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: 0.0843
Rewards/chosen: 1.3063
Rewards/rejected: -3.4555
Rewards/accuracies: 0.9000
Rewards/margins: 4.7618
Logps/rejected: -51.0770
Logps/chosen: -17.6819
Logits/rejected: -3.8211
Logits/chosen: -3.7453

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: 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: 125

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6901	0.3333	25	0.6692	0.0087	-0.0409	0.7200	0.0495	-44.2478	-20.2772	-3.8170	-3.7449
0.5385	0.6667	50	0.4603	0.1249	-0.5004	0.8900	0.6253	-45.1669	-20.0447	-3.8171	-3.7449
0.2906	1.0	75	0.2278	0.4625	-1.4619	0.9000	1.9243	-47.0899	-19.3696	-3.8191	-3.7461
0.0574	1.3333	100	0.1164	0.8924	-2.9525	0.9000	3.8450	-50.0712	-18.5097	-3.8224	-3.7481
0.08	1.6667	125	0.0843	1.3063	-3.4555	0.9000	4.7618	-51.0770	-17.6819	-3.8211	-3.7453

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.19.2
Tokenizers 0.19.1

tsavage68/UTI_M2_125steps_1e7rate_05beta_CSFTDPO

UTI_M2_125steps_1e7rate_05beta_CSFTDPO

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tsavage68/
UTI_M2_125steps_1e7rate_05beta_CSFTDPO