LLMGUARD-roberta-11

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

• Loss: 0.7933
• Accuracy: 0.7696

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-06
• train_batch_size: 16
• eval_batch_size: 8
• 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: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 32
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.2522	1.0	1564	1.9637	0.3163
1.0671	2.0	3128	0.8108	0.7662
0.7146	3.0	4692	0.6397	0.7953
0.6354	4.0	6256	0.5867	0.8027
0.5794	5.0	7820	0.5669	0.8107
0.5303	6.0	9384	0.5505	0.8105
0.504	7.0	10948	0.5560	0.8081
0.4664	8.0	12512	0.5553	0.8071
0.4559	9.0	14076	0.5613	0.8073
0.4393	10.0	15640	0.5734	0.8044
0.4292	11.0	17204	0.5735	0.7995
0.4065	12.0	18768	0.5987	0.8007
0.3905	13.0	20332	0.6292	0.7951
0.3828	14.0	21896	0.6233	0.7927
0.3658	15.0	23460	0.6490	0.7887
0.3663	16.0	25024	0.6571	0.7913
0.3551	17.0	26588	0.6533	0.7852
0.3439	18.0	28152	0.6814	0.7833
0.3316	19.0	29716	0.6764	0.7817
0.3297	20.0	31280	0.6896	0.7798
0.318	21.0	32844	0.6947	0.7775
0.323	22.0	34408	0.7231	0.7767
0.304	23.0	35972	0.7333	0.7707
0.3145	24.0	37536	0.7491	0.7742
0.2975	25.0	39100	0.7502	0.7708
0.2812	26.0	40664	0.7632	0.7719
0.2825	27.0	42228	0.7734	0.7664
0.2828	28.0	43792	0.7767	0.7697
0.2872	29.0	45356	0.7865	0.7683
0.2759	30.0	46920	0.7849	0.7700
0.2758	31.0	48484	0.7915	0.7692
0.2797	32.0	50048	0.7933	0.7696

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0