wassname/grpo_proj2
1
0
Fork 0
mirror of https://github.com/wassname/grpo_proj2.git synced 2026-06-27 15:00:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • 74e228529b feat: finish the self-contained substrate -- 6 -> 24 problems (6 per mode) fix/bf16-gpu-path wassname 2026-06-01 06:30:20 +00:00
  • 409d9c9425 refactor: SVD-diag knob -> parametrized LoRA (fixed A, train B) wassname 2026-06-01 03:43:18 +00:00
  • fbacefd433 fix: bf16/cuda path (7 device+dtype bugs) + GPU-bf16 smoke wassname 2026-06-01 03:21:08 +00:00
  • 25ad306763 readme main wassname 2026-06-01 00:19:30 +00:00
  • b0d1bcd3d5 Rebuild src/ from pseudocode: SVD-basis gradient projection vs GRPO reward hacking wassname 2026-05-31 13:58:54 +00:00