metadata

license: apache-2.0
base_model: ctl/wav2vec2-large-xlsr-cantonese
tags:
  - generated_from_trainer
model-index:
  - name: wav2vec2-large-xls-r-300m-zhhk
    results: []

wav2vec2-large-xls-r-300m-zhhk

This model is a fine-tuned version of ctl/wav2vec2-large-xlsr-cantonese on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 4.6213
Cer: 0.7864

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: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
num_epochs: 30
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.3317	1.35	400	4.5373	0.7915
0.3303	2.71	800	4.5198	0.7915
0.3288	4.06	1200	4.8663	0.8504
0.2901	5.41	1600	4.6080	0.8198
0.2819	6.77	2000	4.4941	0.7316
0.2629	8.12	2400	4.6927	0.8021
0.2363	9.48	2800	4.8796	0.8701
0.2205	10.83	3200	4.6338	0.8087
0.2171	12.18	3600	4.5740	0.7562
0.1875	13.54	4000	4.6072	0.7992
0.1824	14.89	4400	4.6546	0.7669
0.178	16.24	4800	4.6410	0.7961
0.1644	17.6	5200	4.7306	0.8236
0.155	18.95	5600	4.6632	0.7900
0.1396	20.3	6000	4.6239	0.8015
0.1411	21.66	6400	4.6007	0.7793
0.13	23.01	6800	4.5354	0.7475
0.1232	24.37	7200	4.6229	0.7600
0.1239	25.72	7600	4.6382	0.7727
0.1322	27.07	8000	4.6734	0.7902
0.1338	28.43	8400	4.6536	0.7861
0.1353	29.78	8800	4.6213	0.7864

Framework versions

Transformers 4.35.2
Pytorch 2.1.0+cu121
Datasets 2.15.0
Tokenizers 0.15.0