bayan_model_lora_phase2

This model is a fine-tuned version of mohammed525671/final_arabic_grammarly_2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1149
• Gleu: 0.2999
• Bleu: 14.5429
• Chrf++: 37.5397

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 500
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Bleu	Chrf++	Gleu	Validation Loss
0.8908	0.1170	1000	14.1593	37.2520	0.2953	0.1366
0.7238	0.2340	2000	14.1708	37.2646	0.2957	0.1321
0.7564	0.3510	3000	14.1906	37.2828	0.2960	0.1300
0.7263	0.4679	4000	14.2236	37.3122	0.2964	0.1303
0.7220	0.5849	5000	14.4788	37.4797	0.2989	0.1194
0.7334	0.7019	6000	14.4884	37.4898	0.2990	0.1191
0.6696	0.8189	7000	14.5069	37.5061	0.2993	0.1201
0.7510	0.9359	8000	14.5204	37.5154	0.2995	0.1178
0.6947	1.0529	9000	14.5004	37.4985	0.2994	0.1181
0.6888	1.1699	10000	14.5238	37.5202	0.2996	0.1160
0.7907	1.2869	11000	14.5383	37.5302	0.2998	0.1150
0.6893	1.4038	12000	0.1149	0.2999	14.5484	37.5427
0.6581	1.5208	13000	0.1157	0.2999	14.5438	37.5386
0.6910	1.6378	14000	0.1155	0.2999	14.5448	37.5404
0.6837	1.7548	15000	0.1155	0.2999	14.5440	37.5409
0.7107	1.8718	16000	0.1141	0.3000	14.5410	37.5389
0.7501	1.9888	17000	0.1149	0.2999	14.5410	37.5383
0.7501	2.0	17096	0.1149	0.2999	14.5429	37.5397

Framework versions

• PEFT 0.19.1
• Transformers 5.10.2
• Pytorch 2.11.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2