xtreme_s_w2v2_minds14.en-US

This model is a fine-tuned version of facebook/wav2vec2-large-lv60 on the GOOGLE/XTREME_S - MINDS14.EN-US dataset. It achieves the following results on the evaluation set:

• Loss: 0.5337
• F1: 0.9144
• Accuracy: 0.9113

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: 2
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 150.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Accuracy
2.6482	3.95	20	2.6421	0.0242	0.0745
2.6292	7.95	40	2.6359	0.0108	0.0816
2.5993	11.95	60	2.6301	0.0167	0.0674
2.4023	15.95	80	2.5514	0.1105	0.1454
1.4015	19.95	100	1.6843	0.5599	0.5851
0.4379	23.95	120	0.8126	0.7921	0.7908
0.0642	27.95	140	0.7178	0.8158	0.8156
0.0376	31.95	160	0.7286	0.8473	0.8475
0.0185	35.95	180	0.6779	0.8719	0.8723
0.0752	39.95	200	0.7096	0.8578	0.8511
0.0266	43.95	220	0.7655	0.8596	0.8546
0.0078	47.95	240	0.7623	0.8563	0.8511
0.007	51.95	260	0.6620	0.8794	0.8759
0.0047	55.95	280	0.5936	0.9045	0.9007
0.0067	59.95	300	0.8279	0.8546	0.8617
0.0394	63.95	320	0.8766	0.8359	0.8227
0.0051	67.95	340	0.8097	0.8483	0.8475
0.0095	71.95	360	0.6095	0.9083	0.9078
0.0026	75.95	380	0.5286	0.8889	0.8865
0.0023	79.95	400	0.7218	0.8926	0.8936
0.0023	83.95	420	0.6551	0.8997	0.8972
0.0027	87.95	440	0.6664	0.8848	0.8794
0.0019	91.95	460	0.5344	0.9032	0.9043
0.002	95.95	480	0.5863	0.8983	0.9007
0.0015	99.95	500	0.5715	0.9047	0.9043
0.0016	103.95	520	0.5615	0.8956	0.8936
0.0014	107.95	540	0.6353	0.8965	0.8936
0.0014	111.95	560	0.5593	0.9041	0.9007
0.0013	115.95	580	0.6041	0.8977	0.8936
0.0013	119.95	600	0.5794	0.9026	0.9007
0.0012	123.95	620	0.6858	0.9003	0.8972
0.0013	127.95	640	0.6730	0.9002	0.8972
0.0013	131.95	660	0.5707	0.9146	0.9113
0.0012	135.95	680	0.5604	0.9153	0.9113
0.0019	139.95	700	0.5468	0.9114	0.9078
0.0015	143.95	720	0.5361	0.9144	0.9113
0.0012	147.95	740	0.5337	0.9144	0.9113

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0+cu113
• Datasets 2.1.0
• Tokenizers 0.12.1