wav2vec2-large-xls-r-300m-upper-sorbian-pl-frozen-colab

This model is a fine-tuned version of on the common_voice_16_1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7512
• Wer: 0.3985
• Cer: 0.0926

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 60
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.6621	3.23	100	0.6431	0.7048	0.1711
0.4922	6.45	200	0.5838	0.6120	0.1468
0.3591	9.68	300	0.5487	0.5621	0.1322
0.2869	12.9	400	0.5812	0.5436	0.1309
0.2179	16.13	500	0.6222	0.5014	0.1212
0.1731	19.35	600	0.6930	0.4808	0.1141
0.1315	22.58	700	0.6681	0.4721	0.1116
0.1044	25.81	800	0.6849	0.4567	0.1088
0.0876	29.03	900	0.7287	0.4623	0.1125
0.0822	32.26	1000	0.7278	0.4496	0.1097
0.0736	35.48	1100	0.7534	0.4552	0.1117
0.0641	38.71	1200	0.7500	0.4220	0.1025
0.0572	41.94	1300	0.7008	0.4227	0.1024
0.0495	45.16	1400	0.7697	0.4267	0.1011
0.0488	48.39	1500	0.7364	0.4051	0.0947
0.0444	51.61	1600	0.7444	0.4110	0.0952
0.0416	54.84	1700	0.7621	0.3983	0.0936
0.0398	58.06	1800	0.7512	0.3985	0.0926

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.15.2