This model is a fine-tuned version of tsavage68/chat_600STEPS_1e8rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6764
Rewards/chosen: -0.0580
Rewards/rejected: -0.0983
Rewards/accuracies: 0.5209
Rewards/margins: 0.0403
Logps/rejected: -19.7850
Logps/chosen: -17.3346
Logits/rejected: -0.6386
Logits/chosen: -0.6384

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: 5e-07
train_batch_size: 4
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
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
0.6924	0.0977	50	0.6933	0.0017	0.0020	0.4154	-0.0003	-18.7815	-16.7372	-0.5990	-0.5988
0.6889	0.1953	100	0.6896	-0.0103	-0.0178	0.4769	0.0075	-18.9805	-16.8580	-0.6027	-0.6025
0.6916	0.2930	150	0.6883	-0.0335	-0.0443	0.4945	0.0109	-19.2454	-17.0895	-0.6042	-0.6041
0.6896	0.3906	200	0.6862	-0.0223	-0.0382	0.4835	0.0159	-19.1840	-16.9772	-0.6024	-0.6023
0.6921	0.4883	250	0.6851	-0.0229	-0.0418	0.5011	0.0188	-19.2195	-16.9840	-0.6008	-0.6007
0.6846	0.5859	300	0.6811	-0.0344	-0.0622	0.5099	0.0277	-19.4235	-17.0989	-0.6126	-0.6125
0.6801	0.6836	350	0.6812	-0.0454	-0.0734	0.5275	0.0279	-19.5357	-17.2091	-0.6080	-0.6079
0.6757	0.7812	400	0.6796	-0.0410	-0.0732	0.5275	0.0322	-19.5340	-17.1644	-0.6197	-0.6196
0.6822	0.8789	450	0.6794	-0.0353	-0.0681	0.5275	0.0328	-19.4827	-17.1072	-0.6122	-0.6121
0.6728	0.9766	500	0.6787	-0.0351	-0.0699	0.5121	0.0348	-19.5013	-17.1061	-0.6178	-0.6176
0.6186	1.0742	550	0.6781	-0.0359	-0.0723	0.5209	0.0363	-19.5246	-17.1139	-0.6238	-0.6237
0.6205	1.1719	600	0.6777	-0.0476	-0.0850	0.5275	0.0374	-19.6518	-17.2306	-0.6283	-0.6282
0.6273	1.2695	650	0.6772	-0.0533	-0.0918	0.5209	0.0385	-19.7197	-17.2874	-0.6341	-0.6340
0.6067	1.3672	700	0.6762	-0.0522	-0.0927	0.5341	0.0405	-19.7290	-17.2762	-0.6358	-0.6356
0.5987	1.4648	750	0.6759	-0.0558	-0.0971	0.5363	0.0413	-19.7734	-17.3129	-0.6378	-0.6377
0.6302	1.5625	800	0.6761	-0.0560	-0.0969	0.5363	0.0409	-19.7707	-17.3142	-0.6384	-0.6382
0.6155	1.6602	850	0.6765	-0.0572	-0.0973	0.5363	0.0401	-19.7749	-17.3264	-0.6388	-0.6386
0.6308	1.7578	900	0.6760	-0.0572	-0.0982	0.5253	0.0411	-19.7842	-17.3263	-0.6389	-0.6388
0.6092	1.8555	950	0.6760	-0.0570	-0.0981	0.5407	0.0411	-19.7825	-17.3244	-0.6387	-0.6386
0.6166	1.9531	1000	0.6764	-0.0580	-0.0983	0.5209	0.0403	-19.7850	-17.3346	-0.6386	-0.6384

Framework versions

Transformers 4.40.1
Pytorch 2.0.0+cu117
Datasets 2.19.1
Tokenizers 0.19.1

tsavage68/chat_1000_STEPS_01beta_5e7rate_CDPOSFT

chat_1000_STEPS_01beta_5e7rate_CDPOSFT

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tsavage68/
chat_1000_STEPS_01beta_5e7rate_CDPOSFT