This model is a fine-tuned version of tsavage68/UTI_M2_1000steps_1e5rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6865
Rewards/chosen: 0.0030
Rewards/rejected: -0.0105
Rewards/accuracies: 0.7700
Rewards/margins: 0.0134
Logps/rejected: -44.2707
Logps/chosen: -20.2650
Logits/rejected: -3.8169
Logits/chosen: -3.7448

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
0.6933	0.3333	25	0.6935	0.0005	0.0011	0.4100	-0.0006	-44.1551	-20.2900	-3.8168	-3.7448
0.6924	0.6667	50	0.6936	-0.0002	0.0007	0.4200	-0.0009	-44.1593	-20.2964	-3.8168	-3.7448
0.6926	1.0	75	0.6932	0.0004	0.0004	0.4900	0.0000	-44.1621	-20.2904	-3.8169	-3.7449
0.6928	1.3333	100	0.6935	0.0002	0.0009	0.4500	-0.0007	-44.1574	-20.2925	-3.8168	-3.7448
0.6925	1.6667	125	0.6921	0.0005	-0.0016	0.4900	0.0021	-44.1823	-20.2898	-3.8168	-3.7448
0.6907	2.0	150	0.6910	0.0009	-0.0035	0.5800	0.0044	-44.2011	-20.2855	-3.8168	-3.7448
0.6915	2.3333	175	0.6906	0.0017	-0.0035	0.6700	0.0052	-44.2011	-20.2776	-3.8169	-3.7448
0.6894	2.6667	200	0.6896	0.0017	-0.0055	0.6600	0.0072	-44.2212	-20.2776	-3.8169	-3.7448
0.6889	3.0	225	0.6896	0.0016	-0.0056	0.6900	0.0072	-44.2220	-20.2787	-3.8168	-3.7448
0.6886	3.3333	250	0.6891	0.0021	-0.0060	0.6900	0.0082	-44.2264	-20.2732	-3.8168	-3.7448
0.6892	3.6667	275	0.6880	0.0016	-0.0088	0.7700	0.0104	-44.2540	-20.2789	-3.8168	-3.7448
0.6869	4.0	300	0.6883	0.0022	-0.0075	0.7300	0.0097	-44.2411	-20.2728	-3.8168	-3.7448
0.6883	4.3333	325	0.6874	0.0022	-0.0094	0.75	0.0116	-44.2603	-20.2729	-3.8168	-3.7447
0.6874	4.6667	350	0.6875	0.0023	-0.0091	0.7600	0.0114	-44.2574	-20.2714	-3.8168	-3.7448
0.6872	5.0	375	0.6866	0.0025	-0.0107	0.7700	0.0133	-44.2736	-20.2695	-3.8168	-3.7448
0.6871	5.3333	400	0.6872	0.0017	-0.0103	0.7600	0.0120	-44.2692	-20.2774	-3.8168	-3.7448
0.6864	5.6667	425	0.6867	0.0026	-0.0104	0.8400	0.0129	-44.2697	-20.2690	-3.8168	-3.7448
0.6867	6.0	450	0.6861	0.0023	-0.0118	0.7800	0.0141	-44.2843	-20.2715	-3.8169	-3.7449
0.6851	6.3333	475	0.6867	0.0023	-0.0106	0.8100	0.0129	-44.2724	-20.2714	-3.8167	-3.7447
0.688	6.6667	500	0.6869	0.0025	-0.0101	0.7600	0.0125	-44.2667	-20.2696	-3.8168	-3.7447
0.6849	7.0	525	0.6867	0.0020	-0.0110	0.7300	0.0130	-44.2760	-20.2744	-3.8168	-3.7447
0.6864	7.3333	550	0.6869	0.0026	-0.0100	0.75	0.0125	-44.2656	-20.2690	-3.8168	-3.7448
0.6867	7.6667	575	0.6855	0.0028	-0.0126	0.8300	0.0154	-44.2921	-20.2669	-3.8168	-3.7447
0.6877	8.0	600	0.6868	0.0020	-0.0108	0.7600	0.0128	-44.2742	-20.2744	-3.8168	-3.7448
0.6859	8.3333	625	0.6858	0.0024	-0.0124	0.8100	0.0148	-44.2902	-20.2710	-3.8169	-3.7448
0.6864	8.6667	650	0.6869	0.0025	-0.0101	0.7600	0.0127	-44.2672	-20.2691	-3.8168	-3.7447
0.6856	9.0	675	0.6864	0.0026	-0.0111	0.8000	0.0136	-44.2768	-20.2688	-3.8168	-3.7447
0.6865	9.3333	700	0.6865	0.0023	-0.0111	0.7800	0.0135	-44.2773	-20.2711	-3.8168	-3.7448
0.6868	9.6667	725	0.6865	0.0022	-0.0112	0.7600	0.0134	-44.2784	-20.2729	-3.8168	-3.7447
0.6867	10.0	750	0.6869	0.0022	-0.0104	0.7400	0.0126	-44.2698	-20.2721	-3.8169	-3.7448
0.6847	10.3333	775	0.6861	0.0026	-0.0116	0.8100	0.0141	-44.2817	-20.2689	-3.8168	-3.7448
0.6862	10.6667	800	0.6855	0.0030	-0.0124	0.8400	0.0154	-44.2899	-20.2643	-3.8168	-3.7448
0.6868	11.0	825	0.6862	0.0028	-0.0113	0.7800	0.0141	-44.2789	-20.2667	-3.8169	-3.7448
0.6852	11.3333	850	0.6864	0.0029	-0.0107	0.7700	0.0136	-44.2728	-20.2657	-3.8169	-3.7448
0.6871	11.6667	875	0.6865	0.0030	-0.0105	0.7700	0.0134	-44.2707	-20.2650	-3.8169	-3.7448
0.6875	12.0	900	0.6865	0.0030	-0.0105	0.7700	0.0134	-44.2707	-20.2650	-3.8169	-3.7448
0.6852	12.3333	925	0.6865	0.0030	-0.0105	0.7700	0.0134	-44.2707	-20.2650	-3.8169	-3.7448
0.6863	12.6667	950	0.6865	0.0030	-0.0105	0.7700	0.0134	-44.2707	-20.2650	-3.8169	-3.7448
0.6872	13.0	975	0.6865	0.0030	-0.0105	0.7700	0.0134	-44.2707	-20.2650	-3.8169	-3.7448
0.6879	13.3333	1000	0.6865	0.0030	-0.0105	0.7700	0.0134	-44.2707	-20.2650	-3.8169	-3.7448

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.19.2
Tokenizers 0.19.1

tsavage68/UTI_M2_1000steps_1e8rate_01beta_CSFTDPO

UTI_M2_1000steps_1e8rate_01beta_CSFTDPO

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tsavage68/
UTI_M2_1000steps_1e8rate_01beta_CSFTDPO