estimation-model-v3

This model is a fine-tuned version of google/gemma-2-2b-jpn-it on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0232
• Accuracy: 0.5739
• Spearmanr: 0.3397
• Kendalltau: 0.2651
• Pearsonr: 0.3174
• Rmse: 1.4493
• Mae: 1.0966

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 16
• seed: 42
• 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: cosine_with_min_lr
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Spearmanr	Kendalltau	Pearsonr	Rmse	Mae
5.4448	0.1567	500	5.2447	0.3697	0.1294	0.0999	0.1030	1.5116	1.1748
3.5747	0.3135	1000	3.6366	0.4647	0.1303	0.1009	0.0898	1.6046	1.2388
3.2813	0.4702	1500	3.0585	0.4625	0.1429	0.1108	0.1173	1.5283	1.1646
3.1132	0.6270	2000	2.5819	0.5048	0.1707	0.1324	0.1416	1.5356	1.1687
2.5897	0.7837	2500	2.3855	0.5182	0.1987	0.1541	0.1669	1.5235	1.1543
2.0304	0.9404	3000	2.2636	0.5293	0.2220	0.1729	0.1969	1.4558	1.0804
1.9453	1.0972	3500	2.2381	0.5338	0.2413	0.1876	0.2098	1.5274	1.1662
2.4994	1.2539	4000	2.1722	0.5419	0.2640	0.2057	0.2346	1.4880	1.1202
2.3851	1.4107	4500	2.1235	0.5419	0.2764	0.2164	0.2521	1.4205	1.0493
2.1885	1.5674	5000	2.0991	0.5464	0.2864	0.2241	0.2689	1.4017	1.0326
1.8545	1.7241	5500	2.0855	0.5486	0.3038	0.2368	0.2769	1.4451	1.0769
1.9475	1.8809	6000	2.0571	0.5627	0.3133	0.2445	0.2914	1.4352	1.0740
1.5089	2.0376	6500	2.0469	0.5635	0.3228	0.2519	0.3009	1.4361	1.0791
2.0828	2.1944	7000	2.0393	0.5687	0.3290	0.2568	0.3054	1.4403	1.0836
1.7599	2.3511	7500	2.0405	0.5679	0.3300	0.2575	0.3058	1.4577	1.1024
2.1807	2.5078	8000	2.0301	0.5679	0.3332	0.2601	0.3104	1.4349	1.0783
1.9166	2.6646	8500	2.0201	0.5642	0.3360	0.2626	0.3177	1.4096	1.0529
2.1982	2.8213	9000	2.0223	0.5709	0.3384	0.2640	0.3171	1.4399	1.0868
2.1137	2.9781	9500	2.0232	0.5739	0.3397	0.2651	0.3174	1.4493	1.0966

Framework versions

• PEFT 0.15.0
• Transformers 4.49.0
• Pytorch 2.4.1+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1