Qwen2.5-7B-Instruct SFT on CodeForces-CoT
Created on March 1|Last edited on March 3
Comment
Section 1
Initial scan with:
- learning rate in range [1e-5, 2e-5, 4e-5]
- packing=false
- effective bs = 128
- num epochs = 10 (checkpoint every 20% steps)
25
6
4
3
3
7
openthoughts-solutions-w-editorials-mix
6
Add a comment