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: 0.2432
Rewards/chosen: 1.4890
Rewards/rejected: -5.6246
Rewards/accuracies: 0.6500
Rewards/margins: 7.1137
Logps/rejected: -47.2334
Logps/chosen: -14.1398
Logits/rejected: -1.1887
Logits/chosen: -1.1714

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-06
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
0.6278	0.3333	25	0.4801	0.3140	-0.2333	0.6500	0.5473	-29.2623	-18.0567	-1.1541	-1.1498
0.2477	0.6667	50	0.2478	1.1053	-2.8214	0.6500	3.9267	-37.8892	-15.4190	-1.1662	-1.1580
0.1737	1.0	75	0.2445	1.2122	-3.7805	0.6500	4.9927	-41.0862	-15.0625	-1.1740	-1.1629
0.1041	1.3333	100	0.2462	1.2327	-4.0476	0.6500	5.2803	-41.9766	-14.9944	-1.1757	-1.1637
0.1907	1.6667	125	0.2621	1.3998	-4.1257	0.6400	5.5255	-42.2369	-14.4372	-1.1747	-1.1616
0.3639	2.0	150	0.2436	1.5347	-4.4832	0.6500	6.0179	-43.4288	-13.9878	-1.1776	-1.1614
0.26	2.3333	175	0.2436	1.5307	-4.5732	0.6500	6.1039	-43.7287	-14.0010	-1.1779	-1.1616
0.2253	2.6667	200	0.2436	1.5263	-4.6709	0.6500	6.1971	-44.0543	-14.0158	-1.1786	-1.1621
0.208	3.0	225	0.2434	1.5214	-4.7570	0.6500	6.2784	-44.3412	-14.0318	-1.1797	-1.1631
0.2253	3.3333	250	0.2435	1.5247	-4.8509	0.6500	6.3757	-44.6545	-14.0209	-1.1802	-1.1636
0.1733	3.6667	275	0.2433	1.5180	-4.9306	0.6500	6.4486	-44.9200	-14.0431	-1.1808	-1.1642
0.2773	4.0	300	0.2434	1.5162	-5.0012	0.6500	6.5173	-45.1552	-14.0494	-1.1814	-1.1647
0.2426	4.3333	325	0.2434	1.5187	-5.0818	0.6500	6.6005	-45.4240	-14.0408	-1.1823	-1.1655
0.156	4.6667	350	0.2434	1.5119	-5.1300	0.6500	6.6419	-45.5845	-14.0636	-1.1827	-1.1660
0.2253	5.0	375	0.2433	1.5126	-5.2045	0.6500	6.7172	-45.8331	-14.0612	-1.1835	-1.1667
0.2253	5.3333	400	0.2433	1.5050	-5.2614	0.6500	6.7665	-46.0227	-14.0865	-1.1840	-1.1672
0.2253	5.6667	425	0.2433	1.5083	-5.3091	0.6500	6.8173	-46.1816	-14.0758	-1.1847	-1.1677
0.208	6.0	450	0.2434	1.5053	-5.3459	0.6500	6.8513	-46.3044	-14.0855	-1.1850	-1.1681
0.2773	6.3333	475	0.2433	1.5037	-5.3838	0.6500	6.8875	-46.4306	-14.0908	-1.1853	-1.1683
0.3119	6.6667	500	0.2433	1.5033	-5.4082	0.6500	6.9115	-46.5121	-14.0923	-1.1858	-1.1688
0.208	7.0	525	0.2433	1.5061	-5.4546	0.6500	6.9607	-46.6668	-14.0830	-1.1865	-1.1695
0.1733	7.3333	550	0.2433	1.5023	-5.4820	0.6500	6.9843	-46.7581	-14.0957	-1.1867	-1.1696
0.2599	7.6667	575	0.2433	1.4981	-5.5026	0.6500	7.0007	-46.8266	-14.1095	-1.1869	-1.1698
0.2599	8.0	600	0.2433	1.4959	-5.5358	0.6500	7.0317	-46.9373	-14.1169	-1.1875	-1.1704
0.2253	8.3333	625	0.2432	1.4946	-5.5465	0.6500	7.0411	-46.9730	-14.1212	-1.1882	-1.1710
0.104	8.6667	650	0.2433	1.4914	-5.5586	0.6500	7.0500	-47.0135	-14.1321	-1.1880	-1.1708
0.2253	9.0	675	0.2433	1.4928	-5.5797	0.6500	7.0725	-47.0836	-14.1273	-1.1886	-1.1714
0.2253	9.3333	700	0.2433	1.4954	-5.5899	0.6500	7.0853	-47.1178	-14.1188	-1.1886	-1.1713
0.2253	9.6667	725	0.2433	1.4911	-5.6004	0.6500	7.0915	-47.1527	-14.1328	-1.1886	-1.1714
0.3119	10.0	750	0.2432	1.4901	-5.6089	0.6500	7.0990	-47.1810	-14.1364	-1.1888	-1.1716
0.2079	10.3333	775	0.2433	1.4923	-5.6105	0.6500	7.1028	-47.1863	-14.1288	-1.1888	-1.1717
0.2253	10.6667	800	0.2433	1.4920	-5.6175	0.6500	7.1095	-47.2096	-14.1299	-1.1887	-1.1714
0.2426	11.0	825	0.2432	1.4932	-5.6125	0.6500	7.1057	-47.1930	-14.1260	-1.1891	-1.1719
0.2946	11.3333	850	0.2432	1.4899	-5.6309	0.6500	7.1207	-47.2542	-14.1371	-1.1888	-1.1716
0.1733	11.6667	875	0.2433	1.4901	-5.6244	0.6500	7.1145	-47.2326	-14.1363	-1.1888	-1.1716
0.156	12.0	900	0.2433	1.4904	-5.6265	0.6500	7.1169	-47.2397	-14.1353	-1.1889	-1.1717
0.1906	12.3333	925	0.2433	1.4895	-5.6198	0.6500	7.1092	-47.2172	-14.1384	-1.1889	-1.1717
0.2426	12.6667	950	0.2432	1.4854	-5.6283	0.6500	7.1136	-47.2455	-14.1520	-1.1886	-1.1714
0.2079	13.0	975	0.2432	1.4890	-5.6246	0.6500	7.1137	-47.2334	-14.1398	-1.1887	-1.1714
0.3119	13.3333	1000	0.2432	1.4890	-5.6246	0.6500	7.1137	-47.2334	-14.1398	-1.1887	-1.1714

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.19.2
Tokenizers 0.19.1

tsavage68/UTI2_L3_1000steps_1e6rate_03beta_CSFTDPO

UTI2_L3_1000steps_1e6rate_03beta_CSFTDPO

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tsavage68/
UTI2_L3_1000steps_1e6rate_03beta_CSFTDPO