kids_phoneme_sm_model

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the https://huggingface.co/datasets/mirfan899/kids_phoneme_sm dataset. It achieves the following results on the evaluation set:

• Loss: 0.5405
• Cer: 0.2770

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: 4e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
3.2595	0.74	500	3.7094	1.0
2.8393	1.48	1000	3.2563	1.0
2.7916	2.22	1500	3.0450	1.0
1.9585	2.96	2000	1.0280	0.8428
1.0099	3.7	2500	0.6477	0.5162
0.7968	4.44	3000	0.5551	0.4592
0.6977	5.19	3500	0.5107	0.4065
0.609	5.93	4000	0.4763	0.3916
0.5941	6.67	4500	0.4817	0.3850
0.5411	7.41	5000	0.4755	0.3639
0.5021	8.15	5500	0.4649	0.3622
0.4884	8.89	6000	0.4630	0.3569
0.4484	9.63	6500	0.4675	0.3420
0.4432	10.37	7000	0.4192	0.3402
0.399	11.11	7500	0.4508	0.3310
0.4215	11.85	8000	0.4406	0.3345
0.366	12.59	8500	0.4620	0.3248
0.3708	13.33	9000	0.4594	0.3327
0.3352	14.07	9500	0.4649	0.3121
0.3468	14.81	10000	0.4413	0.3020
0.3283	15.56	10500	0.4948	0.2915
0.3222	16.3	11000	0.4870	0.3025
0.3081	17.04	11500	0.4779	0.2919
0.3099	17.78	12000	0.4927	0.2871
0.2485	18.52	12500	0.5013	0.2831
0.3163	19.26	13000	0.4929	0.2888
0.2555	20.0	13500	0.5234	0.2888
0.2705	20.74	14000	0.5259	0.2818
0.2632	21.48	14500	0.5105	0.2831
0.2374	22.22	15000	0.5284	0.2845
0.2565	22.96	15500	0.5237	0.2875
0.2394	23.7	16000	0.5368	0.2818
0.2458	24.44	16500	0.5386	0.2814
0.2383	25.19	17000	0.5366	0.2788
0.2152	25.93	17500	0.5320	0.2770
0.231	26.67	18000	0.5441	0.2779
0.2061	27.41	18500	0.5448	0.2796
0.245	28.15	19000	0.5413	0.2796
0.2119	28.89	19500	0.5379	0.2774
0.2155	29.63	20000	0.5405	0.2770

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.0
• Datasets 2.13.0
• Tokenizers 0.13.3