llama3-2k-non-spa_2

This model is a fine-tuned version of save_model/llama3-2k-non-spa_1 on the datasets/training-llama3-2k-non-spa_2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6049
• Rewards/chosen: -2.6010
• Rewards/rejected: -4.3580
• Rewards/accuracies: 0.7429
• Rewards/margins: 1.7570
• Rewards/mix Margin: 1.7570
• Logps/rejected: -322.7352
• Logps/chosen: -348.4507
• Logits/rejected: -0.3859
• Logits/chosen: -0.4144

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
0.3094	0.8	1000	0.6049	-2.6010	-4.3580	0.7429	1.7570	1.7570	-322.7352	-348.4507	-0.3859	-0.4144

Framework versions

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