This model is a fine-tuned version of DUAL-GPO/phi-2-gpo-iter-0 on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.0108
Rewards/chosen: -0.0004
Rewards/rejected: -0.0003
Rewards/accuracies: 0.5
Rewards/margins: -0.0000
Logps/rejected: -278.6819
Logps/chosen: -306.4212
Logits/rejected: 0.0927
Logits/chosen: -0.0062

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-06
train_batch_size: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.0104	1.6	100	0.0107	-0.0005	-0.0009	0.5145	0.0005	-278.7423	-306.4289	0.0889	-0.0100

Framework versions

PEFT 0.7.1
Transformers 4.36.2
Pytorch 2.1.2+cu118
Datasets 2.14.6
Tokenizers 0.15.2

DUAL-GPO/phi-2-gpo-iter-1

phi-2-gpo-iter-1

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

DUAL-GPO/
phi-2-gpo-iter-1