GPT2-Pretrain, Inference

• complete pretraining code
• obtained pretrained model
• load model from huggingface
• inference through huggingface space
• (huggingface) added tokenizer along with model code
• code cleanup.

GPT2: sebastian-gutenberg (500Mb chunk)

• GPT2-sebastian-gutenberg-project-code
• works on 500 MB single chunk
• original source code
• Modified working code

# for multiple chunks, we should modify the code from 
for epoch in epochs:
    for chunk in chunks

# instead of 
for chunk in chunks:
    for epoch in epochs:

# will do in future