vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-06 11:19:03 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Update README.md

Browse source
This commit is contained in:
Li, Zonghang 2025-04-11 01:41:43 +08:00 • committed by GitHub
parent 4845abf25e
commit 6d13836c44
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
README.md
View file

@ -59,6 +59,8 @@ And, if your devices are more powerful, you could unlock even more possibilities
| DeepSeek-R1-Distill-Llama-70B | 10978 ms | OOM | - | **724 ms** |
| Qwen-2.5-72B | 12227 ms | OOM | - | **867 ms** |
> As video recording consumes some RAM, prima.cpp proactively reduces memory usage, resulting in slightly higher latency in the video compared to the table.
> In current implementation, each device is assigned at least one model layer. For example, this leads to a 1:1:29:1 split for Llama 3-8B, which makes prima.cpp less efficient. In future updates, we will have a 0:0:32:0 split and idle devices removed, then llama.cpp would become a special case of prima.cpp when serving small models.
## 🔑 Key Features