dpo_t5_2

This model is a fine-tuned version of zera09/long_t5_4 on the None dataset. It achieves the following results on the evaluation set:

• eval_loss: 0.3486
• eval_runtime: 189.2688
• eval_samples_per_second: 8.454
• eval_steps_per_second: 1.057
• eval_rewards/chosen: 0.4122
• eval_rewards/rejected: -0.7810
• eval_rewards/accuracies: 0.8781
• eval_rewards/margins: 1.1932
• eval_logps/rejected: -6.7047
• eval_logps/chosen: -26.4297
• eval_logits/rejected: -19.0942
• eval_logits/chosen: -18.7712
• epoch: 2.3
• step: 460

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-07
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• training_steps: 1000

Framework versions

• Transformers 4.45.2
• Pytorch 2.2.1
• Datasets 3.0.1
• Tokenizers 0.20.1