This model is a fine-tuned version of tsavage68/IE_M2_1000steps_1e7rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.3747
Rewards/chosen: -0.1730
Rewards/rejected: -3.5887
Rewards/accuracies: 0.4600
Rewards/margins: 3.4157
Logps/rejected: -76.9091
Logps/chosen: -43.9359
Logits/rejected: -2.8759
Logits/chosen: -2.8145

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: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.558	0.4	50	0.4553	-0.0349	-0.8002	0.4600	0.7653	-49.0237	-42.5545	-2.9038	-2.8422
0.3818	0.8	100	0.3747	-0.1730	-3.5887	0.4600	3.4157	-76.9091	-43.9359	-2.8759	-2.8145

Framework versions

Transformers 4.44.2
Pytorch 2.0.0+cu117
Datasets 3.0.0
Tokenizers 0.19.1

tsavage68/IE_M2_100steps_1e7rate_01beta_cSFTDPO

IE_M2_100steps_1e7rate_01beta_cSFTDPO

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tsavage68/
IE_M2_100steps_1e7rate_01beta_cSFTDPO