Hacker Times
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
minimaxir
35 days ago
|
parent
|
context
|
favorite
| on:
Making LLM Training Faster with Unsloth and NVIDIA
You can use modern off-the-shelf models for those types of tasks, however a smaller-but-bespoke model will usually be more cost-efficient if used at scale.
najarvg
35 days ago
[–]
And smaller bespoke models running locally are better for regulated workflows (healthcare, banking etc) as well
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: