{"id":1854,"date":"2019-04-15T19:36:06","date_gmt":"2019-04-16T03:36:06","guid":{"rendered":"https:\/\/live-cometml.pantheonsite.io\/blog\/approach-pre-trained-deep-learning-models-with-caution\/"},"modified":"2019-04-15T19:36:06","modified_gmt":"2019-04-16T03:36:06","slug":"approach-pre-trained-deep-learning-models-with-caution","status":"publish","type":"post","link":"https:\/\/www.comet.com\/site\/blog\/approach-pre-trained-deep-learning-models-with-caution\/","title":{"rendered":"Approach pre-trained deep learning models with caution"},"content":{"rendered":"\n

 <\/p>\n\n\n\n

Pre-trained models are easy to use, but are you glossing over details that could impact your model performance?<\/h4>\n\n\n\n

How many times have you run the following snippets:<\/p>\n\n\n\n

import torchvision.models as models\ninception = models.inception_v3(pretrained=True)<\/code><\/pre>\n\n\n\n
or<\/p>\n\n\n\n
from keras.applications.inception_v3 import InceptionV3\nbase_model = InceptionV3(weights='imagenet', include_top=False)<\/code><\/pre>\n\n\n\n
It seems like using these pre-trained models have become a new standard for industry best practices. After all, why\u00a0wouldn\u2019t<\/em>\u00a0you take advantage of a model that\u2019s been trained on more data and compute than you could ever muster by yourself?<\/p>\n\n\n\n
\n
See the discussion on\u00a0<\/em>Reddit<\/a>\u00a0and\u00a0<\/em>HackerNews<\/a><\/p>\n<\/blockquote>\n\n\n\n
Long live pre-trained models!<\/h2>\n\n\n\n
There are several substantial benefits to leveraging pre-trained models:<\/p>\n\n\n\n