Modelpublic

ctaguchi/ssc-rwm-mms-model-mix-adapt-max-lowlr

sourceHugging Faceupdated 10mo agoView on Hugging Face

0likes4downloads

Model Card

ssc-rwm-mms-model-mix-adapt-max-lowlr

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.8955
—Cer: 0.2104
—Wer: 0.5980

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
—trainbatchsize: 1
—evalbatchsize: 6
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 2
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_steps: 100
—num_epochs: 5
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.8419	0.3019	200	0.8862	0.2211	0.6454
0.826	0.6038	400	0.9426	0.2187	0.6285
0.8465	0.9057	600	0.9068	0.2198	0.6294
0.7648	1.2068	800	0.9459	0.2165	0.6163
0.825	1.5087	1000	0.9425	0.2149	0.6112
0.8225	1.8106	1200	0.9395	0.2144	0.6109
0.7771	2.1117	1400	0.8818	0.2168	0.6138
0.7675	2.4136	1600	0.8921	0.2129	0.6095
0.7874	2.7155	1800	0.8872	0.2142	0.6060
0.7384	3.0166	2000	0.8983	0.2119	0.6058
0.7053	3.3185	2200	0.9122	0.2123	0.5995
0.6995	3.6204	2400	0.8902	0.2125	0.6012
0.7518	3.9223	2600	0.8890	0.2106	0.6009
0.7021	4.2234	2800	0.8882	0.2105	0.6001
0.7198	4.5253	3000	0.8828	0.2122	0.6032
0.6693	4.8272	3200	0.8955	0.2104	0.5980

Framework versions

—Transformers 4.57.2
—Pytorch 2.9.1+cu128
—Datasets 3.6.0
—Tokenizers 0.22.0