Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
boltzmann-brain
2 days ago
|
parent
|
context
|
favorite
| on:
Speculative Speculative Decoding (SSD)
> Our implementation is up to 2x faster than optimized speculative decoding baselines and up to 5x faster than autoregressive decoding with open source inference engines
what about per-FLOP?
help
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
what about per-FLOP?