[Abhishek-B-R]

orchestrate.py plan never shows the kernel shape. extract.py writes it under model_shape, but cmd_plan reads shape, which nothing writes, so the column is always empty.

Same command also prints each what-if line twice when the plan has 5 or fewer kernels, since min(len(kernels), 10) lands on 1, 3 or 5 again.

Repro: build a plan with extract.generate_optimization_plan for 3 kernels and run cmd_plan on it.

before
  1     matmul                                  12.00   60.0   OPTIMIZING
  If top-3 kernels achieve 1.5x -> model speedup: 1.43x
  If top-3 kernels achieve 1.5x -> model speedup: 1.43x

after
  1     matmul   M=2048, N=4096, K=4096         12.00   60.0   OPTIMIZING
  If top-3 kernels achieve 1.5x -> model speedup: 1.43x

Two line change in cmd_plan, still falls back to shape for hand written plans. Doesn't touch the time budget code that #13 changes.

[Abhishek-B-R]

@jaberjaber23 friendly bump: this makes orchestrate.py plan read the model_shape key that extract.py writes (falling back to shape for hand-written plans) and stops the what-if rows repeating for plans with 5 or fewer kernels. It merges cleanly with main and doesn't overlap the time-budget change in #13. Happy to adjust anything if the approach doesn't fit.

[Abhishek-B-R]

@jaberjaber23 following up once more: the branch still merges cleanly with main, and it's a single commit so orchestrate.py plan reads the model_shape key that extract.py writes (falling back to shape for hand-written plans) and no longer repeats what-if rows for plans with 5 or fewer kernels. If you'd prefer a different approach, I'm happy to rework it.