llama3-2k-rm-ref-spa_3

This model is a fine-tuned version of save_model/llama3-2k-rm-ref-spa_2 on the datasets/training-llama3-2k-rm-ref-spa_3 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7454
• Rewards/chosen: -0.8829
• Rewards/rejected: -0.9416
• Rewards/accuracies: 0.5468
• Rewards/margins: 0.0588
• Rewards/mix Margin: 0.0588
• Logps/rejected: -295.0553
• Logps/chosen: -336.0750
• Logits/rejected: -0.4778
• Logits/chosen: -0.4730

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: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• total_eval_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Rewards/mix Margin	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6863	0.53	1000	0.7454	-0.8829	-0.9416	0.5468	0.0588	0.0588	-295.0553	-336.0750	-0.4778	-0.4730

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.15.2