base_sami_22k_ftmix_ep20

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1144.6145
• 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: 0.0005
• train_batch_size: 16
• eval_batch_size: 8
• 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_ratio: 0.25
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
5425.9882	1.0	1525	1144.5670	1.0	1.0
4416.5207	2.0	3050	1142.7595	1.0	1.0
4474.501	3.0	4575	1144.2609	1.0	1.0
4499.2403	4.0	6100	1163.1985	1.0	1.0
4492.3184	5.0	7625	1145.3702	1.0	1.0
4439.5584	6.0	9150	1145.4093	1.0	1.0
4400.9305	7.0	10675	1144.0641	1.0	1.0
4388.3134	8.0	12200	1142.6368	1.0	1.0
4337.4364	9.0	13725	1141.1510	1.0	1.0
4328.8672	10.0	15250	1140.2180	1.0	1.0
4315.4475	11.0	16775	1140.9918	1.0	1.0
4340.9885	12.0	18300	1141.5726	1.0	1.0
4303.5944	13.0	19825	1140.9758	1.0	1.0
4305.4659	14.0	21350	1141.2748	1.0	1.0
4296.8341	15.0	22875	1140.5507	1.0	1.0
4293.78	16.0	24400	1141.3170	1.0	1.0
4291.8813	17.0	25925	1140.3838	1.0	1.0
4291.3587	18.0	27450	1141.1962	1.0	1.0
4288.1016	19.0	28975	1140.9142	1.0	1.0
4287.0193	20.0	30500	1141.5001	1.0	1.0

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1
• Datasets 3.2.0
• Tokenizers 0.21.0