To avoid GitHub's 100MB file limit, the fine-tuned model is hosted on HuggingFace Hub:

Model Used by the API

dlwts/gpt2_qa_finetuned https://huggingface.co/dlwts/gpt2_qa_finetuned

The API automatically downloads this model at runtime:

tokenizer_llm = AutoTokenizer.from_pretrained("dlwts/gpt2_qa_finetuned") model_llm = AutoModelForCausalLM.from_pretrained("dlwts/gpt2_qa_finetuned")