gemma9_on_korean_summary_events

This model is a fine-tuned version of rtzr/ko-gemma-2-9b-it on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4183

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 5
• total_train_batch_size: 10
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.5952	0.1316	20	1.0820
0.9103	0.2632	40	0.7513
0.7022	0.3947	60	0.5833
0.5149	0.5263	80	0.4630
0.4837	0.6579	100	0.4376
0.449	0.7895	120	0.4213
0.431	0.9211	140	0.4080
0.3811	1.0526	160	0.4000
0.3227	1.1842	180	0.3964
0.283	1.3158	200	0.3974
0.2984	1.4474	220	0.3993
0.3102	1.5789	240	0.3851
0.3045	1.7105	260	0.3847
0.3034	1.8421	280	0.3851
0.2779	1.9737	300	0.3793
0.2191	2.1053	320	0.3991
0.1971	2.2368	340	0.4157
0.1908	2.3684	360	0.4209
0.1766	2.5	380	0.4190
0.1749	2.6316	400	0.4183

Framework versions

• PEFT 0.12.0
• Transformers 4.43.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1