kpriyanshu256/whisper-large-v2-br-1000-32-1e-05

openai/whisper-large-v2-breton

This model is a fine-tuned version of openai/whisper-large-v2 on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7162
• Wer: 39.9271

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
• distributed_type: multi-GPU
• 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: 50
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7423	0.1	100	0.8363	57.1553
0.4361	1.07	200	0.6833	46.7176
0.2227	2.03	300	0.6483	42.5929
0.1472	3.0	400	0.6511	42.4627
0.0892	3.1	500	0.6633	40.9604
0.0651	4.07	600	0.6807	39.7534
0.0416	5.04	700	0.6870	41.2383
0.0352	6.0	800	0.7315	39.9010
0.022	6.1	900	0.7201	40.4307
0.0195	7.07	1000	0.7162	39.9271

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2