mms-1b-nyagen-male-model

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

• Loss: 0.1298
• Wer: 0.1952

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: 4
• eval_batch_size: 4
• seed: 42
• 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_steps: 100
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.345	0.2463	100	0.4506	0.4026
0.353	0.4926	200	0.2066	0.2924
0.2989	0.7389	300	0.1829	0.2715
0.2402	0.9852	400	0.1674	0.2546
0.2302	1.2315	500	0.1582	0.2436
0.2279	1.4778	600	0.1602	0.2428
0.2264	1.7241	700	0.1507	0.2287
0.2276	1.9704	800	0.1496	0.2353
0.2088	2.2167	900	0.1460	0.2208
0.1881	2.4631	1000	0.1455	0.2165
0.2079	2.7094	1100	0.1418	0.2168
0.196	2.9557	1200	0.1404	0.2086
0.1782	3.2020	1300	0.1373	0.2078
0.1741	3.4483	1400	0.1343	0.1944
0.1948	3.6946	1500	0.1318	0.2137
0.1904	3.9409	1600	0.1307	0.2043
0.1762	4.1872	1700	0.1313	0.2003
0.1718	4.4335	1800	0.1275	0.1932
0.1595	4.6798	1900	0.1276	0.1952
0.1811	4.9261	2000	0.1290	0.1983
0.1477	5.1724	2100	0.1298	0.1952

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0