xbilek25/whisper-small-train-v3.0

basic_train_basic_test 1000 similar params: per_device_train_batch_size=32, # bylo 16 a pod tim 1 gradient_accumulation_steps=2, warmup_steps=300, max_steps=3000

This model is a fine-tuned version of openai/whisper-small on the xbilek25/train_set_1st_1000_de_en_de dataset. It achieves the following results on the evaluation set:

• Loss: 0.6605
• Wer: 25.4637

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: 16
• 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
• training_steps: 3000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0072	9.01	600	0.5727	22.6644
0.0009	19.0	1200	0.6166	23.9460
0.0006	28.01	1800	0.6416	24.7892
0.0004	38.0	2400	0.6551	25.2951
0.0004	47.01	3000	0.6605	25.4637

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.15.2