Ahmed107/hamsa-meetingsv2.5

Hamsa-meetings

This model is a fine-tuned version of nadsoft/hamsa-v0.1-beta on the nadsoft/nadsoft-meetings-v2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9346
• Wer Ortho: 43.4495
• Wer: 43.4495

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: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.4883	2.91	250	0.6170	40.2644	40.2644
0.1678	5.81	500	0.6893	43.6899	43.6899
0.0749	8.72	750	0.7367	42.0673	42.0673
0.0352	11.63	1000	0.7829	42.6683	42.6683
0.0214	14.53	1250	0.8553	43.9904	43.9904
0.0146	17.44	1500	0.9061	43.3894	43.3894
0.0112	20.35	1750	0.9225	44.2909	44.2909
0.0104	23.26	2000	0.9346	43.4495	43.4495

Framework versions

• Transformers 4.36.0
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0