Pageee/DistilFT-Frisian-10hc

DistilFT-Frisian-10h

This model is a fine-tuned version of distil-small.en on the mozilla-foundation/common_voice_6_fy_NL dataset. It achieves the following results on the evaluation set:

• Loss: 0.6854
• Wer: 35.8047

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: 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: 300
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0989	0.5348	500	1.2112	59.2408
0.5734	1.0695	1000	0.8419	46.6405
0.4798	1.6043	1500	0.7341	42.1137
0.2483	2.1390	2000	0.6788	39.4190
0.2367	2.6738	2500	0.6554	37.7865
0.1197	3.2086	3000	0.6613	36.7706
0.0969	3.7433	3500	0.6591	36.7279
0.0468	4.2781	4000	0.6777	35.8688
0.0358	4.8128	4500	0.6771	35.7583
0.028	5.3476	5000	0.6854	35.8047

Framework versions

• Transformers 4.41.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1