wav2vec2 large danish npsc nst This model is a fine tuned version of chcaa/xls r 300m danish on the None dataset. It achieves the following results on the evaluation set: Loss: 0.0587 Wer: 0.0669 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: 0.0001 train batch size: 16 eval batch size: 16 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: 2000 num epochs: 15.0 mixed precision training: Native AMP Training results Training Loss Epoch Step Validation Loss Wer : : : : : : : : : : 3.0474 0.05 500 2.9879 1.0 1.7255 0.11 1000 1.1271 0.9732 0.8932 0.16 1500 0.5474 0.5983 0.7358 0.21 2000 0.4152 0.4584 0.5939 0.26 2500 0.3455 0.3860 0.5437 0.32 3000 0.3024 0.3473 0.5128 0.37 3500 0.2708 0.3043 0.4682 0.42 4000 0.2462 0.2841 0.4392 0.47 4500 0.2374 0.2639 0.4022 0.53 5000 0.2182 0.2457 0.4106 0.58 5500 0.2066 0.2331 0.3883 0.63 6000 0.1997 0.2267 0.3701…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy