Fauna-v0.7 - Rootflo

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

• Loss: 0.0558
• Wer: 60.7190

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-06
• train_batch_size: 72
• eval_batch_size: 96
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 576
• total_eval_batch_size: 384
• optimizer: Use adamw_bnb_8bit with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 1500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0301	1.0695	500	0.0498	58.1393
0.0278	2.1390	1000	0.0515	58.0461
0.0261	3.2086	1500	0.0538	59.1088
0.0241	4.2781	2000	0.0558	60.7190

Framework versions

• Transformers 4.46.1
• Pytorch 2.5.1+cu124
• Datasets 3.0.2
• Tokenizers 0.20.3