mtzig/v3b_mistral_lora_bugged

v3b_mistral_lora

This model is a fine-tuned version of peiyi9979/math-shepherd-mistral-7b-prm on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0126
• Accuracy: 0.9936
• Precision: 0.8793
• Recall: 0.9808
• F1: 0.9273

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 765837
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• total_eval_batch_size: 32
• 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: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
No log	0	0	0.3911	0.9098	0.2222	0.4615	0.3
0.7904	0.0095	20	0.3855	0.9163	0.24	0.4615	0.3158
0.6496	0.0190	40	0.3578	0.9300	0.2603	0.3654	0.304
0.5209	0.0284	60	0.3011	0.9469	0.2941	0.1923	0.2326
0.482	0.0379	80	0.2597	0.9525	0.3478	0.1538	0.2133
0.4165	0.0474	100	0.2602	0.9549	0.4524	0.3654	0.4043
0.5055	0.0569	120	0.2183	0.9614	0.5345	0.5962	0.5636
0.2795	0.0664	140	0.1609	0.9541	0.4699	0.75	0.5778
0.3314	0.0758	160	0.1574	0.9316	0.3659	0.8654	0.5143
0.257	0.0853	180	0.1001	0.9597	0.5119	0.8269	0.6324
0.3089	0.0948	200	0.0747	0.9750	0.6567	0.8462	0.7395
0.329	0.1043	220	0.0722	0.9742	0.6471	0.8462	0.7333
0.4088	0.1138	240	0.0581	0.9815	0.7458	0.8462	0.7928
0.2598	0.1233	260	0.0531	0.9815	0.7458	0.8462	0.7928
0.1978	0.1327	280	0.0428	0.9863	0.8070	0.8846	0.8440
0.1661	0.1422	300	0.0452	0.9839	0.75	0.9231	0.8276
0.2944	0.1517	320	0.0511	0.9815	0.7101	0.9423	0.8099
0.2463	0.1612	340	0.0379	0.9847	0.7463	0.9615	0.8403
0.1767	0.1707	360	0.0379	0.9871	0.7812	0.9615	0.8621
0.1581	0.1801	380	0.0315	0.9903	0.8571	0.9231	0.8889
0.1162	0.1896	400	0.0378	0.9863	0.7692	0.9615	0.8547
0.1442	0.1991	420	0.0395	0.9855	0.7656	0.9423	0.8448
0.1957	0.2086	440	0.0343	0.9903	0.8333	0.9615	0.8929
0.2593	0.2181	460	0.0308	0.9919	0.875	0.9423	0.9074
0.1137	0.2275	480	0.0236	0.9936	0.8929	0.9615	0.9259
0.2213	0.2370	500	0.0264	0.9879	0.7846	0.9808	0.8718
0.2092	0.2465	520	0.0336	0.9895	0.8197	0.9615	0.8850
0.2247	0.2560	540	0.0287	0.9928	0.8909	0.9423	0.9159
0.2684	0.2655	560	0.0286	0.9919	0.8621	0.9615	0.9091
0.1485	0.2749	580	0.0354	0.9895	0.8197	0.9615	0.8850
0.1168	0.2844	600	0.0274	0.9928	0.8909	0.9423	0.9159
0.1899	0.2939	620	0.0262	0.9911	0.8475	0.9615	0.9009
0.1626	0.3034	640	0.0274	0.9895	0.8197	0.9615	0.8850
0.1919	0.3129	660	0.0435	0.9847	0.7324	1.0	0.8455
0.1953	0.3224	680	0.0215	0.9936	0.8929	0.9615	0.9259
0.1978	0.3318	700	0.0260	0.9903	0.8333	0.9615	0.8929
0.311	0.3413	720	0.0168	0.9944	0.9091	0.9615	0.9346
0.226	0.3508	740	0.0166	0.9928	0.8525	1.0	0.9204
0.1394	0.3603	760	0.0195	0.9919	0.85	0.9808	0.9107
0.1702	0.3698	780	0.0358	0.9855	0.75	0.9808	0.85
0.1413	0.3792	800	0.0269	0.9911	0.8361	0.9808	0.9027
0.2015	0.3887	820	0.0238	0.9911	0.8361	0.9808	0.9027
0.1752	0.3982	840	0.0253	0.9895	0.8	1.0	0.8889
0.212	0.4077	860	0.0250	0.9911	0.8475	0.9615	0.9009
0.1556	0.4172	880	0.0247	0.9928	0.8909	0.9423	0.9159
0.1495	0.4266	900	0.0141	0.9919	0.85	0.9808	0.9107
0.1666	0.4361	920	0.0187	0.9936	0.9074	0.9423	0.9245
0.1475	0.4456	940	0.0199	0.9911	0.8361	0.9808	0.9027
0.1459	0.4551	960	0.0166	0.9919	0.8387	1.0	0.9123
