week5-phi-1.5-pref-shakespeare

Model Card for Model ID

Summary

<!-- Provide a quick summary of what the model is/does. -->

This is a preference tuned model for text completion based on Phi 1.5. It has been tuned on a filtered version of the The Complete Works of William Shakespeare, which can be found and downloaded from here: https://www.gutenberg.org/ebooks/100.

Model Description

<!-- Provide a longer summary of what this model is. -->

• Developed by: Course Organizers
• Finetuned from model: microsoft/phi-1_5

Training Details

This model has been trained using the TRL library and OPROTrainer class from Huggingface.

Training Data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

The Complete Works of William Shakespeare, which can be found and downloaded from here: https://www.gutenberg.org/ebooks/100

Training Hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-06
• per_device_train_batch_size: 1
• lr_scheduler_type: cosine
• weight_decay: 0.01
• num_epochs: 1

Framework Versions

• accelerate==0.26.1
• datasets==2.16.1
• transformers==4.45.2
• trl==0.11.2

Compute Infrastructure and Hardware

Slurm cluster with 8 x H100 Nvidia GPUs.