facebook-mms-1b-all-common_voice_fleurs-amh-200hrs-v1

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

• Loss: 0.3264
• Wer: 0.1207
• Cer: 0.0366

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use 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_ratio: 0.05
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.309	1.0	2293	0.5970	0.3579	0.1065
0.5773	2.0	4586	0.3532	0.2227	0.0633
0.4096	3.0	6879	0.3304	0.1958	0.0560
0.3602	4.0	9172	0.3361	0.2031	0.0575
0.3407	5.0	11465	0.3173	0.2024	0.0570
0.3035	6.0	13758	0.2646	0.2164	0.0579
0.2462	7.0	16051	0.2196	0.1644	0.0468
0.2073	8.0	18344	0.2239	0.1559	0.0446
0.1712	9.0	20637	0.2348	0.1525	0.0441
0.1504	10.0	22930	0.2257	0.1531	0.0436
0.139	11.0	25223	0.2197	0.1416	0.0410
0.1218	12.0	27516	0.2497	0.1415	0.0413
0.1066	13.0	29809	0.2531	0.1392	0.0410
0.0953	14.0	32102	0.2425	0.1422	0.0413
0.0862	15.0	34395	0.2478	0.1370	0.0403
0.0793	16.0	36688	0.2572	0.1361	0.0398
0.0748	17.0	38981	0.2469	0.1415	0.0410
0.0674	18.0	41274	0.2633	0.1328	0.0390
0.0608	19.0	43567	0.2889	0.1305	0.0384
0.0573	20.0	45860	0.2731	0.1305	0.0385
0.0556	21.0	48153	0.2794	0.1310	0.0390
0.0552	22.0	50446	0.2897	0.1283	0.0381
0.0522	23.0	52739	0.2844	0.1255	0.0375
0.0505	24.0	55032	0.2764	0.1268	0.0379
0.0481	25.0	57325	0.2927	0.1345	0.0395
0.0493	26.0	59618	0.2942	0.1279	0.0379
0.0494	27.0	61911	0.2938	0.1313	0.0388
0.0444	28.0	64204	0.3004	0.1258	0.0377
0.0421	29.0	66497	0.2940	0.1252	0.0379
0.0403	30.0	68790	0.3049	0.1236	0.0369
0.0395	31.0	71083	0.2992	0.1230	0.0372
0.0392	32.0	73376	0.3000	0.1269	0.0381
0.0392	33.0	75669	0.2924	0.1259	0.0376
0.0379	34.0	77962	0.3070	0.1228	0.0369
0.0365	35.0	80255	0.3140	0.1202	0.0364
0.0352	36.0	82548	0.3092	0.1221	0.0364
0.0343	37.0	84841	0.3229	0.1233	0.0370
0.0346	38.0	87134	0.3046	0.1218	0.0367
0.033	39.0	89427	0.3369	0.1196	0.0362
0.0329	40.0	91720	0.3181	0.1214	0.0364
0.0318	41.0	94013	0.3143	0.1205	0.0362
0.0319	42.0	96306	0.3043	0.1267	0.0374
0.0343	43.0	98599	0.3056	0.1206	0.0364
0.0314	44.0	100892	0.3116	0.1214	0.0365
0.0309	45.0	103185	0.3264	0.1207	0.0366

Framework versions

• Transformers 4.48.2
• Pytorch 2.5.1+cu121
• Datasets 2.20.0
• Tokenizers 0.21.0