Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
SlavikCA
16 days ago
|
parent
|
context
|
favorite
| on:
GLM-5.2 – How to Run Locally
And with MTP (or other speculation techniques) you can ~double that.
phamilton
16 days ago
[–]
MTP on a MoE is hit or miss. If you're bottlenecked on memory, MTP can increase the number of active experts (like any batch processing would), which can eat away gains from it.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: