qwen2.5-3b-sft3-25-2

This model is a fine-tuned version of Qwen/Qwen2.5-3B on the hZzy/SFT_new_mix_full2 dataset. It achieves the following results on the evaluation set:

• Loss: 2.0083

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-06
• train_batch_size: 10
• eval_batch_size: 10
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 320
• total_eval_batch_size: 40
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.5077	0.0549	5	2.4993
2.4825	0.1099	10	2.4981
2.4996	0.1648	15	2.4941
2.4918	0.2198	20	2.4830
2.4775	0.2747	25	2.4775
2.4498	0.3297	30	2.4512
2.4412	0.3846	35	2.4408
2.4251	0.4396	40	2.4016
2.3855	0.4945	45	2.3850
2.3874	0.5495	50	2.3679
2.3528	0.6044	55	2.3480
2.3324	0.6593	60	2.3311
2.3166	0.7143	65	2.3162
2.3002	0.7692	70	2.3039
2.3037	0.8242	75	2.2930
2.2807	0.8791	80	2.2831
2.2813	0.9341	85	2.2735
2.2683	0.9890	90	2.2645
2.2559	1.0440	95	2.2556
2.2543	1.0989	100	2.2489
2.2405	1.1538	105	2.2407
2.2318	1.2088	110	2.2327
2.2314	1.2637	115	2.2249
2.2107	1.3187	120	2.2172
2.2057	1.3736	125	2.2096
2.1878	1.4286	130	2.2020
2.1925	1.4835	135	2.1945
2.1719	1.5385	140	2.1868
2.1704	1.5934	145	2.1795
2.172	1.6484	150	2.1741
2.1671	1.7033	155	2.1679
2.1591	1.7582	160	2.1620
2.1446	1.8132	165	2.1566
2.1457	1.8681	170	2.1514
2.144	1.9231	175	2.1465
2.1469	1.9780	180	2.1418
2.1249	2.0330	185	2.1374
2.1111	2.0879	190	2.1333
2.118	2.1429	195	2.1293
2.1157	2.1978	200	2.1257
2.1066	2.2527	205	2.1221
2.1073	2.3077	210	2.1188
2.1037	2.3626	215	2.1155
2.0922	2.4176	220	2.1124
2.1008	2.4725	225	2.1094
2.0915	2.5275	230	2.1065
2.1005	2.5824	235	2.1037
2.0791	2.6374	240	2.1010
2.0892	2.6923	245	2.0985
2.0739	2.7473	250	2.0960
2.0861	2.8022	255	2.0935
2.0753	2.8571	260	2.0912
2.077	2.9121	265	2.0889
2.0588	2.9670	270	2.0867
2.0611	3.0220	275	2.0844
2.0585	3.0769	280	2.0824
2.0546	3.1319	285	2.0805
2.059	3.1868	290	2.0785
2.0482	3.2418	295	2.0767
2.0586	3.2967	300	2.0747
2.0457	3.3516	305	2.0729
2.0431	3.4066	310	2.0711
2.0593	3.4615	315	2.0694
2.0319	3.5165	320	2.0678
2.029	3.5714	325	2.0660
2.0438	3.6264	330	2.0644
2.035	3.6813	335	2.0628
2.0276	3.7363	340	2.0612
2.0345	3.7912	345	2.0598
2.027	3.8462	350	2.0583
2.0319	3.9011	355	2.0570
2.0354	3.9560	360	2.0555
2.0278	4.0110	365	2.0541
2.0202	4.0659	370	2.0530
2.0126	4.1209	375	2.0517
2.0246	4.1758	380	2.0504
2.0089	4.2308	385	2.0491
2.0237	4.2857	390	2.0478
2.0076	4.3407	395	2.0467
2.0119	4.3956	400	2.0455
1.9964	4.4505	405	2.0445
2.001	4.5055	410	2.0433
2.0073	4.5604	415	2.0421
2.0021	4.6154	420	2.0411
1.9958	4.6703	425	2.0399
