mms-1b-all-lg-CV-Fleurs-1hrs-v1

This model is a fine-tuned version of facebook/mms-1b-all on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4100
• Wer: 0.4579
• Cer: 0.0897

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: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
13.1561	1.0	33	6.8333	1.0125	0.8338
5.7233	2.0	66	4.8759	0.9995	0.9681
4.3803	3.0	99	3.3296	1.0	0.8922
1.104	4.0	132	0.3399	0.4080	0.0826
0.3525	5.0	165	0.2893	0.3694	0.0764
0.3231	6.0	198	0.2836	0.3696	0.0763
0.2939	7.0	231	0.2792	0.3760	0.0751
0.2799	8.0	264	0.2819	0.3784	0.0756
0.2632	9.0	297	0.2732	0.3737	0.0741
0.2481	10.0	330	0.2814	0.3648	0.0735
0.2371	11.0	363	0.2811	0.3572	0.0728
0.2377	12.0	396	0.2955	0.3701	0.0746
0.234	13.0	429	0.2803	0.3628	0.0723
0.2241	14.0	462	0.2891	0.3635	0.0744
0.2252	15.0	495	0.2841	0.3615	0.0728
0.2279	16.0	528	0.2903	0.3742	0.0757
0.2386	17.0	561	0.2980	0.3681	0.0750
0.2563	18.0	594	0.3717	0.4037	0.0799
0.2718	19.0	627	0.3242	0.3841	0.0784
0.3255	20.0	660	0.3398	0.3876	0.0792
0.3285	21.0	693	0.3553	0.4299	0.0822
0.3719	22.0	726	0.4071	0.5228	0.1024
0.4119	23.0	759	0.4100	0.4579	0.0897

Framework versions

• Transformers 4.48.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0