ierg4320_en_test

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9548
• Wer: 0.4153

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.0982	0.6211	400	2.9438	0.9938
1.7877	1.2422	800	1.0128	0.6645
0.8688	1.8634	1200	0.8012	0.5662
0.6808	2.4845	1600	0.8370	0.5366
0.6252	3.1056	2000	0.7710	0.5063
0.5532	3.7267	2400	0.7258	0.5041
0.5073	4.3478	2800	0.7337	0.4864
0.4834	4.9689	3200	0.7023	0.4777
0.4419	5.5901	3600	0.7542	0.4708
0.4326	6.2112	4000	0.7187	0.4647
0.4024	6.8323	4400	0.7212	0.4671
0.3809	7.4534	4800	0.7139	0.4582
0.3752	8.0745	5200	0.7296	0.4512
0.337	8.6957	5600	0.7207	0.4578
0.3305	9.3168	6000	0.7233	0.4528
0.3329	9.9379	6400	0.7178	0.4565
0.3047	10.5590	6800	0.7077	0.4518
0.2957	11.1801	7200	0.7788	0.4512
0.2913	11.8012	7600	0.7483	0.4528
0.2685	12.4224	8000	0.7644	0.4426
0.2666	13.0435	8400	0.7640	0.4427
0.2495	13.6646	8800	0.7959	0.4401
0.2501	14.2857	9200	0.7978	0.4494
0.2369	14.9068	9600	0.8217	0.4403
0.2282	15.5280	10000	0.8052	0.4359
0.2293	16.1491	10400	0.8688	0.4357
0.2165	16.7702	10800	0.8566	0.4385
0.2067	17.3913	11200	0.8504	0.4307
0.2034	18.0124	11600	0.8358	0.4346
0.1963	18.6335	12000	0.8729	0.4307
0.1846	19.2547	12400	0.8562	0.4349
0.189	19.8758	12800	0.8408	0.4266
0.1787	20.4969	13200	0.8424	0.4288
0.1757	21.1180	13600	0.8947	0.4327
0.1691	21.7391	14000	0.9070	0.4291
0.1652	22.3602	14400	0.8735	0.4299
0.1619	22.9814	14800	0.9224	0.4315
0.1544	23.6025	15200	0.9199	0.4278
0.1551	24.2236	15600	0.9089	0.4240
0.1449	24.8447	16000	0.9296	0.4229
0.1465	25.4658	16400	0.9476	0.4198
0.1434	26.0870	16800	0.9167	0.4193
0.1431	26.7081	17200	0.9492	0.4141
0.1335	27.3292	17600	0.9597	0.4185
0.1326	27.9503	18000	0.9516	0.4156
0.1333	28.5714	18400	0.9490	0.4153
0.1293	29.1925	18800	0.9605	0.4172
0.1257	29.8137	19200	0.9548	0.4153

Framework versions

• Transformers 4.46.3
• Pytorch 2.5.1.post302
• Datasets 3.1.0
• Tokenizers 0.20.4