charvibannur/OpenAssistant-Phi2-QLoRA

out

This model is a fine-tuned version of microsoft/phi-2 on the open-assistant-guanaco dataset. It achieves the following results on the evaluation set:

• Loss: 0.2759

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 1
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.3107	0.5	276	0.3292
0.2879	1.0	552	0.3003
0.3	1.5	828	0.2873
0.3341	1.99	1104	0.2759

Framework versions

• PEFT 0.7.2.dev0
• Transformers 4.36.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0