🔴Check out my latest URDU ASR model with 25 WER here -> kingabzpro/whisper-large-v3-turbo-urdu

---

wav2vec2-60-Urdu-V8

This model is a fine-tuned version of Harveenchadha/vakyansh-wav2vec2-urdu-urm-60 on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 11.4832
• Wer: 0.5729
• Cer: 0.3170

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 7.5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• 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: 200
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
19.671	8.33	100	7.7671	0.8795	0.4492
2.085	16.67	200	9.2759	0.6201	0.3320
0.6633	25.0	300	8.7025	0.5738	0.3104
0.388	33.33	400	10.2286	0.5852	0.3128
0.2822	41.67	500	11.1953	0.5738	0.3174
0.2293	50.0	600	11.4832	0.5729	0.3170

Framework versions

• Transformers 4.16.2
• Pytorch 1.10.0+cu111
• Datasets 1.18.3
• Tokenizers 0.11.0