gemma3-peft-multiclass

This model is a fine-tuned version of google/gemma-3-270m-it on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.2109
• Accuracy: 0.42
• F1 Macro: 0.4174
• Precision Macro: 0.4194
• Recall Macro: 0.4193

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
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	Precision Macro	Recall Macro
16.4047	0.0710	10	3.6137	0.352	0.3320	0.3414	0.3439
14.2141	0.1421	20	3.4200	0.356	0.3440	0.3466	0.3498
13.5641	0.2131	30	3.2952	0.354	0.3528	0.3651	0.3571
13.9938	0.2842	40	3.1136	0.35	0.3475	0.3472	0.3480
12.8039	0.3552	50	2.9981	0.352	0.3483	0.3495	0.3518
11.6164	0.4263	60	2.8842	0.36	0.3536	0.3576	0.3604
11.4141	0.4973	70	2.7804	0.382	0.3811	0.3892	0.3827
10.9391	0.5684	80	2.7409	0.382	0.3795	0.3942	0.3874
11.2023	0.6394	90	2.6599	0.368	0.3605	0.3686	0.3621
11.1117	0.7105	100	2.5964	0.382	0.3689	0.3782	0.3741
10.8336	0.7815	110	2.5515	0.382	0.3739	0.3737	0.3780
10.7383	0.8526	120	2.5144	0.378	0.3737	0.3775	0.3778
10.7766	0.9236	130	2.5157	0.376	0.3646	0.3887	0.3816
9.9812	0.9947	140	2.4519	0.388	0.3848	0.3864	0.3848
9.3602	1.0639	150	2.4290	0.39	0.3832	0.3870	0.3889
9.6672	1.1350	160	2.4054	0.398	0.3901	0.3951	0.3922
9.6578	1.2060	170	2.3940	0.404	0.4040	0.4068	0.4061
10.1414	1.2771	180	2.4032	0.388	0.3807	0.4050	0.3947
9.7828	1.3481	190	2.3542	0.4	0.3963	0.3984	0.3969
9.8086	1.4192	200	2.3286	0.406	0.4018	0.4031	0.4027
9.1289	1.4902	210	2.3212	0.39	0.3819	0.3964	0.3933
9.5641	1.5613	220	2.3019	0.414	0.4137	0.4157	0.4171
9.6461	1.6323	230	2.2907	0.412	0.4055	0.4088	0.4080
9.1906	1.7034	240	2.2955	0.422	0.4008	0.4265	0.4151
9.0484	1.7744	250	2.2959	0.406	0.4017	0.4147	0.4101
9.2789	1.8455	260	2.2869	0.392	0.3850	0.4055	0.3972
9.3695	1.9165	270	2.2800	0.394	0.3898	0.4064	0.4001
9.1937	1.9876	280	2.2526	0.414	0.4118	0.4137	0.4130
8.7906	2.0568	290	2.2432	0.422	0.4037	0.4311	0.4120
9.143	2.1279	300	2.2473	0.418	0.3911	0.4353	0.4057
9.0289	2.1989	310	2.2312	0.404	0.3995	0.4058	0.4032
8.8203	2.2700	320	2.2432	0.402	0.3979	0.4116	0.4062
9.082	2.3410	330	2.2344	0.416	0.4145	0.4229	0.4206
9.0086	2.4121	340	2.2278	0.418	0.4169	0.4236	0.4204
8.7383	2.4831	350	2.2259	0.408	0.3991	0.4129	0.4085
8.668	2.5542	360	2.2204	0.398	0.3902	0.3975	0.3971
9.1867	2.6252	370	2.2122	0.412	0.4063	0.4102	0.4078
8.8688	2.6963	380	2.2088	0.432	0.4235	0.4312	0.4259
8.85	2.7673	390	2.2119	0.432	0.4255	0.4327	0.4270
9.3047	2.8384	400	2.2091	0.432	0.4264	0.4295	0.4278
9.1258	2.9094	410	2.2139	0.428	0.4250	0.4273	0.4264
8.825	2.9805	420	2.2109	0.42	0.4174	0.4194	0.4193

Framework versions

• PEFT 0.17.1
• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 4.4.1
• Tokenizers 0.22.1