gemma-2-2b_HelpSteer2_sftsd2_rmsd2

This model is a fine-tuned version of RylanSchaeffer/gemma-2-2b_HelpSteer2_sftsd2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6617
• Accuracy: 0.6114

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: 4
• eval_batch_size: 4
• seed: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 1250

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0	0	2.0297	0.4635
0.926	0.0789	100	0.9822	0.5440
0.6578	0.1578	200	0.6888	0.5983
0.6758	0.2366	300	0.6775	0.5955
0.6664	0.3155	400	0.6640	0.6161
0.6987	0.3944	500	0.6683	0.5927
0.6837	0.4733	600	0.6538	0.6245
0.641	0.5522	700	0.6502	0.6301
0.6706	0.6310	800	0.6548	0.6292
0.678	0.7099	900	0.6588	0.6227
0.654	0.7888	1000	0.6494	0.6358
0.6682	0.8677	1100	0.6538	0.6301
0.69	0.9466	1200	0.6845	0.6030

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1