Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
IanCal
on Sept 15, 2023
|
parent
|
context
|
favorite
| on:
Optimizing LLMs from a Dataset Perspective
Not sure on the licensing but yes you can do that technically.
Phi-1 and therefore phi-1.5 are partially trained on gpt3.5 generated synthetic textbooks.
saurik
on Sept 15, 2023
[–]
The premise here is specifically not to train it on generated output of the bigger model but to merely use the bigger model to better curate non-generated (and thereby untainted) inputs for the training set of the smaller model.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
Phi-1 and therefore phi-1.5 are partially trained on gpt3.5 generated synthetic textbooks.