This model is a fine-tuned version of tsavage68/IE_L3_1000steps_1e6rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.1816
Rewards/chosen: -0.5946
Rewards/rejected: -5.5441
Rewards/accuracies: 0.7400
Rewards/margins: 4.9495
Logps/rejected: -94.1076
Logps/chosen: -84.7799
Logits/rejected: -0.7931
Logits/chosen: -0.7224

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
0.5177	0.4	50	0.4502	-0.0545	-0.6708	0.7400	0.6163	-77.8632	-82.9793	-0.7959	-0.7377
0.1418	0.8	100	0.1816	-0.5946	-5.5441	0.7400	4.9495	-94.1076	-84.7799	-0.7931	-0.7224

Framework versions

Transformers 4.44.2
Pytorch 2.0.0+cu117
Datasets 3.0.0
Tokenizers 0.19.1

RichardErkhov/tsavage68_-_IE_L3_100steps_1e7rate_03beta_cSFTDPO-4bits

Original model description:

IE_L3_100steps_1e7rate_03beta_cSFTDPO

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

RichardErkhov/
tsavage68_-_IE_L3_100steps_1e7rate_03beta_cSFTDPO-4bits