gpt-peter-2.7B

pszemraj/gpt-peter-2.7B

• This model is a fine-tuned version of EleutherAI/gpt-neo-2.7B on about 80k WhatsApp and iMessage texts.
• The model is too large to use the inference API. linked here is a notebook for testing in Colab.
  • alternatively, you can message a bot on telegram where I test LLMs for dialogue generation
  • the telegram bot code and the model training code can be found in this repository

Usage in python

Install the transformers library if you don't have it:

pip install -U transformers

load the model into a pipeline object:

from transformers import pipeline
import torch
my_chatbot = pipeline('text-generation', 
                      'pszemraj/gpt-peter-2.7B',
                      device=0 if torch.cuda.is_available() else -1,
                    )

generate text!

my_chatbot('Did you ever hear the tragedy of Darth Plagueis The Wise?')

(example above for simplicity, but adding generation parameters such as no_repeat_ngram_size are recommended to get better generations)

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 6e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 32
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 1

Framework versions

• Transformers 4.17.0
• Pytorch 1.10.0+cu113
• Datasets 2.0.0
• Tokenizers 0.11.6