This model is a fine-tuned version of tsavage68/Summary4500_M2_200steps_1e7rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0015
Rewards/chosen: -2.4830
Rewards/rejected: -19.5796
Rewards/accuracies: 0.9980
Rewards/margins: 17.0965
Logps/rejected: -217.9950
Logps/chosen: -102.0166
Logits/rejected: -2.2164
Logits/chosen: -2.1724

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-07
train_batch_size: 1
eval_batch_size: 1
seed: 42
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
0.2747	0.0112	50	0.2959	-0.0178	-1.1712	0.9980	1.1534	-156.6336	-93.7990	-2.3394	-2.2921
0.0016	0.0224	100	0.0056	-1.1134	-9.6882	0.9980	8.5748	-185.0238	-97.4512	-2.2610	-2.2152
0.0	0.0336	150	0.0026	-1.7239	-13.2278	0.9980	11.5039	-196.8225	-99.4863	-2.2486	-2.2032
0.0002	0.0448	200	0.0019	-1.7692	-15.4130	0.9980	13.6438	-204.1066	-99.6372	-2.2353	-2.1900
0.0	0.0559	250	0.0018	-1.8456	-16.0238	0.9980	14.1782	-206.1425	-99.8917	-2.2328	-2.1878
0.0	0.0671	300	0.0017	-1.9778	-16.6087	0.9980	14.6309	-208.0922	-100.3325	-2.2295	-2.1847
0.0002	0.0783	350	0.0016	-2.2605	-17.8597	0.9980	15.5992	-212.2620	-101.2747	-2.2241	-2.1798
0.0002	0.0895	400	0.0016	-2.4489	-19.0704	0.9980	16.6215	-216.2979	-101.9028	-2.2152	-2.1712
0.0	0.1007	450	0.0016	-2.4489	-19.0868	0.9980	16.6378	-216.3524	-101.9029	-2.2152	-2.1712
0.0	0.1119	500	0.0016	-2.4479	-19.1013	0.9980	16.6533	-216.4006	-101.8995	-2.2154	-2.1714
0.0	0.1231	550	0.0015	-2.4619	-19.3418	0.9980	16.8799	-217.2026	-101.9461	-2.2168	-2.1728
0.0	0.1343	600	0.0015	-2.4651	-19.3838	0.9980	16.9188	-217.3425	-101.9567	-2.2170	-2.1729
0.0	0.1454	650	0.0015	-2.4805	-19.4813	0.9980	17.0008	-217.6673	-102.0080	-2.2170	-2.1731
0.0	0.1566	700	0.0015	-2.4872	-19.5320	0.9980	17.0448	-217.8365	-102.0304	-2.2172	-2.1732
0.0	0.1678	750	0.0015	-2.4850	-19.5486	0.9980	17.0636	-217.8918	-102.0233	-2.2162	-2.1722
0.0	0.1790	800	0.0015	-2.4848	-19.5861	0.9980	17.1012	-218.0167	-102.0225	-2.2170	-2.1729
0.0	0.1902	850	0.0016	-2.4833	-19.5564	0.9980	17.0731	-217.9177	-102.0174	-2.2161	-2.1721
0.0	0.2014	900	0.0015	-2.4875	-19.5685	0.9980	17.0810	-217.9580	-102.0313	-2.2166	-2.1726
0.0	0.2126	950	0.0015	-2.4836	-19.5789	0.9980	17.0953	-217.9927	-102.0184	-2.2164	-2.1724
0.0	0.2238	1000	0.0015	-2.4830	-19.5796	0.9980	17.0965	-217.9950	-102.0166	-2.2164	-2.1724

Framework versions

Transformers 4.42.4
Pytorch 2.0.0+cu117
Datasets 2.20.0
Tokenizers 0.19.1

tsavage68/Summary4500_M2_1000steps_1e7rate_03beta_CSFTDPO

Hyponatremia_M2_1000steps_1e7rate_03beta_CSFTDPO

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tsavage68/
Summary4500_M2_1000steps_1e7rate_03beta_CSFTDPO