CafeBERT_nli

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

• Loss: 1.2989
• Accuracy: 0.8306
• Precision Macro: 0.8307
• Recall Macro: 0.8308
• F1 Macro: 0.8306
• F1 Weighted: 0.8306

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 256
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision Macro	Recall Macro	F1 Macro	F1 Weighted
1.0641	1.0	72	0.6313	0.7565	0.7672	0.7575	0.7562	0.7561
0.64	2.0	144	0.5313	0.8044	0.8077	0.8042	0.8039	0.8040
0.3679	3.0	216	0.5117	0.8062	0.8078	0.8067	0.8060	0.8059
0.2855	4.0	288	0.5816	0.8098	0.8150	0.8101	0.8087	0.8087
0.1571	5.0	360	0.6372	0.8058	0.8060	0.8058	0.8058	0.8059
0.1165	6.0	432	0.6929	0.8177	0.8186	0.8177	0.8178	0.8178
0.0855	7.0	504	0.7374	0.8084	0.8090	0.8087	0.8084	0.8084
0.0704	8.0	576	0.8241	0.8075	0.8107	0.8071	0.8075	0.8075
0.0593	9.0	648	0.9712	0.8098	0.8108	0.8094	0.8095	0.8096
0.0415	10.0	720	0.8643	0.8155	0.8165	0.8153	0.8155	0.8155
0.034	11.0	792	0.9662	0.8124	0.8149	0.8120	0.8123	0.8123
0.0273	12.0	864	1.0114	0.8182	0.8188	0.8181	0.8182	0.8182
0.0189	13.0	936	1.2237	0.8155	0.8195	0.8159	0.8156	0.8155
0.0068	14.0	1008	1.2312	0.8244	0.8265	0.8247	0.8244	0.8244
0.011	15.0	1080	1.2062	0.8315	0.8316	0.8316	0.8314	0.8314
0.003	16.0	1152	1.2550	0.8279	0.8280	0.8280	0.8280	0.8279
0.0024	17.0	1224	1.2774	0.8302	0.8303	0.8303	0.8302	0.8302
0.003	18.0	1296	1.2946	0.8293	0.8295	0.8295	0.8292	0.8292
0.0023	19.0	1368	1.2969	0.8306	0.8307	0.8308	0.8306	0.8306
0.0012	20.0	1440	1.2989	0.8306	0.8307	0.8308	0.8306	0.8306

Framework versions

• Transformers 4.55.0
• Pytorch 2.7.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4