v2b_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: 0.3066
• Accuracy: 0.8603
• Precision: 0.8713
• Recall: 0.5889
• F1: 0.7028

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	0.5994	0.7384	0.6545	0.1423	0.2338
0.572	0.0186	20	0.5946	0.7395	0.6731	0.1383	0.2295
0.5829	0.0371	40	0.5678	0.7494	0.7143	0.1779	0.2848
0.4808	0.0557	60	0.5129	0.7694	0.6744	0.3439	0.4555
0.498	0.0742	80	0.4658	0.7805	0.6708	0.4269	0.5217
0.2531	0.0928	100	0.4835	0.8016	0.7312	0.4625	0.5666
0.2925	0.1113	120	0.5003	0.8016	0.8304	0.3676	0.5096
0.1912	0.1299	140	0.4575	0.8004	0.8411	0.3557	0.5
0.1991	0.1484	160	0.4109	0.8115	0.8374	0.4071	0.5479
0.2153	0.1670	180	0.3718	0.8337	0.8456	0.4980	0.6269
0.1638	0.1855	200	0.3657	0.8237	0.8672	0.4387	0.5827
0.2033	0.2041	220	0.3455	0.8370	0.8354	0.5217	0.6423
0.2448	0.2226	240	0.3438	0.8381	0.8497	0.5138	0.6404
0.2337	0.2412	260	0.3705	0.8282	0.8828	0.4466	0.5932
0.1698	0.2597	280	0.3724	0.8215	0.8710	0.4269	0.5729
0.1607	0.2783	300	0.3455	0.8293	0.8722	0.4585	0.6010
0.1671	0.2968	320	0.3371	0.8337	0.8503	0.4941	0.625
0.1809	0.3154	340	0.3406	0.8514	0.8287	0.5929	0.6912
0.1672	0.3340	360	0.3520	0.8392	0.8699	0.5020	0.6366
0.153	0.3525	380	0.3273	0.8459	0.8562	0.5415	0.6634
0.2	0.3711	400	0.3307	0.8448	0.8599	0.5336	0.6585
0.2082	0.3896	420	0.3143	0.8603	0.8396	0.6206	0.7136
0.2051	0.4082	440	0.3139	0.8570	0.8563	0.5889	0.6979
0.0959	0.4267	460	0.3130	0.8570	0.8523	0.5929	0.6993
0.1955	0.4453	480	0.3044	0.8592	0.8462	0.6087	0.7080
0.1904	0.4638	500	0.3389	0.8404	0.8759	0.5020	0.6382
0.1809	0.4824	520	0.3319	0.8459	0.8701	0.5296	0.6585
0.1605	0.5009	540	0.3016	0.8614	0.8678	0.5968	0.7073
0.2123	0.5195	560	0.2983	0.8603	0.8396	0.6206	0.7136
0.2279	0.5380	580	0.3046	0.8559	0.8361	0.6047	0.7018
0.2224	0.5566	600	0.3395	0.8381	0.8741	0.4941	0.6313
0.1655	0.5751	620	0.3388	0.8359	0.8777	0.4822	0.6224
0.1468	0.5937	640	0.3022	0.8592	0.8424	0.6126	0.7094
0.1421	0.6122	660	0.3297	0.8437	0.8784	0.5138	0.6484
0.2483	0.6308	680	0.3060	0.8525	0.8529	0.5731	0.6856
0.1411	0.6494	700	0.3171	0.8481	0.8537	0.5534	0.6715
0.2015	0.6679	720	0.3120	0.8525	0.8614	0.5652	0.6826
0.2216	0.6865	740	0.3030	0.8503	0.8598	0.5573	0.6763
0.1936	0.7050	760	0.3091	0.8503	0.8598	0.5573	0.6763
0.135	0.7236	780	0.3023	0.8525	0.8529	0.5731	0.6856
0.1332	0.7421	800	0.3207	0.8437	0.8836	0.5099	0.6466
0.249	0.7607	820	0.3031	0.8592	0.8706	0.5850	0.6998
0.2033	0.7792	840	0.3076	0.8592	0.875	0.5810	0.6983
0.1418	0.7978	860	0.2998	0.8614	0.8678	0.5968	0.7073
0.1826	0.8163	880	0.3014	0.8625	0.8728	0.5968	0.7089
0.1538	0.8349	900	0.3092	0.8614	0.8855	0.5810	0.7017
0.1762	0.8534	920	0.3011	0.8603	0.8671	0.5929	0.7042
0.1561	0.8720	940	0.2998	0.8603	0.8671	0.5929	0.7042
0.1633	0.8905	960	0.3064	0.8570	0.8690	0.5771	0.6936
0.1452	0.9091	980	0.3034	0.8603	0.8713	0.5889	0.7028
0.086	0.9276	1000	0.3051	0.8581	0.8698	0.5810	0.6967
0.1909	0.9462	1020	0.3055	0.8581	0.8698	0.5810	0.6967
0.2017	0.9647	1040	0.3058	0.8581	0.8743	0.5771	0.6952
0.1828	0.9833	1060	0.3066	0.8603	0.8713	0.5889	0.7028

Framework versions

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