper_med_alex.en on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3767
• Wer: 23.2232

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: 24
• eval_batch_size: 16
• 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
• training_steps: 25609
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2094	1.1583	300	0.1578	10.2392
0.15	2.3166	600	0.1294	7.8546
0.1244	3.4749	900	0.1208	7.2765
0.1077	4.6332	1200	0.1073	6.7227
0.0931	5.7915	1500	0.0945	6.6980
0.0804	6.9498	1800	0.0908	6.4261
0.075	8.1081	2100	0.0797	3.9691
0.0661	9.2664	2400	0.0669	3.1484
0.0597	10.4247	2700	0.0743	4.4795
0.0551	11.5830	3000	0.0669	3.6286
0.0518	12.7413	3300	0.0565	4.1814
0.0475	13.8996	3600	0.0552	2.6162
0.046	15.0579	3900	0.0508	3.1180
0.0412	16.2162	4200	0.0512	2.7882
0.0391	17.3745	4500	0.0525	2.1304
0.038	18.5328	4800	0.0462	1.7037
0.0359	19.6911	5100	0.0409	1.5496
0.0328	20.8494	5400	0.0422	1.7406
0.0306	89.0618	5700	0.0334	1.2605
0.2114	115.3791	6000	0.0279	0.9614
0.1869	121.1517	6300	0.0320	0.8169
0.14	51.1622	6600	0.1414	6.1036
0.1151	53.4865	6900	0.1405	6.0445
0.1194	55.8031	7200	0.1394	6.1036
0.0277	28.9575	7500	0.0355	1.0875
0.0253	30.1158	7800	0.0335	1.0133
0.0244	31.2741	8100	0.0325	0.8898
0.9305	175.0	8400	0.8205	30.4552
0.7302	181.25	8700	0.6930	27.5458
0.624	187.5	9000	0.5692	22.5246
0.516	193.75	9300	0.4793	20.1783
0.4555	200.0	9600	0.4468	18.3013
0.4128	206.25	9900	0.3961	16.0488
0.3649	212.5	10200	0.3497	14.6879
0.323	218.75	10500	0.3275	13.5617
0.3023	225.0	10800	0.3095	12.8109
0.2778	231.25	11100	0.2806	12.1539
0.2519	237.5	11400	0.2660	11.6847
0.2381	243.75	11700	0.2523	11.1685
0.216	250.0	12000	0.2321	10.3707
0.1987	256.25	12300	0.2220	9.7607
0.1827	262.5	12600	0.2024	9.1506
0.1827	266.6667	12800	0.1893	8.3529
0.1608	270.8333	13000	0.1803	7.6959
0.1443	275.0	13200	0.1742	7.5551
0.1443	279.1667	13400	0.1710	7.5082
0.1357	283.3333	13600	0.1703	7.5082
0.1279	287.5	13800	0.1583	6.9451
0.1279	291.6667	14000	0.1499	6.5228
0.111	295.8333	14200	0.1434	6.0535
0.0977	300.0	14400	0.1393	5.9596
0.0977	304.1667	14600	0.1373	5.8189
0.0923	308.3333	14800	0.1370	5.7719
0.0873	312.5	15000	0.1311	5.0211
0.0757	318.75	15300	0.1235	4.8803
0.0678	325.0	15600	0.1192	4.7865
0.0631	331.25	15900	0.1175	4.7865
0.0588	337.5	16200	0.1125	4.6926
0.0517	343.75	16500	0.1075	4.5049
0.0461	350.0	16800	0.1044	4.3641
0.0431	356.25	17100	0.1032	4.3641
0.0402	362.5	17400	0.0997	4.3641
0.0351	368.75	17700	0.0969	4.2703
0.0311	375.0	18000	0.0951	3.0971
0.029	381.25	18300	0.0945	3.0502
0.0275	387.5	18600	0.0925	3.3318
0.024	393.75	18900	0.0908	2.8156
0.0213	400.0	19200	0.0899	2.6279
0.0202	406.25	19500	0.0896	2.6279
0.0189	412.5	19800	0.0888	2.5340
0.0169	418.75	20100	0.0877	2.5340
0.0153	425.0	20400	0.0885	2.5340
0.0146	431.25	20700	0.0886	2.5340
0.0139	437.5	21000	0.0892	2.5340
0.0123	443.75	21300	0.0892	2.5340
0.0112	450.0	21600	0.0897	2.4871
0.0107	456.25	21900	0.0896	2.5809
0.8558	569.2308	22200	0.8489	38.8470
0.6744	576.9231	22500	0.6970	26.5260
0.5946	584.6154	22800	0.6345	24.7551
0.5608	592.3077	23100	0.6144	24.3029
0.5341	600.0	23400	0.5610	23.5870
0.4896	607.6923	23700	0.5067	26.2622
0.4598	615.3846	24000	0.4778	25.5087
0.444	623.0769	24300	0.4675	25.0565
0.4335	647.3684	24600	0.4388	22.9599
0.4055	655.2632	24900	0.4039	24.6271
0.3846	663.1579	25200	0.3839	23.3694
0.3725	671.0526	25500	0.3767	23.2232

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0