How to speed up pred_prob
Open
performance
- Dominant language
- C++
- Stars
- 9.1k
- Forks
- 1.3k
- PR merge metrics
- No merged PRs in 30d
Description
I need to predict scores (labels) on 800M unlabeled samples.
the running time on colab default setting is as following
```
# 300,000 : 649.1985890865326
# 200,000 : 428.93789553642273
# 100,000 : 221.7507927417755
# 10,000 : 19.62497639656067
# 1,000 : 1.960167646408081
```
with this calculation 800M would take 20 days
Contributor guide
Assessment
This issue has not been assessed yet.