torgo_xlsr_finetune_M01_keep_all

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.4135
• Wer: 0.2436

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.5955	0.56	1000	3.3189	1.0
2.8346	1.11	2000	2.3479	0.9970
1.1249	1.67	3000	1.3541	0.6641
0.7802	2.23	4000	1.5194	0.5856
0.6145	2.79	5000	1.4507	0.5063
0.535	3.34	6000	1.4502	0.4964
0.5027	3.9	7000	1.4042	0.4167
0.4289	4.46	8000	1.4100	0.4091
0.4027	5.02	9000	1.3281	0.3538
0.3501	5.57	10000	1.5689	0.3862
0.3202	6.13	11000	1.2194	0.3416
0.3159	6.69	12000	1.3865	0.3366
0.2933	7.25	13000	1.3632	0.3222
0.2748	7.8	14000	1.6112	0.3408
0.2666	8.36	15000	1.5043	0.3267
0.2578	8.92	16000	1.3961	0.2974
0.2589	9.48	17000	1.1875	0.2939
0.2389	10.03	18000	1.3837	0.3100
0.2531	10.59	19000	1.3161	0.2978
0.2132	11.15	20000	1.3118	0.2848
0.1979	11.71	21000	1.3639	0.3012
0.1912	12.26	22000	1.4812	0.2798
0.2104	12.82	23000	1.4553	0.2642
0.1711	13.38	24000	1.3557	0.2592
0.2064	13.94	25000	1.3746	0.2737
0.1764	14.49	26000	1.3845	0.2665
0.1706	15.05	27000	1.3203	0.2570
0.1601	15.61	28000	1.4216	0.2592
0.1589	16.16	29000	1.3192	0.2440
0.15	16.72	30000	1.3713	0.2585
0.1464	17.28	31000	1.4574	0.2577
0.159	17.84	32000	1.3796	0.2516
0.1336	18.39	33000	1.4397	0.2474
0.1423	18.95	34000	1.3613	0.2459
0.1081	19.51	35000	1.4135	0.2436

Framework versions

• Transformers 4.26.1
• Pytorch 2.1.2
• Datasets 2.16.1
• Tokenizers 0.13.3