xtreme_s_w2v2_t5lephone-small_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: 1.5203
• F1: 0.7526
• Accuracy: 0.7518

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
2.589	3.95	20	2.6401	0.0108	0.0816
2.5223	7.95	40	2.6493	0.0339	0.0816
2.5085	11.95	60	2.6236	0.0539	0.1028
2.1252	15.95	80	2.5006	0.1458	0.1667
1.3711	19.95	100	2.2712	0.2344	0.2837
1.5092	23.95	120	2.0599	0.3631	0.3936
0.4962	27.95	140	1.8475	0.4881	0.4894
0.4169	31.95	160	1.8262	0.5358	0.5142
0.1579	35.95	180	1.6481	0.5967	0.6028
0.0927	39.95	200	1.4470	0.6748	0.6560
0.1363	43.95	220	1.2725	0.6836	0.6879
0.1324	47.95	240	1.4330	0.6653	0.6702
0.0294	51.95	260	1.2978	0.7079	0.7163
0.0326	55.95	280	1.3869	0.6823	0.6879
0.0444	59.95	300	1.5764	0.7051	0.6986
0.0527	63.95	320	2.2013	0.5899	0.5851
0.1542	67.95	340	1.5203	0.7053	0.6986
0.0127	71.95	360	1.7149	0.7105	0.7128
0.0105	75.95	380	1.2471	0.7853	0.7837
0.009	79.95	400	1.5720	0.7065	0.7057
0.0081	83.95	420	1.9395	0.6656	0.6702
0.2345	87.95	440	1.5704	0.7408	0.7411
0.0076	91.95	460	1.4706	0.7554	0.7589
0.0064	95.95	480	1.5746	0.7491	0.7518
0.3105	99.95	500	1.6824	0.7273	0.7376
0.0058	103.95	520	1.3799	0.7474	0.7624
0.0055	107.95	540	1.4086	0.7350	0.7518
0.0051	111.95	560	1.2832	0.7874	0.7979
0.0052	115.95	580	1.3474	0.7752	0.7801
0.0046	119.95	600	1.6125	0.7451	0.7482
0.0044	123.95	620	1.5927	0.7486	0.7518
0.0044	127.95	640	1.5551	0.7487	0.7518
0.0041	131.95	660	1.5117	0.7631	0.7660
0.0041	135.95	680	1.5210	0.7577	0.7624
0.0041	139.95	700	1.5145	0.7655	0.7660
0.004	143.95	720	1.5053	0.7665	0.7660
0.004	147.95	740	1.5203	0.7526	0.7518

Framework versions

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