vphu123/whisper_totaldataa2

whisper-base-lastversion

This model is a fine-tuned version of openai/whisper-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1732

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.000025
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• distributed_type: gpu
• gradient_accumulation_steps: 2
• total_train_batch_size: 256
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5000
• training_steps: 80000

Training results

Training Loss	Epoch	Step	Validation Loss
0.3116	1	5000	0.6231
0.2104	3	10000	0.4287
0.1729	4	15000	0.3421
0.1472	6	20000	0.3211
0.128	7	25000	0.2811
0.1065	9	30000	0.2649
0.0995	10	35000	0.2523
0.0812	12	40000	0.2401
0.066	14	45000	0.2311
0.0574	15	50000	0.2132
0.0463	17	55000	0.2077
0.04	18	60000	0.1957
0.0314	19	65000	0.1813
0.0305	20	70000	0.1802
0.0298	21	75000	0.1755
0.0265	22	80000	0.1732

Framework versions

• Transformers 4.28.1
• Pytorch 2.1.0a0+gitcc01568
• Datasets 2.13.1
• Tokenizers 0.13.3