pankaja0285/era_v1_session17

BERT,GPT, ViT model arch

era_v1_session17

Description: Using Transformer based architecture train BERT, GPT and ViT models.

Problem statement:
- Re-write the code in such a way where there is one transformer.py file that you can use to train all 3 model
- Share the last few log details for all 3 models (BERT, GPT and ViT)
- Share the link to a single repo where all 3 ipynb files can be found

Folder structure

|_ bert_data
  |__ names.tsv
  |__ training.txt
  |__ values.tsv
  |__ vocab.txt
|_ gpt_data
  |__ english.txt
|_ vit_data
  |__ test
  |__ train
|_ checkpoint
  |__ README.md
|_ models
  |__ README.md
|_ transformer_model_scripts
  |__ vit_utils
  |__ BERT_utils.py
  |__ GPT_utils.py
  |__ transformer_models.py
|_ README.md
|_ S17_Bert.ipynb
|_ S17_Gpt.ipynb
|_ S17_Vit.ipynb

Details of how to run:

- clone the repo with

    
        git clone [email protected]:pankaja0285/era_v1_session17.git
    

- FOR BERT Model: click to open the S17_Bert.ipynb and proceed to run cell by cell
- FOR GPT Model: click to open the S17_Gpt.ipynb and proceed to run cell by cell
- FOR ViT Model: click to open the S17_Vit.ipynb and proceed to run cell by cell