xlm-roberta-xl-lora

This model is a fine-tuned version of facebook/xlm-roberta-xl on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.5846
• Precision: 0.8927
• Recall: 0.9038
• F1: 0.8982
• Accuracy: 0.9154

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 64
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 63
• num_epochs: 50
• label_smoothing_factor: 0.2

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	2.0	126	3.4068	0.2417	0.2988	0.2672	0.2522
No log	4.0	252	2.5708	0.5402	0.6641	0.5958	0.6379
No log	6.0	378	2.2050	0.6278	0.7262	0.6734	0.7242
2.8519	8.0	504	2.0050	0.7250	0.7922	0.7571	0.7955
2.8519	10.0	630	1.8831	0.8083	0.8427	0.8252	0.8531
2.8519	12.0	756	1.7923	0.8453	0.8630	0.8540	0.8756
2.8519	14.0	882	1.7371	0.8496	0.8693	0.8593	0.8843
1.8053	16.0	1008	1.7031	0.8529	0.8753	0.8640	0.8886
1.8053	18.0	1134	1.6692	0.8691	0.8812	0.8751	0.8969
1.8053	20.0	1260	1.6555	0.8699	0.8856	0.8777	0.8991
1.8053	22.0	1386	1.6359	0.8824	0.8903	0.8863	0.9054
1.6089	24.0	1512	1.6303	0.8756	0.8919	0.8837	0.9043
1.6089	26.0	1638	1.6169	0.8806	0.8935	0.8870	0.9063
1.6089	28.0	1764	1.6105	0.8876	0.8952	0.8914	0.9088
1.6089	30.0	1890	1.6067	0.8861	0.8981	0.8920	0.9089
1.5373	32.0	2016	1.5998	0.8870	0.8989	0.8929	0.9109
1.5373	34.0	2142	1.5967	0.8900	0.8996	0.8948	0.9121
1.5373	36.0	2268	1.5939	0.8912	0.9015	0.8964	0.9137
1.5373	38.0	2394	1.5922	0.8914	0.9014	0.8964	0.9135
1.501	40.0	2520	1.5894	0.8920	0.9021	0.8970	0.9142
1.501	42.0	2646	1.5874	0.8900	0.9029	0.8964	0.9139
1.501	44.0	2772	1.5865	0.8930	0.9043	0.8986	0.9155
1.501	46.0	2898	1.5866	0.8906	0.9036	0.8971	0.9146
1.4812	48.0	3024	1.5853	0.8907	0.9033	0.8970	0.9148
1.4812	50.0	3150	1.5846	0.8927	0.9038	0.8982	0.9154

Framework versions

• Transformers 4.31.0
• Pytorch 2.1.0
• Datasets 2.14.5
• Tokenizers 0.13.3