Live data from Hacker News

Show HN: Overfitted a 900KB Transformer to Compress a 100MB CSV into 7MB

news.ycombinator.com

81–83 of 83 posts

Re: Show HN: Overfitted a 900KB Transformer to Compress a 100MB CSV into 7MB

#81
so trained model only works for that specific file, and retraining is required for a new file? maybe this is could be good for crypto instead of compression, the question is how do you share the model before hand to allow destination to decrypt it, if you use conventional stuff like rsa, then the question would be why don't you use for the actual message directly.

Re: Show HN: Overfitted a 900KB Transformer to Compress a 100MB CSV into 7MB

#82

so trained model only works for that specific file, and retraining is required for a new file? maybe this is could be good for crypto instead of compression, the question is how do you share the model before hand to allow destination to decrypt it, if you use conventional stuff like rsa, then the question would be why don't you use for the actual message directly.

Ummm I don't have a lot of idea about crypto but in compression the idea was that you just share the model along with the metadata file to the other user and they de compress it, not sure how things will differ in crypto
Post reply on HN