cdli-whisper-en-ug-openai-large-v3-full-a40-fixed-nogc-clean

This model is a fine-tuned version of openai/whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.8364
• Wer: 0.2963
• Cer: 0.2133

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 150
• training_steps: 1500

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.5356	1.5445	250	0.7888	0.3111	0.2203
0.4352	3.0866	500	0.7592	0.2968	0.2120
0.3735	4.6311	750	0.7697	0.3011	0.2153
0.2898	6.1732	1000	0.7953	0.2971	0.2138
0.2843	7.7177	1250	0.8035	0.2980	0.2136
0.2508	9.2599	1500	0.8364	0.2963	0.2133

Framework versions

• Transformers 4.57.6
• Pytorch 2.7.1+cu118
• Datasets 3.6.0
• Tokenizers 0.22.2