KMB_SimCSE_test

This model is a fine-tuned version of CocoRoF/KoModernBERT-chp-11 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0438
• Pearson Cosine: 0.7947
• Spearman Cosine: 0.7992
• Pearson Manhattan: 0.7493
• Spearman Manhattan: 0.7655
• Pearson Euclidean: 0.7507
• Spearman Euclidean: 0.7666
• Pearson Dot: 0.6408
• Spearman Dot: 0.6472

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 4.0

Training results

Training Loss	Epoch	Step	Validation Loss	Pearson Cosine	Spearman Cosine	Pearson Manhattan	Spearman Manhattan	Pearson Euclidean	Spearman Euclidean	Pearson Dot	Spearman Dot
0.761	0.1172	250	0.1397	0.7191	0.7366	0.7129	0.7205	0.7135	0.7210	0.4342	0.4302
0.6275	0.2343	500	0.1240	0.7535	0.7638	0.7442	0.7505	0.7442	0.7506	0.4527	0.4533
0.5326	0.3515	750	0.1149	0.7540	0.7698	0.7320	0.7461	0.7327	0.7466	0.4786	0.4737
0.4917	0.4686	1000	0.1028	0.7630	0.7778	0.7395	0.7532	0.7395	0.7531	0.5428	0.5404
0.4451	0.5858	1250	0.0959	0.7634	0.7803	0.7505	0.7649	0.7508	0.7652	0.5909	0.5929
0.4682	0.7029	1500	0.1057	0.7687	0.7855	0.7541	0.7681	0.7545	0.7685	0.5271	0.5190
0.4489	0.8201	1750	0.0994	0.7658	0.7800	0.7505	0.7624	0.7514	0.7627	0.5765	0.5760
0.4696	0.9372	2000	0.1055	0.7618	0.7835	0.7514	0.7669	0.7526	0.7675	0.5910	0.5835
0.3474	1.0544	2250	0.0818	0.7663	0.7777	0.7527	0.7636	0.7536	0.7642	0.5774	0.5748
0.319	1.1715	2500	0.0752	0.7753	0.7858	0.7589	0.7692	0.7592	0.7692	0.5929	0.5919
0.3682	1.2887	2750	0.0767	0.7736	0.7851	0.7556	0.7667	0.7564	0.7671	0.5784	0.5785
0.3033	1.4058	3000	0.0716	0.7836	0.7962	0.7590	0.7723	0.7600	0.7727	0.5987	0.5976
0.3247	1.5230	3250	0.0768	0.7779	0.7911	0.7613	0.7731	0.7621	0.7735	0.5638	0.5623
0.26	1.6401	3500	0.0686	0.7792	0.7902	0.7615	0.7733	0.7623	0.7734	0.6004	0.5998
0.3216	1.7573	3750	0.0707	0.7851	0.7950	0.7668	0.7787	0.7677	0.7791	0.6098	0.6136
0.3166	1.8744	4000	0.0719	0.7799	0.7911	0.7550	0.7693	0.7563	0.7701	0.5737	0.5754
0.315	1.9916	4250	0.0710	0.7818	0.7925	0.7657	0.7780	0.7672	0.7790	0.5918	0.5930
0.2117	2.1087	4500	0.0545	0.7772	0.7890	0.7551	0.7702	0.7567	0.7712	0.6059	0.6096
0.1725	2.2259	4750	0.0544	0.7780	0.7868	0.7593	0.7714	0.7605	0.7721	0.6065	0.6128
0.1985	2.3430	5000	0.0540	0.7818	0.7916	0.7621	0.7733	0.7626	0.7734	0.6017	0.6078
0.1871	2.4602	5250	0.0527	0.7830	0.7898	0.7576	0.7718	0.7587	0.7724	0.5843	0.5894
0.17	2.5773	5500	0.0521	0.7877	0.7959	0.7621	0.7746	0.7633	0.7753	0.6240	0.6246
0.174	2.6945	5750	0.0528	0.7876	0.7949	0.7594	0.7713	0.7603	0.7716	0.6196	0.6234
0.1896	2.8116	6000	0.0506	0.7848	0.7891	0.7595	0.7712	0.7606	0.7718	0.6052	0.6083
0.1897	2.9288	6250	0.0549	0.7819	0.7902	0.7521	0.7664	0.7533	0.7667	0.5957	0.5981
0.105	3.0459	6500	0.0450	0.7887	0.7931	0.7516	0.7669	0.7527	0.7675	0.6385	0.6450
0.1055	3.1631	6750	0.0460	0.7875	0.7927	0.7515	0.7652	0.7525	0.7657	0.6256	0.6332
0.1145	3.2802	7000	0.0453	0.7925	0.7977	0.7548	0.7671	0.7559	0.7678	0.6316	0.6408
0.1252	3.3974	7250	0.0470	0.7889	0.7947	0.7561	0.7683	0.7571	0.7693	0.6257	0.6283
0.1058	3.5145	7500	0.0446	0.7913	0.7958	0.7572	0.7714	0.7578	0.7715	0.6221	0.6338
0.1144	3.6317	7750	0.0433	0.7939	0.7989	0.7534	0.7673	0.7542	0.7677	0.6519	0.6583
0.0971	3.7488	8000	0.0438	0.7952	0.7993	0.7537	0.7675	0.7547	0.7679	0.6345	0.6383
0.1107	3.8660	8250	0.0432	0.7953	0.7992	0.7507	0.7673	0.7518	0.7675	0.6355	0.6411
0.1232	3.9831	8500	0.0438	0.7947	0.7992	0.7493	0.7655	0.7507	0.7666	0.6408	0.6472

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.21.0