Summary_L3_1000steps_1e6rate_05beta_CSFTDPO

This model is a fine-tuned version of tsavage68/Summary_L3_1000steps_1e7rate_SFT2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5961
• Rewards/chosen: 0.1158
• Rewards/rejected: -2.7330
• Rewards/accuracies: 0.1400
• Rewards/margins: 2.8488
• Logps/rejected: -20.7298
• Logps/chosen: -9.1512
• Logits/rejected: -1.1135
• Logits/chosen: -1.1149

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• 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
0.555	0.2004	50	0.5962	0.0976	-1.3577	0.1400	1.4553	-17.9791	-9.1876	-1.0985	-1.1002
0.6585	0.4008	100	0.5962	0.1094	-1.5231	0.1400	1.6326	-18.3100	-9.1639	-1.1003	-1.1019
0.6238	0.6012	150	0.5961	0.1341	-2.2789	0.1400	2.4130	-19.8216	-9.1145	-1.1048	-1.1065
0.6065	0.8016	200	0.5961	0.1193	-2.7271	0.1400	2.8464	-20.7179	-9.1442	-1.1137	-1.1150
0.6238	1.0020	250	0.5961	0.1211	-2.7359	0.1400	2.8570	-20.7355	-9.1407	-1.1133	-1.1146
0.6238	1.2024	300	0.5961	0.1211	-2.7359	0.1400	2.8570	-20.7355	-9.1407	-1.1133	-1.1146
0.6238	1.4028	350	0.5961	0.1226	-2.7319	0.1400	2.8545	-20.7275	-9.1376	-1.1131	-1.1144
0.5718	1.6032	400	0.5961	0.1226	-2.7319	0.1400	2.8545	-20.7275	-9.1376	-1.1131	-1.1144
0.5892	1.8036	450	0.5961	0.1196	-2.7246	0.1400	2.8442	-20.7129	-9.1435	-1.1135	-1.1147
0.5718	2.0040	500	0.5961	0.1211	-2.7256	0.1400	2.8467	-20.7150	-9.1406	-1.1135	-1.1147
0.5718	2.2044	550	0.5961	0.1207	-2.7233	0.1400	2.8439	-20.7103	-9.1414	-1.1134	-1.1147
0.5545	2.4048	600	0.5961	0.1207	-2.7233	0.1400	2.8439	-20.7103	-9.1414	-1.1134	-1.1147
0.5199	2.6052	650	0.5961	0.1207	-2.7233	0.1400	2.8439	-20.7103	-9.1414	-1.1134	-1.1147
0.6238	2.8056	700	0.5961	0.1207	-2.7233	0.1400	2.8439	-20.7103	-9.1414	-1.1134	-1.1147
0.6065	3.0060	750	0.5961	0.1181	-2.7332	0.1400	2.8513	-20.7302	-9.1466	-1.1134	-1.1147
0.6412	3.2064	800	0.5961	0.1124	-2.7370	0.1400	2.8494	-20.7378	-9.1580	-1.1135	-1.1148
0.6585	3.4068	850	0.5961	0.1124	-2.7370	0.1400	2.8494	-20.7378	-9.1580	-1.1135	-1.1148
0.6238	3.6072	900	0.5961	0.1148	-2.7352	0.1400	2.8500	-20.7342	-9.1532	-1.1135	-1.1149
0.5372	3.8076	950	0.5961	0.1148	-2.7352	0.1400	2.8500	-20.7342	-9.1532	-1.1135	-1.1149
0.6238	4.0080	1000	0.5961	0.1158	-2.7330	0.1400	2.8488	-20.7298	-9.1512	-1.1135	-1.1149

Framework versions

• Transformers 4.41.2
• Pytorch 2.0.0+cu117
• Datasets 2.20.0
• Tokenizers 0.19.1