Models are loaded and run through Hugging Face Transformers pipelines.
Small language models
Using small-parameter, freely available language models on minimal hardware.
Project requirement
No hardware for running large models.
Technical question
How far can small models go in document generation?
Validation
DistilGPT2, GPT-Neo 125M, Facebook OPT 125M and 350M, and GPT-2 Medium were compared.
Decision
Small models are usable with tailored prompting; handling complex relationships is still limited and needs fine-tuning.