2.0025	4.7253	430	2.0389
2.0023	4.7802	435	2.0379
2.0028	4.8352	440	2.0369
1.9874	4.8901	445	2.0360
2.002	4.9451	450	2.0350
1.9943	5.0	455	2.0341
1.9939	5.0549	460	2.0333
1.9997	5.1099	465	2.0326
1.9845	5.1648	470	2.0318
1.987	5.2198	475	2.0308
1.98	5.2747	480	2.0300
1.9828	5.3297	485	2.0292
1.9776	5.3846	490	2.0285
1.9832	5.4396	495	2.0277
1.9774	5.4945	500	2.0271
1.9789	5.5495	505	2.0264
1.9744	5.6044	510	2.0256
1.984	5.6593	515	2.0249
1.9693	5.7143	520	2.0244
1.9732	5.7692	525	2.0236
1.9638	5.8242	530	2.0229
1.9807	5.8791	535	2.0224
1.967	5.9341	540	2.0216
1.9776	5.9890	545	2.0211
1.9738	6.0440	550	2.0207
1.9533	6.0989	555	2.0201
1.9609	6.1538	560	2.0197
1.9735	6.2088	565	2.0193
1.9706	6.2637	570	2.0186
1.9704	6.3187	575	2.0182
1.9563	6.3736	580	2.0178
1.9689	6.4286	585	2.0174
1.954	6.4835	590	2.0170
1.961	6.5385	595	2.0165
1.9603	6.5934	600	2.0160
1.9473	6.6484	605	2.0158
1.9628	6.7033	610	2.0154
1.9614	6.7582	615	2.0149
1.9567	6.8132	620	2.0146
1.9649	6.8681	625	2.0144
1.9547	6.9231	630	2.0139
1.9634	6.9780	635	2.0136
1.9388	7.0330	640	2.0134
1.959	7.0879	645	2.0132
1.9566	7.1429	650	2.0128
1.9527	7.1978	655	2.0126
1.9554	7.2527	660	2.0124
1.9566	7.3077	665	2.0121
1.9558	7.3626	670	2.0119
1.9516	7.4176	675	2.0117
1.9552	7.4725	680	2.0115
1.9451	7.5275	685	2.0113
1.943	7.5824	690	2.0111
1.9488	7.6374	695	2.0110
1.9539	7.6923	700	2.0107
1.9386	7.7473	705	2.0105
1.9428	7.8022	710	2.0103
1.9558	7.8571	715	2.0102
1.9483	7.9121	720	2.0100
1.9457	7.9670	725	2.0099
1.941	8.0220	730	2.0098
1.9457	8.0769	735	2.0097
1.9401	8.1319	740	2.0095
1.9464	8.1868	745	2.0094
1.9387	8.2418	750	2.0093
1.9421	8.2967	755	2.0093
1.9435	8.3516	760	2.0092
1.9491	8.4066	765	2.0091
1.9457	8.4615	770	2.0090
1.9488	8.5165	775	2.0089
1.9409	8.5714	780	2.0089
1.9376	8.6264	785	2.0089
1.9383	8.6813	790	2.0088
1.9465	8.7363	795	2.0087
1.9518	8.7912	800	2.0086
1.942	8.8462	805	2.0086
1.9472	8.9011	810	2.0086
1.9515	8.9560	815	2.0085
1.9384	9.0110	820	2.0084
1.9448	9.0659	825	2.0084
1.9349	9.1209	830	2.0084
1.9506	9.1758	835	2.0084
1.9368	9.2308	840	2.0084
1.9298	9.2857	845	2.0084
1.9443	9.3407	850	2.0084
1.9514	9.3956	855	2.0084
1.95	9.4505	860	2.0083
1.948	9.5055	865	2.0083
1.9466	9.5604	870	2.0083
1.9387	9.6154	875	2.0083
1.9397	9.6703	880	2.0083
1.9431	9.7253	885	2.0083
1.9452	9.7802	890	2.0083
1.9387	9.8352	895	2.0083
1.9336	9.8901	900	2.0083
1.9408	9.9451	905	2.0083
1.945	10.0	910	2.0083

Framework versions

• Transformers 4.42.0
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.19.1