fine_tuned_ancient_semitic_BERT

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

• Loss: 0.3771
• F1: 0.5652
• F5: 0.5748
• Precision: 0.5417
• Recall: 0.5909

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: 5e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1	F5	Precision	Recall
No log	1.0	17	0.3386	0.0	0.0	0.0	0.0
No log	2.0	34	0.3252	0.0	0.0	0.0	0.0
No log	3.0	51	0.3188	0.0	0.0	0.0	0.0
No log	4.0	68	0.3410	0.0	0.0	0.0	0.0
No log	5.0	85	0.3069	0.1379	0.1080	0.5	0.08
No log	6.0	102	0.3209	0.1379	0.1080	0.5	0.08
No log	7.0	119	0.3432	0.2222	0.2132	0.25	0.2
No log	8.0	136	0.3606	0.2727	0.2592	0.3158	0.24
No log	9.0	153	0.3319	0.2927	0.2700	0.375	0.24
No log	10.0	170	0.3741	0.4074	0.4193	0.3793	0.44
No log	11.0	187	0.3008	0.3784	0.3336	0.5833	0.28
No log	12.0	204	0.3237	0.4231	0.4294	0.4074	0.44
No log	13.0	221	0.2848	0.5	0.4752	0.5789	0.44
No log	14.0	238	0.3058	0.52	0.52	0.52	0.52
No log	15.0	255	0.2912	0.5417	0.5332	0.5652	0.52
No log	16.0	272	0.3005	0.4681	0.4569	0.5	0.44
No log	17.0	289	0.3122	0.5556	0.5717	0.5172	0.6
No log	18.0	306	0.3670	0.5667	0.6052	0.4857	0.68
No log	19.0	323	0.2818	0.5926	0.6098	0.5517	0.64
No log	20.0	340	0.3012	0.5882	0.5927	0.5769	0.6
No log	21.0	357	0.3288	0.6154	0.6246	0.5926	0.64
No log	22.0	374	0.3251	0.6250	0.6152	0.6522	0.6
No log	23.0	391	0.3145	0.6250	0.6152	0.6522	0.6
No log	24.0	408	0.3128	0.68	0.68	0.68	0.68
No log	25.0	425	0.3190	0.6538	0.6636	0.6296	0.68

Framework versions

• Transformers 4.38.2
• Pytorch 2.3.0a0+ebedce2
• Datasets 2.17.1
• Tokenizers 0.15.2