wav2vec2-large-mms-1b-kyrgyz-colab

This model is a fine-tuned version of facebook/mms-1b-all on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2384
• Wer: 0.2312

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: 0.001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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: 100
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.9324	0.2347	100	0.3080	0.2811
0.4397	0.4695	200	0.2887	0.2666
0.4336	0.7042	300	0.2816	0.2581
0.4064	0.9390	400	0.2602	0.2518
0.3604	1.1737	500	0.2447	0.2433
0.3752	1.4085	600	0.2603	0.2457
0.3801	1.6432	700	0.2531	0.2433
0.4021	1.8779	800	0.2418	0.2441
0.3807	2.1127	900	0.2625	0.2431
0.3764	2.3474	1000	0.2421	0.2410
0.3669	2.5822	1100	0.2423	0.2422
0.3529	2.8169	1200	0.2447	0.2337
0.3432	3.0516	1300	0.2351	0.2345
0.3535	3.2864	1400	0.2469	0.2331
0.3481	3.5211	1500	0.2353	0.2355
0.3301	3.7559	1600	0.2414	0.2309
0.3696	3.9906	1700	0.2384	0.2312

Framework versions

• Transformers 4.49.0.dev0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0