v3_mistral_balance1_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: 0.0130
• Accuracy: 0.9980
• Precision: 0.9818
• Recall: 0.9474
• F1: 0.9643

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: 8569382
• 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	0.3256	0.9369	0.1429	0.2456	0.1806
0.4529	0.0258	20	0.2883	0.9484	0.1493	0.1754	0.1613
0.2483	0.0515	40	0.1461	0.9672	0.2	0.0526	0.0833
0.1622	0.0773	60	0.1080	0.9687	0.35	0.1228	0.1818
0.1243	0.1031	80	0.0879	0.9697	0.4524	0.3333	0.3838
0.0678	0.1289	100	0.0700	0.9692	0.4719	0.7368	0.5753
0.0301	0.1546	120	0.0474	0.9836	0.65	0.9123	0.7591
0.0105	0.1804	140	0.0342	0.9911	0.8421	0.8421	0.8421
0.041	0.2062	160	0.0333	0.9926	0.875	0.8596	0.8673
0.0291	0.2320	180	0.0268	0.9930	0.8308	0.9474	0.8852
0.0366	0.2577	200	0.0262	0.9916	0.7941	0.9474	0.864
0.0133	0.2835	220	0.0206	0.9921	0.8154	0.9298	0.8689
0.0075	0.3093	240	0.0188	0.9955	0.9444	0.8947	0.9189
0.0036	0.3351	260	0.0168	0.9945	0.9107	0.8947	0.9027
0.0081	0.3608	280	0.0182	0.9960	0.9153	0.9474	0.9310
0.0155	0.3866	300	0.0145	0.9980	0.9818	0.9474	0.9643
0.0075	0.4124	320	0.0165	0.9975	0.9643	0.9474	0.9558
0.0033	0.4381	340	0.0139	0.9975	0.9643	0.9474	0.9558
0.01	0.4639	360	0.0136	0.9970	0.9474	0.9474	0.9474
0.0018	0.4897	380	0.0146	0.9970	0.9474	0.9474	0.9474
0.0006	0.5155	400	0.0138	0.9975	0.9815	0.9298	0.9550
0.003	0.5412	420	0.0135	0.9965	0.9310	0.9474	0.9391
0.0035	0.5670	440	0.0141	0.9965	0.9808	0.8947	0.9358
0.0024	0.5928	460	0.0148	0.9965	0.9808	0.8947	0.9358
0.0203	0.6186	480	0.0136	0.9970	0.9474	0.9474	0.9474
0.0293	0.6443	500	0.0164	0.9970	0.9811	0.9123	0.9455
0.0078	0.6701	520	0.0149	0.9970	0.9474	0.9474	0.9474
0.0291	0.6959	540	0.0147	0.9975	0.9643	0.9474	0.9558
0.0119	0.7216	560	0.0136	0.9970	0.9474	0.9474	0.9474
0.002	0.7474	580	0.0138	0.9980	0.9818	0.9474	0.9643
0.0009	0.7732	600	0.0140	0.9980	0.9818	0.9474	0.9643
0.0022	0.7990	620	0.0134	0.9980	0.9818	0.9474	0.9643
0.0149	0.8247	640	0.0136	0.9980	0.9818	0.9474	0.9643
0.0397	0.8505	660	0.0140	0.9980	0.9818	0.9474	0.9643
0.0058	0.8763	680	0.0135	0.9980	0.9818	0.9474	0.9643
0.0153	0.9021	700	0.0132	0.9980	0.9818	0.9474	0.9643
0.0122	0.9278	720	0.0132	0.9980	0.9818	0.9474	0.9643
0.0276	0.9536	740	0.0132	0.9980	0.9818	0.9474	0.9643
0.0042	0.9794	760	0.0130	0.9980	0.9818	0.9474	0.9643

Framework versions

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