Add the ability to load Pauli strings#3
Conversation
|
There are now three ways to transfer a Hamiltonian from
|
572b65a to
6ce5b80
Compare
6ce5b80 to
d243d71
Compare
|
I ran another comparison (and by "I", I mean "Claude Code"), and got the below results. I see that the Pauli strings formats still agree with each other, but the tensors format still gives a slightly smaller resource estimate. Generated an H2 molecule Hamiltonian using
|
| Format | Clifford Count | T Count | Qubit Count |
|---|---|---|---|
| Tensors (.tensors.npz) | 59,071,402 | 30,753,652 | 15 |
| Dense Pauli (.dat) | 59,914,766 | 31,273,590 | 15 |
| Sparse Pauli (.dat) | 59,914,766 | 31,273,590 | 15 |
Key Findings
-
Dense and Sparse Pauli formats are identical - They produce exactly the same resource estimates, as expected since they represent the same Hamiltonian just in different notations.
-
Tensors format differs slightly - About 1.4-1.7% fewer gates, likely due to the internal fermion-to-qubit mapping process having minor numerical differences.
-
All formats are viable - The small differences (<2%) show that all three formats work correctly and produce very similar resource estimates for the same physical system.
If we're going to claim it in the paper, we should really add it. And it should be easy to quickly add at least a very basic version of this.
Reuben is working on grouping, and he may end up modifying this and/or creating a second reader that does something similar but supports grouping.