llama-3-8b-instruct-gapo-v2-rougeL-beta2-he-scale-gamma0.3-lr2.0e-6

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the princeton-nlp/llama3-ultrafeedback-armorm dataset. It achieves the following results on the evaluation set:

• Loss: 0.3499
• Rewards/chosen: -11.9406
• Rewards/rejected: -16.3570
• Rewards/accuracies: 0.9004
• Rewards/margins: 4.4164
• Logps/rejected: -8.1785
• Logps/chosen: -5.9703
• Logits/rejected: -1.3845
• Logits/chosen: -1.3878

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: 2e-06
• train_batch_size: 2
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• total_eval_batch_size: 16
• 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	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.3805	0.8550	400	0.3499	-11.9406	-16.3570	0.9004	4.4164	-8.1785	-5.9703	-1.3845	-1.3878

Framework versions

• Transformers 4.44.2
• Pytorch 2.2.0
• Datasets 2.21.0
• Tokenizers 0.19.1