IKI-Category-multilabel_bge

This model is a fine-tuned version of BAAI/bge-base-en-v1.5 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4541
• Precision-micro: 0.75
• Precision-samples: 0.7708
• Precision-weighted: 0.7517
• Recall-micro: 0.7880
• Recall-samples: 0.7858
• Recall-weighted: 0.7880
• F1-micro: 0.7685
• F1-samples: 0.7537
• F1-weighted: 0.7615

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: 4.5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Precision-micro	Precision-samples	Precision-weighted	Recall-micro	Recall-samples	Recall-weighted	F1-micro	F1-samples	F1-weighted
0.8999	0.99	94	0.8742	0.3889	0.0272	0.1308	0.0169	0.0188	0.0169	0.0323	0.0202	0.0280
0.7377	2.0	189	0.6770	0.4727	0.4996	0.5333	0.5639	0.5782	0.5639	0.5143	0.4883	0.4998
0.5582	2.99	283	0.5552	0.5111	0.5585	0.5685	0.7229	0.7357	0.7229	0.5988	0.5959	0.6175
0.3943	4.0	378	0.4713	0.5616	0.6397	0.5869	0.7904	0.8071	0.7904	0.6567	0.6761	0.6611
0.2883	4.99	472	0.4555	0.6384	0.6969	0.6444	0.7446	0.7641	0.7446	0.6874	0.6901	0.6854
0.2112	6.0	567	0.4459	0.6443	0.6968	0.6637	0.7855	0.7942	0.7855	0.7079	0.7123	0.7068
0.1608	6.99	661	0.4212	0.6508	0.7071	0.6586	0.7904	0.7931	0.7904	0.7138	0.7161	0.7116
0.1247	8.0	756	0.4177	0.6633	0.7145	0.6650	0.7976	0.8006	0.7976	0.7243	0.7193	0.7195
0.1031	8.99	850	0.4435	0.7277	0.7523	0.7306	0.7855	0.7875	0.7855	0.7555	0.7425	0.7487
0.0851	10.0	945	0.4522	0.7380	0.7623	0.7465	0.7807	0.7795	0.7807	0.7588	0.7432	0.7516
0.074	10.99	1039	0.4548	0.7359	0.7663	0.7368	0.7855	0.7910	0.7855	0.7599	0.7490	0.7521
0.0648	12.0	1134	0.4430	0.7425	0.7676	0.7437	0.7783	0.7781	0.7783	0.76	0.7461	0.7540
0.0605	12.99	1228	0.4478	0.7366	0.7651	0.7379	0.7952	0.7948	0.7952	0.7648	0.7545	0.7579
0.0566	14.0	1323	0.4574	0.7506	0.7708	0.7519	0.7904	0.7893	0.7904	0.7700	0.7546	0.7625
0.0546	14.92	1410	0.4541	0.75	0.7708	0.7517	0.7880	0.7858	0.7880	0.7685	0.7537	0.7615

Category	Precision	Recall	F1	Suport
Active mobility	0.70	0.894	0.7908	19.0
Alternative fuels	0.804	0.865	0.833	52.0
Aviation improvements	0.700	1.00	0.824	7.0
Comprehensive transport planning	0.750	0.571	0.649	21.0
Digital solutions	0.708	0.772	0.739	22.0
Economic instruments	0.742	0.821	0.780	28.0
Education and behavioral change	0.727	0.727	0.727	11.0
Electric mobility	0.766	0.922	0.837	64.0
Freight efficiency improvements	0.768	0.650	0.703	20.0
Improve infrastructure	0.638	0.857	0.732	35.0
Land use	1.00	0.625	0.769	8.0
Other Transport Category	0.600	0.27	0.375	11.0
Public transport improvement	0.777	0.833	0.804	42.0
Shipping improvements	0.846	0.846	0.846	13.0
Transport demand management	0.666	0.40	0.500	15.0
Vehicle improvements	0.783	0.766	0.774	47.0

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Carbon Emitted: 0.0473 kg of CO2
• Hours Used: 0.996 hours

Training Hardware

• On Cloud: yes
• GPU Model: 1 x Tesla T4
• CPU Model: Intel(R) Xeon(R) CPU @ 2.30GHz
• RAM Size: 12.67 GB

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.1