phi-2-finetuned-1k_stories_100_genre

This model is a fine-tuned version of microsoft/phi-2 on 1k_stories_100_genre dataset using TRL library's SFTTrainer class.

Model description

Use prompts in the format : f"Generate a '{genre_of_the_story} ' story titled '{title of the story}'."

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Supervised Fine-Tuning using QLoRA. Used bitsandbytes for 4-bit quantization and LoRA adapters with r=4 for only q, v projection matrices.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Framework versions

• PEFT 0.10.0
• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2