wav2vec2-xls-r-300m-en-phoneme-ctc-41h

It achieves the following results on the evaluation set:

• Loss: 0.3051
• Per: 0.0887
• Phoneme 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: 0.0001
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 300
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Phoneme Accuracy
5.9962	1.0	597	3.5377	0.9980	0.0020
3.3735	2.0	1194	0.4537	0.1548	0.8452
0.7613	3.0	1791	0.2519	0.1117	0.8883
0.4123	4.0	2388	0.2086	0.1031	0.8969
0.3445	5.0	2985	0.1980	0.0996	0.9004
0.2721	6.0	3582	0.1892	0.0982	0.9018
0.2487	7.0	4179	0.1879	0.0970	0.9030
0.227	8.0	4776	0.1938	0.0965	0.9035
0.2117	9.0	5373	0.1855	0.0958	0.9042
0.2044	10.0	5970	0.1947	0.0960	0.9040
0.1717	11.0	6567	0.2006	0.0952	0.9048
0.1692	12.0	7164	0.1995	0.0932	0.9068
0.1615	13.0	7761	0.2060	0.0937	0.9063
0.147	14.0	8358	0.2094	0.0944	0.9056
0.1429	15.0	8955	0.2146	0.0942	0.9058
0.131	16.0	9552	0.2171	0.0934	0.9066
0.1264	17.0	10149	0.2151	0.0918	0.9082
0.1254	18.0	10746	0.2267	0.0919	0.9081
0.116	19.0	11343	0.2384	0.0917	0.9083
0.1143	20.0	11940	0.2317	0.0927	0.9073
0.1079	21.0	12537	0.2433	0.0913	0.9087
0.1023	22.0	13134	0.2490	0.0919	0.9081
0.0951	23.0	13731	0.2544	0.0912	0.9088
0.0953	24.0	14328	0.2543	0.0909	0.9091
0.0946	25.0	14925	0.2630	0.0909	0.9091
0.0929	26.0	15522	0.2656	0.0914	0.9086
0.0849	27.0	16119	0.2683	0.0901	0.9099
0.0854	28.0	16716	0.2659	0.0910	0.9090
0.0815	29.0	17313	0.2549	0.0907	0.9093
0.0829	30.0	17910	0.2875	0.0916	0.9084
0.0764	31.0	18507	0.2767	0.0910	0.9090
0.0734	32.0	19104	0.2618	0.0896	0.9104
0.0709	33.0	19701	0.2863	0.0908	0.9092
0.0747	34.0	20298	0.2839	0.0907	0.9093
0.0706	35.0	20895	0.2814	0.0896	0.9104
0.0688	36.0	21492	0.2893	0.0898	0.9102
0.0657	37.0	22089	0.3021	0.0894	0.9106
0.0647	38.0	22686	0.2915	0.0888	0.9112
0.0649	39.0	23283	0.2944	0.0893	0.9107
0.0621	40.0	23880	0.2880	0.0886	0.9114
0.0623	41.0	24477	0.2984	0.0891	0.9109
0.0571	42.0	25074	0.2993	0.0886	0.9114
0.0577	43.0	25671	0.3016	0.0886	0.9114
0.0597	44.0	26268	0.3063	0.0888	0.9112
0.0558	45.0	26865	0.3075	0.0888	0.9112
0.0622	46.0	27462	0.3051	0.0887	0.9113

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1