Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
hsuduebc2
11 months ago
|
parent
|
context
|
favorite
| on:
DeepSeek releases Janus Pro, a text-to-image gener...
I read that they training data was not scrapped from the internet but generated through meta llama model. Maybe this is the reason behind such behavior?
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: