qwen25-0.5b-prefix-tuning

This model is a fine-tuned version of Qwen/Qwen2.5-0.5B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0820
• Bleu: 0.0062
• Rougel: 0.0738

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: 8
• eval_batch_size: 2
• 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
• num_epochs: 12

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Rougel
No log	0	0	3.1247	0.0014	0.0523
1.8148	0.16	20	3.0069	0.0019	0.0538
1.6696	0.32	40	2.9177	0.0021	0.0594
1.7065	0.48	60	2.8417	0.0021	0.0595
1.6933	0.64	80	2.7770	0.0021	0.0608
1.7361	0.8	100	2.7165	0.0021	0.0704
1.6731	0.96	120	2.6617	0.0013	0.0708
1.7361	1.12	140	2.6140	0.0013	0.0713
1.684	1.28	160	2.5721	0.0019	0.0717
1.5104	1.44	180	2.5317	0.0017	0.0682
1.674	1.6	200	2.4960	0.0017	0.0696
1.5632	1.76	220	2.4623	0.0027	0.0699
1.5437	1.92	240	2.4329	0.0028	0.0754
1.4255	2.08	260	2.4085	0.0028	0.0748
1.6271	2.24	280	2.3852	0.0041	0.0752
1.4799	2.4	300	2.3632	0.0041	0.0744
1.5331	2.56	320	2.3437	0.0040	0.0735
1.514	2.7200	340	2.3266	0.0041	0.0740
1.4895	2.88	360	2.3120	0.0046	0.0754
1.4881	3.04	380	2.2989	0.0051	0.0753
1.4331	3.2	400	2.2879	0.0051	0.0753
1.4564	3.36	420	2.2765	0.0051	0.0753
1.3368	3.52	440	2.2671	0.0056	0.0756
1.4613	3.68	460	2.2583	0.0060	0.0759
1.4115	3.84	480	2.2504	0.0065	0.0762
1.505	4.0	500	2.2427	0.0073	0.0769
1.4828	4.16	520	2.2347	0.0073	0.0765
1.4893	4.32	540	2.2285	0.0072	0.0761
1.5359	4.48	560	2.2214	0.0072	0.0759
1.4248	4.64	580	2.2151	0.0072	0.0759
1.5041	4.8	600	2.2087	0.0072	0.0756
1.5033	4.96	620	2.2031	0.0072	0.0754
1.4361	5.12	640	2.1981	0.0071	0.0747
1.4457	5.28	660	2.1928	0.0072	0.0747
1.4021	5.44	680	2.1872	0.0071	0.0751
1.3328	5.6	700	2.1821	0.0071	0.0751
1.5165	5.76	720	2.1769	0.0071	0.0751
1.339	5.92	740	2.1722	0.0071	0.0749
1.3154	6.08	760	2.1672	0.0071	0.0749
1.3402	6.24	780	2.1630	0.0071	0.0751
1.5663	6.4	800	2.1584	0.0071	0.0751
1.3663	6.5600	820	2.1538	0.0071	0.0749
1.4353	6.72	840	2.1489	0.0071	0.0749
1.3013	6.88	860	2.1447	0.0063	0.0746
1.2856	7.04	880	2.1405	0.0063	0.0746
1.4434	7.2	900	2.1364	0.0063	0.0746
1.4239	7.36	920	2.1324	0.0063	0.0742
1.3253	7.52	940	2.1287	0.0063	0.0742
1.2715	7.68	960	2.1254	0.0063	0.0742
1.404	7.84	980	2.1222	0.0063	0.0742
1.3504	8.0	1000	2.1188	0.0063	0.0744
1.3449	8.16	1020	2.1154	0.0063	0.0744
1.3085	8.32	1040	2.1121	0.0063	0.0742
1.3783	8.48	1060	2.1093	0.0063	0.0744
1.4059	8.64	1080	2.1065	0.0063	0.0744
1.3948	8.8	1100	2.1040	0.0063	0.0742
1.4517	8.96	1120	2.1015	0.0063	0.0744
1.4014	9.12	1140	2.0992	0.0063	0.0744
1.4002	9.28	1160	2.0971	0.0063	0.0746
1.2871	9.44	1180	2.0950	0.0063	0.0746
1.4064	9.6	1200	2.0930	0.0063	0.0746
1.4157	9.76	1220	2.0914	0.0062	0.0742
1.364	9.92	1240	2.0900	0.0062	0.0742
1.4199	10.08	1260	2.0887	0.0062	0.0742
1.4118	10.24	1280	2.0876	0.0062	0.0738
1.3544	10.4	1300	2.0867	0.0062	0.0738
1.3636	10.56	1320	2.0859	0.0062	0.0738
1.4293	10.72	1340	2.0850	0.0062	0.0738
1.3867	10.88	1360	2.0842	0.0062	0.0738
1.2971	11.04	1380	2.0836	0.0062	0.0738
1.3792	11.2	1400	2.0832	0.0062	0.0738
1.3477	11.36	1420	2.0827	0.0062	0.0738
1.3216	11.52	1440	2.0824	0.0062	0.0738
1.3645	11.68	1460	2.0822	0.0062	0.0737
1.3225	11.84	1480	2.0821	0.0062	0.0738
1.2991	12.0	1500	2.0820	0.0062	0.0738

Framework versions

• PEFT 0.14.0
• Transformers 4.51.1
• Pytorch 2.5.1+cu124
• Datasets 3.5.0
• Tokenizers 0.21.0