WAVLM_TITML_IDN_model

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the audiofolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.7585
• Accuracy: 0.8181

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: 0.0003
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
8.0217	0.98	31	7.7416	0.0472
5.1076	2.0	63	3.5170	0.0472
3.0131	2.98	94	2.9921	0.0876
3.0119	4.0	126	2.9580	0.0928
2.685	4.98	157	2.6591	0.0793
2.4513	6.0	189	2.3831	0.1257
2.4415	6.98	220	2.3518	0.1415
2.2998	8.0	252	2.2327	0.1864
2.1987	8.98	283	2.1297	0.1549
2.1206	10.0	315	2.0529	0.2118
2.0542	10.98	346	1.9592	0.2507
1.9693	12.0	378	1.8652	0.2792
1.8677	12.98	409	1.7811	0.3668
1.7369	14.0	441	1.7902	0.2493
1.6551	14.98	472	1.6558	0.3406
1.6176	16.0	504	1.5724	0.3585
1.5666	16.98	535	1.5822	0.4207
1.5103	18.0	567	1.5028	0.4379
1.4695	18.98	598	1.4276	0.4970
1.3016	20.0	630	1.3621	0.4798
1.2025	20.98	661	1.2016	0.5778
1.1211	22.0	693	1.2346	0.5644
1.0204	22.98	724	1.0743	0.6445
0.9365	24.0	756	1.0121	0.6759
0.8553	24.98	787	0.9246	0.7290
0.7698	26.0	819	0.8603	0.7612
0.7336	26.98	850	0.8072	0.7867
0.6965	28.0	882	0.7770	0.8009
0.6662	28.98	913	0.7640	0.8136
0.63	29.52	930	0.7585	0.8181

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu118
• Datasets 2.14.6
• Tokenizers 0.14.1