You can download different types of file (clean and malicious) from a large list of organizations and educational institutions, such as:
ViruSign: http://www.virusign.com/
MalShare: http://malshare.com/
Malware DB: http://ytisf.github.io/theZoo/ Endgame
Malware BEnchmark for Research (EMBER): One of the largest datasets, this contains 1.1 million SHA256 hashes from PE files that were scanned sometime in 2017.
I highly recommend you download it and try to build your models using it. You can download it from https://pubdata.endgame.com/ember/ember_dataset.tar.bz2 (1.6 GB, expands to 9.2 GB):