sft-zephyr-7b-sft-qlora-ultrafeedback-binarized-20241011-162008

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.9805

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: 0.0002
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.0557	0.0105	20	1.2313
1.0725	0.0209	40	1.1635
1.0261	0.0314	60	1.1410
1.1577	0.0419	80	1.1209
1.1619	0.0523	100	1.1067
1.1234	0.0628	120	1.0981
1.0256	0.0733	140	1.0901
1.1511	0.0837	160	1.0850
1.2364	0.0942	180	1.0802
1.1676	0.1047	200	1.0761
1.2327	0.1151	220	1.0731
1.0082	0.1256	240	1.0695
0.9324	0.1361	260	1.0666
1.0435	0.1465	280	1.0630
0.8484	0.1570	300	1.0588
0.962	0.1675	320	1.0588
0.9531	0.1779	340	1.0549
0.8902	0.1884	360	1.0518
1.1103	0.1988	380	1.0485
1.0641	0.2093	400	1.0455
0.9541	0.2198	420	1.0431
1.0081	0.2302	440	1.0427
0.9761	0.2407	460	1.0407
1.0654	0.2512	480	1.0391
1.1185	0.2616	500	1.0367
1.0337	0.2721	520	1.0357
0.9059	0.2826	540	1.0335
1.1223	0.2930	560	1.0318
1.1514	0.3035	580	1.0300
1.0715	0.3140	600	1.0294
1.1336	0.3244	620	1.0263
1.0148	0.3349	640	1.0246
1.0242	0.3454	660	1.0238
1.1316	0.3558	680	1.0220
1.0114	0.3663	700	1.0216
1.1682	0.3768	720	1.0207
1.1026	0.3872	740	1.0180
1.0854	0.3977	760	1.0182
0.8933	0.4082	780	1.0164
1.0233	0.4186	800	1.0153
1.1105	0.4291	820	1.0140
0.8441	0.4396	840	1.0124
0.963	0.4500	860	1.0113
1.0488	0.4605	880	1.0093
0.8147	0.4710	900	1.0084
1.0005	0.4814	920	1.0081
0.959	0.4919	940	1.0071
0.8878	0.5024	960	1.0062
1.238	0.5128	980	1.0048
0.9114	0.5233	1000	1.0032
1.0474	0.5338	1020	1.0017
0.9858	0.5442	1040	1.0009
0.9642	0.5547	1060	1.0007
0.9116	0.5651	1080	0.9992
0.9444	0.5756	1100	0.9978
1.0698	0.5861	1120	0.9970
0.9379	0.5965	1140	0.9959
0.8902	0.6070	1160	0.9950
1.0654	0.6175	1180	0.9941
1.1352	0.6279	1200	0.9935
1.0493	0.6384	1220	0.9922
0.9792	0.6489	1240	0.9913
0.8634	0.6593	1260	0.9903
0.8152	0.6698	1280	0.9898
1.0059	0.6803	1300	0.9890
0.9244	0.6907	1320	0.9884
0.9918	0.7012	1340	0.9876
1.0536	0.7117	1360	0.9872
0.9883	0.7221	1380	0.9866
0.9426	0.7326	1400	0.9863
0.8653	0.7431	1420	0.9855
0.863	0.7535	1440	0.9849
0.9217	0.7640	1460	0.9847
1.0365	0.7745	1480	0.9844
0.8865	0.7849	1500	0.9841
1.1006	0.7954	1520	0.9836
0.9393	0.8059	1540	0.9832
0.8455	0.8163	1560	0.9826
1.1479	0.8268	1580	0.9823
1.0578	0.8373	1600	0.9820
0.7279	0.8477	1620	0.9818
0.973	0.8582	1640	0.9815
1.1227	0.8687	1660	0.9812
0.9897	0.8791	1680	0.9811
0.8196	0.8896	1700	0.9810
0.9309	0.9001	1720	0.9808
0.8774	0.9105	1740	0.9808
0.9671	0.9210	1760	0.9807
1.0849	0.9314	1780	0.9807
1.0233	0.9419	1800	0.9806
0.9742	0.9524	1820	0.9806
1.029	0.9628	1840	0.9806
1.0048	0.9733	1860	0.9806
0.9348	0.9838	1880	0.9805
0.8959	0.9942	1900	0.9805

Framework versions

• PEFT 0.12.0
• Transformers 4.45.2
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.20.0