metadata

license: llama3
base_model: tsavage68/Summary_L3_1000steps_1e7rate_SFT2
tags:
  - trl
  - dpo
  - generated_from_trainer
model-index:
  - name: Summary_L3_100steps_1e8rate_05beta_CSFTDPO
    results: []

Summary_L3_100steps_1e8rate_05beta_CSFTDPO

This model is a fine-tuned version of tsavage68/Summary_L3_1000steps_1e7rate_SFT2 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6879
Rewards/chosen: -0.0012
Rewards/rejected: -0.0138
Rewards/accuracies: 0.1000
Rewards/margins: 0.0126
Logps/rejected: -15.2914
Logps/chosen: -9.3853
Logits/rejected: -1.0958
Logits/chosen: -1.0972

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-08
train_batch_size: 1
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 4
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
0.6824	0.2004	50	0.6901	0.0066	-0.0020	0.0850	0.0086	-15.2678	-9.3695	-1.0960	-1.0974
0.6926	0.4008	100	0.6879	-0.0012	-0.0138	0.1000	0.0126	-15.2914	-9.3853	-1.0958	-1.0972

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.20.0
Tokenizers 0.19.1