# How to reduce the size of a machine learning model?

I am a beginner in this field. I needed a summariser model to summarise the text content on my server. I found a few good models on Hugging Face, the models work pretty well.

However the models are too big, about 1 to 2 giga bytes in size and I am running a pretty small server for a hooby project.

Is there a way to reduce the size of the model or find a compressed version of it ? Do I need to look for a new model?

Currently I am using this model.

• I have no experience with summarizer models but the task is one of the hardest and being also NLP usually the models are really large. I don't expect to find small, even medium models for this task. Sep 16 at 7:05
• In addition to what @rapaio mentioned, have you looked at some of the other models that mentioned on that page? For example the distillbart-xsum-12-1 model seems to be smaller (less parameters and much faster inference speed). Sep 16 at 7:11