0.1264	0.4646	980	0.0155	0.9936	0.8793	0.9808	0.9273
0.1389	0.4740	1000	0.0152	0.9944	0.9091	0.9615	0.9346
0.1433	0.4835	1020	0.0147	0.9944	0.9091	0.9615	0.9346
0.1389	0.4930	1040	0.0126	0.9968	0.9444	0.9808	0.9623
0.1452	0.5025	1060	0.0230	0.9903	0.8125	1.0	0.8966
0.1623	0.5120	1080	0.0128	0.9952	0.9259	0.9615	0.9434
0.1179	0.5215	1100	0.0156	0.9952	0.8966	1.0	0.9455
0.1256	0.5309	1120	0.0175	0.9952	0.9107	0.9808	0.9444
0.1536	0.5404	1140	0.0221	0.9919	0.85	0.9808	0.9107
0.1873	0.5499	1160	0.0234	0.9903	0.8125	1.0	0.8966
0.1234	0.5594	1180	0.0174	0.9944	0.9091	0.9615	0.9346
0.205	0.5689	1200	0.0162	0.9944	0.8814	1.0	0.9369
0.1975	0.5783	1220	0.0155	0.9944	0.9091	0.9615	0.9346
0.1302	0.5878	1240	0.0131	0.9960	0.9123	1.0	0.9541
0.2996	0.5973	1260	0.0172	0.9944	0.8814	1.0	0.9369
0.1381	0.6068	1280	0.0145	0.9936	0.8929	0.9615	0.9259
0.1559	0.6163	1300	0.0142	0.9936	0.8929	0.9615	0.9259
0.1588	0.6257	1320	0.0194	0.9911	0.8361	0.9808	0.9027
0.1101	0.6352	1340	0.0149	0.9936	0.8929	0.9615	0.9259
0.1533	0.6447	1360	0.0152	0.9936	0.8667	1.0	0.9286
0.1567	0.6542	1380	0.0139	0.9928	0.8772	0.9615	0.9174
0.1771	0.6637	1400	0.0176	0.9936	0.8793	0.9808	0.9273
0.1531	0.6731	1420	0.0142	0.9928	0.8772	0.9615	0.9174
0.1366	0.6826	1440	0.0157	0.9928	0.8525	1.0	0.9204
0.1597	0.6921	1460	0.0149	0.9944	0.8814	1.0	0.9369
0.1527	0.7016	1480	0.0156	0.9919	0.8387	1.0	0.9123
0.2199	0.7111	1500	0.0137	0.9944	0.8814	1.0	0.9369
0.2215	0.7205	1520	0.0140	0.9944	0.8814	1.0	0.9369
0.199	0.7300	1540	0.0167	0.9936	0.8667	1.0	0.9286
0.1505	0.7395	1560	0.0106	0.9960	0.9273	0.9808	0.9533
0.2082	0.7490	1580	0.0146	0.9928	0.8525	1.0	0.9204
0.1431	0.7585	1600	0.0108	0.9944	0.8947	0.9808	0.9358
0.1388	0.7680	1620	0.0125	0.9952	0.8966	1.0	0.9455
0.1221	0.7774	1640	0.0100	0.9960	0.9273	0.9808	0.9533
0.1571	0.7869	1660	0.0108	0.9960	0.9273	0.9808	0.9533
0.1472	0.7964	1680	0.0115	0.9944	0.8947	0.9808	0.9358
0.1236	0.8059	1700	0.0121	0.9936	0.8793	0.9808	0.9273
0.1477	0.8154	1720	0.0113	0.9952	0.9107	0.9808	0.9444
0.1781	0.8248	1740	0.0139	0.9936	0.8793	0.9808	0.9273
0.1517	0.8343	1760	0.0129	0.9936	0.8793	0.9808	0.9273
0.1811	0.8438	1780	0.0123	0.9944	0.8947	0.9808	0.9358
0.1592	0.8533	1800	0.0136	0.9944	0.8814	1.0	0.9369
0.0799	0.8628	1820	0.0149	0.9944	0.8814	1.0	0.9369
0.1294	0.8722	1840	0.0130	0.9944	0.8814	1.0	0.9369
0.2076	0.8817	1860	0.0121	0.9936	0.8793	0.9808	0.9273
0.1628	0.8912	1880	0.0121	0.9936	0.8793	0.9808	0.9273
0.156	0.9007	1900	0.0128	0.9936	0.8793	0.9808	0.9273
0.0762	0.9102	1920	0.0128	0.9936	0.8793	0.9808	0.9273
0.1372	0.9196	1940	0.0124	0.9936	0.8793	0.9808	0.9273
0.0837	0.9291	1960	0.0123	0.9936	0.8793	0.9808	0.9273
0.1353	0.9386	1980	0.0126	0.9936	0.8793	0.9808	0.9273
0.1914	0.9481	2000	0.0127	0.9936	0.8793	0.9808	0.9273
0.1072	0.9576	2020	0.0127	0.9936	0.8793	0.9808	0.9273
0.2165	0.9671	2040	0.0129	0.9936	0.8793	0.9808	0.9273
0.1417	0.9765	2060	0.0127	0.9936	0.8793	0.9808	0.9273
0.101	0.9860	2080	0.0126	0.9936	0.8793	0.9808	0.9273
0.2136	0.9955	2100	0.0126	0.9936	0.8793	0.9808	0.9273

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3