This model is a fine-tuned version of tsavage68/UTI_L3_1000steps_1e5rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6909
Rewards/chosen: 0.0027
Rewards/rejected: -0.0021
Rewards/accuracies: 0.5600
Rewards/margins: 0.0049
Logps/rejected: -43.2766
Logps/chosen: -29.2159
Logits/rejected: -1.1412
Logits/chosen: -1.1365

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: 1e-08
train_batch_size: 2
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 4
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6931	0.3333	25	0.6920	0.0001	-0.0023	0.1400	0.0024	-43.2770	-29.2245	-1.1409	-1.1362
0.6955	0.6667	50	0.6923	0.0082	0.0062	0.5	0.0020	-43.2487	-29.1975	-1.1412	-1.1365
0.6918	1.0	75	0.6917	0.0083	0.0050	0.5300	0.0033	-43.2530	-29.1974	-1.1411	-1.1364
0.7011	1.3333	100	0.6945	0.0070	0.0091	0.4700	-0.0022	-43.2391	-29.2017	-1.1414	-1.1367
0.6875	1.6667	125	0.6921	-0.0051	-0.0076	0.5100	0.0026	-43.2949	-29.2418	-1.1411	-1.1364
0.694	2.0	150	0.6935	0.0011	0.0014	0.4900	-0.0003	-43.2648	-29.2211	-1.1411	-1.1364
0.6926	2.3333	175	0.6930	-0.0018	-0.0026	0.4700	0.0008	-43.2781	-29.2309	-1.1410	-1.1363
0.6957	2.6667	200	0.6898	0.0072	0.0001	0.5800	0.0071	-43.2691	-29.2008	-1.1410	-1.1363
0.6949	3.0	225	0.6913	0.0019	-0.0023	0.4800	0.0042	-43.2772	-29.2187	-1.1412	-1.1365
0.6951	3.3333	250	0.6904	-0.0004	-0.0064	0.5600	0.0060	-43.2910	-29.2264	-1.1410	-1.1363
0.6936	3.6667	275	0.6912	0.0049	0.0006	0.5500	0.0043	-43.2676	-29.2086	-1.1411	-1.1364
0.6901	4.0	300	0.6922	-0.0012	-0.0035	0.5100	0.0024	-43.2812	-29.2288	-1.1414	-1.1367
0.6932	4.3333	325	0.6919	-0.0003	-0.0033	0.5400	0.0029	-43.2804	-29.2260	-1.1411	-1.1364
0.6905	4.6667	350	0.6909	0.0004	-0.0043	0.5200	0.0048	-43.2840	-29.2234	-1.1412	-1.1365
0.6892	5.0	375	0.6901	0.0065	0.0000	0.5200	0.0065	-43.2695	-29.2032	-1.1411	-1.1364
0.6956	5.3333	400	0.6925	0.0113	0.0096	0.5500	0.0017	-43.2374	-29.1872	-1.1413	-1.1365
0.6798	5.6667	425	0.6890	0.0022	-0.0065	0.6100	0.0086	-43.2910	-29.2177	-1.1411	-1.1364
0.692	6.0	450	0.6877	0.0019	-0.0094	0.5700	0.0113	-43.3010	-29.2186	-1.1410	-1.1363
0.6882	6.3333	475	0.6901	-0.0026	-0.0092	0.5500	0.0066	-43.3002	-29.2336	-1.1409	-1.1363
0.6918	6.6667	500	0.6896	0.0010	-0.0067	0.5800	0.0077	-43.2917	-29.2216	-1.1411	-1.1364
0.6905	7.0	525	0.6902	0.0041	-0.0021	0.5400	0.0061	-43.2764	-29.2114	-1.1408	-1.1361
0.6949	7.3333	550	0.6884	0.0043	-0.0055	0.5300	0.0098	-43.2879	-29.2105	-1.1413	-1.1365
0.6945	7.6667	575	0.6885	0.0086	-0.0012	0.6100	0.0098	-43.2735	-29.1963	-1.1413	-1.1366
0.6903	8.0	600	0.6914	0.0027	-0.0013	0.5200	0.0040	-43.2737	-29.2159	-1.1409	-1.1362
0.6902	8.3333	625	0.6905	0.0041	-0.0016	0.6100	0.0057	-43.2748	-29.2111	-1.1410	-1.1363
0.689	8.6667	650	0.6903	0.0016	-0.0045	0.5200	0.0061	-43.2844	-29.2195	-1.1410	-1.1363
0.6973	9.0	675	0.6887	0.0005	-0.0089	0.5900	0.0094	-43.2992	-29.2234	-1.1410	-1.1363
0.6976	9.3333	700	0.6913	0.0040	-0.0001	0.5300	0.0041	-43.2698	-29.2117	-1.1410	-1.1363
0.6914	9.6667	725	0.6921	0.0047	0.0022	0.5600	0.0026	-43.2622	-29.2091	-1.1410	-1.1364
0.6921	10.0	750	0.6923	0.0028	0.0006	0.5200	0.0022	-43.2675	-29.2157	-1.1411	-1.1364
0.6946	10.3333	775	0.6912	0.0017	-0.0027	0.5400	0.0044	-43.2784	-29.2192	-1.1412	-1.1365
0.6901	10.6667	800	0.6908	0.0029	-0.0022	0.5600	0.0051	-43.2769	-29.2152	-1.1412	-1.1365
0.7002	11.0	825	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365
0.6928	11.3333	850	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365
0.6915	11.6667	875	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365
0.6927	12.0	900	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365
0.6923	12.3333	925	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365
0.6935	12.6667	950	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365
0.6877	13.0	975	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365
0.692	13.3333	1000	0.6909	0.0027	-0.0021	0.5600	0.0049	-43.2766	-29.2159	-1.1412	-1.1365

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.19.2
Tokenizers 0.19.1

tsavage68/UTI2_L3_1000steps_1e8rate_03beta_CSFTDPO

UTI2_L3_1000steps_1e8rate_01beta_CSFTDPO

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tsavage68/
UTI2_L3_1000steps_1e8rate_03beta_CSFTDPO