xls-r-1b-bigcgen-combined-5hrs

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the BIGCGEN - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.6944
• Wer: 0.6601

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: 3e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use 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: 500
• num_epochs: 30.0

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.4228	100	3.7787	1.0
No log	0.8457	200	2.8642	1.0
No log	1.2664	300	1.1326	0.9986
No log	1.6892	400	0.8747	0.8281
5.818	2.1099	500	0.7865	0.8027
5.818	2.5328	600	0.6710	0.6954
5.818	2.9556	700	0.7234	0.7939
5.818	3.3763	800	0.6657	0.6706
5.818	3.7992	900	0.6836	0.7246
1.2021	4.2199	1000	0.6894	0.6897
1.2021	4.6427	1100	0.6464	0.6642
1.2021	5.0634	1200	0.6663	0.6777
1.2021	5.4863	1300	0.6701	0.6704
1.2021	5.9091	1400	0.6834	0.6899
0.8307	6.3298	1500	0.6944	0.6600

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0