c2de63094b39d7a708ab29fbef13418d

This model is a fine-tuned version of google/gemma-2b on the contemmcm/hate-speech-and-offensive-language dataset. It achieves the following results on the evaluation set:

• Loss: 1.9597
• Data Size: 1.0
• Epoch Runtime: 120.1581
• Accuracy: 0.8805
• F1 Macro: 0.7220

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: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	10.3970	0	8.0531	0.1605	0.1126
No log	1	619	3.4281	0.0078	8.0725	0.7025	0.4142
No log	2	1238	1.8057	0.0156	10.7769	0.8571	0.5585
0.0638	3	1857	1.3290	0.0312	17.6640	0.9000	0.6010
0.0638	4	2476	1.3352	0.0625	20.4095	0.9050	0.6162
1.4392	5	3095	1.1541	0.125	30.0605	0.9062	0.7165
0.1139	6	3714	1.2512	0.25	42.7309	0.9075	0.6404
1.4476	7	4333	1.2949	0.5	70.7366	0.8776	0.7006
1.1024	8.0	4952	1.1059	1.0	124.6609	0.9048	0.7329
0.8604	9.0	5571	1.2030	1.0	117.7966	0.8965	0.7118
0.7213	10.0	6190	1.6289	1.0	118.2617	0.8925	0.6738
0.5029	11.0	6809	1.9951	1.0	121.5040	0.9020	0.7012
0.2815	12.0	7428	1.9597	1.0	120.1581	0.8805	0.7220

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1