tinyllama-1.1B-intermediate-step-715k-1.5T-dpo-lora

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6877
• Rewards/chosen: 0.0254
• Rewards/rejected: 0.0135
• Rewards/accuracies: 0.5645
• Rewards/margins: 0.0119
• Logps/rejected: -315.6106
• Logps/chosen: -402.4017
• Logits/rejected: -4.0818
• Logits/chosen: -4.1916

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: 2
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 32
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6916	1.0	968	0.6921	0.0039	0.0011	0.5070	0.0028	-315.7343	-402.6164	-4.0813	-4.1913
0.6904	2.0	1936	0.6884	0.0191	0.0086	0.5570	0.0105	-315.6588	-402.4643	-4.0824	-4.1920
0.6876	3.0	2904	0.6877	0.0254	0.0135	0.5645	0.0119	-315.6106	-402.4017	-4.0818	-4.1916

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1