LightOnOCR-2-ft-takukuru

This model is a fine-tuned version of lightonai/LightOnOCR-2-1B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4290

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: 6e-05
• train_batch_size: 2
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 10
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.6963	0.1134	50	0.6549
0.5308	0.2268	100	0.5731
0.4599	0.3401	150	0.5378
0.4328	0.4535	200	0.5106
0.4138	0.5669	250	0.4893
0.4592	0.6803	300	0.4746
0.4014	0.7937	350	0.4629
0.3840	0.9070	400	0.4484
0.3883	1.0204	450	0.4434
0.3366	1.1338	500	0.4407
0.2948	1.2472	550	0.4367
0.2986	1.3605	600	0.4366
0.3041	1.4739	650	0.4348
0.3367	1.5873	700	0.4314
0.3053	1.7007	750	0.4296
0.3113	1.8141	800	0.4291
0.2956	1.9274	850	0.4290

Framework versions

• Transformers 5.0.0.dev0
• Pytorch 2.8.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.1