hubert-large-ll60k-gui-ufe

This model is a fine-tuned version of facebook/hubert-large-ll60k on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.0332
• Cer: 1.0

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
12.7179	0.4329	100	2.9635	1.0
5.9177	0.8658	200	2.9109	1.0
5.9451	1.2987	300	2.9614	1.0
5.9373	1.7316	400	2.9688	1.0
6.0992	2.1645	500	2.9031	1.0
5.8871	2.5974	600	2.8871	1.0
5.9773	3.0303	700	2.9620	1.0
6.0069	3.4632	800	2.9517	1.0
5.9892	3.8961	900	2.9592	1.0
5.9731	4.3290	1000	2.9473	1.0
5.9879	4.7619	1100	2.9479	1.0
6.0309	5.1948	1200	2.9910	1.0
5.9473	5.6277	1300	2.9572	1.0
6.0225	6.0606	1400	2.9512	1.0
6.0506	6.4935	1500	2.9782	1.0
6.0465	6.9264	1600	2.9807	1.0
6.0775	7.3593	1700	2.9781	1.0
6.0678	7.7922	1800	2.9842	1.0
6.0392	8.2251	1900	2.9986	1.0
6.1305	8.6580	2000	3.0216	1.0
6.1608	9.0909	2100	3.0306	1.0
6.1697	9.5238	2200	3.0266	1.0
6.1319	9.9567	2300	3.0332	1.0

Framework versions

• Transformers 5.1.0
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.2