GPT2-Codeparrot

Generative Pre-trained Transformer 2 (GPT-2) is a large language model from OpenAI that was first introduced in gpt2. It is a decoder-only Transformer model trained using a masked language modeling (MLM) objective. This means the model is trained to predict the next word in a sequence, given the previous words. GPT-2 models are known for their ability to generate realistic and coherent text, making them useful for a variety of natural language processing tasks such as text generation, translation, and question answering.

Model description

This model is a base GPT-2 architecture with [insert number] parameters. It was trained on the huggingface-course/codeparrot-ds-valid dataset, which is a small subset of the original WebText dataset used to train GPT-2. Due to the limited training data, this model may not perform as well as other pre-trained GPT-2 models available on Hugging Face.

Intended uses & limitations

This model is intended for personal learning and exploration of the GPT-2 architecture. Due to its limited training data, it may not be suitable for real-world applications.

Training and evaluation data

This model was trained using the Transformers library with the following specifications:

• Training Data: huggingface-course/codeparrot-ds-valid
• Training Script: Training_a_causal_language_model_from_scratch

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0005
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 1000
• num_epochs: 1
• mixed_precision_training: Native AMP

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1