v4_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: 0.2841
• Accuracy: 0.8687
• Precision: 0.8392
• Recall: 0.6575
• F1: 0.7373

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: 6
• eval_batch_size: 8
• seed: 89234
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 3
• total_train_batch_size: 72
• 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	0.5995	0.7340	0.6	0.1535	0.2445
0.6057	0.0254	20	0.5918	0.7373	0.625	0.1575	0.2516
0.5356	0.0507	40	0.5521	0.7517	0.6790	0.2165	0.3284
0.5141	0.0761	60	0.5021	0.7627	0.5890	0.5079	0.5455
0.3594	0.1015	80	0.4427	0.7980	0.7448	0.4252	0.5414
0.3988	0.1268	100	0.4067	0.8245	0.7684	0.5354	0.6311
0.3205	0.1522	120	0.3738	0.8201	0.8138	0.4646	0.5915
0.3026	0.1776	140	0.3680	0.8289	0.8235	0.4961	0.6192
0.2886	0.2030	160	0.3467	0.8433	0.8590	0.5276	0.6537
0.2345	0.2283	180	0.3289	0.8587	0.7972	0.6654	0.7253
0.2964	0.2537	200	0.3322	0.8377	0.8497	0.5118	0.6388
0.2655	0.2791	220	0.3495	0.8278	0.8657	0.4567	0.5979
0.3252	0.3044	240	0.3189	0.8455	0.8314	0.5630	0.6714
0.2561	0.3298	260	0.3228	0.8532	0.8201	0.6102	0.6998
0.1661	0.3552	280	0.3141	0.8499	0.8598	0.5551	0.6746
0.1812	0.3805	300	0.3330	0.8300	0.8378	0.4882	0.6169
0.3265	0.4059	320	0.2961	0.8543	0.8280	0.6063	0.7
0.2217	0.4313	340	0.2970	0.8664	0.8065	0.6890	0.7431
0.2058	0.4567	360	0.3054	0.8521	0.8333	0.5906	0.6912
0.225	0.4820	380	0.3018	0.8576	0.8531	0.5945	0.7007
0.2045	0.5074	400	0.3174	0.8510	0.8742	0.5472	0.6731
0.2368	0.5328	420	0.3156	0.8477	0.8537	0.5512	0.6699
0.2162	0.5581	440	0.2928	0.8609	0.8441	0.6181	0.7136
0.1664	0.5835	460	0.2978	0.8598	0.8325	0.6260	0.7146
0.2282	0.6089	480	0.3031	0.8587	0.8539	0.5984	0.7037
0.1983	0.6342	500	0.2958	0.8543	0.8177	0.6181	0.7040
0.1843	0.6596	520	0.3055	0.8609	0.8556	0.6063	0.7097
0.1915	0.6850	540	0.2818	0.8675	0.8160	0.6811	0.7425
0.1582	0.7104	560	0.2887	0.8675	0.8641	0.6260	0.7260
0.2003	0.7357	580	0.2872	0.8653	0.8511	0.6299	0.7240
0.2345	0.7611	600	0.2827	0.8687	0.8293	0.6693	0.7407
0.2107	0.7865	620	0.2954	0.8642	0.8701	0.6063	0.7146
0.2562	0.8118	640	0.2938	0.8642	0.8503	0.6260	0.7211
0.1054	0.8372	660	0.2917	0.8642	0.8503	0.6260	0.7211
0.2837	0.8626	680	0.2842	0.8664	0.8376	0.6496	0.7317
0.1779	0.8879	700	0.2841	0.8709	0.8477	0.6575	0.7406
0.2277	0.9133	720	0.2847	0.8675	0.8384	0.6535	0.7345
0.2099	0.9387	740	0.2828	0.8720	0.8485	0.6614	0.7434
0.2167	0.9641	760	0.2835	0.8709	0.8477	0.6575	0.7406
0.1901	0.9894	780	0.2841	0.8687	0.8392	0.6575	0.7373

Framework versions

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