kiranpantha/w2v-bert-2.0-nepali

Wave2Vec2-Bert2.0 - Kiran Pantha

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the OpenSLR54 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2182
• Wer: 0.2604

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.4889	0.1800	300	0.8423	0.8076
0.7028	0.3599	600	0.6309	0.5951
0.5787	0.5399	900	0.5455	0.5167
0.476	0.7199	1200	0.4670	0.5109
0.4094	0.8998	1500	0.4415	0.4382
0.3345	1.0798	1800	0.3395	0.3951
0.2545	1.2597	2100	0.3266	0.3609
0.2444	1.4397	2400	0.2814	0.3204
0.2214	1.6197	2700	0.2593	0.2947
0.1846	1.7996	3000	0.2256	0.2685
0.1783	1.9796	3300	0.2182	0.2604

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1