v0.4.0 release

This is the release note of v0.4.0.

Highlights

• Proper ONNX export, and optimizing peak memory usage and execution speed on some ONNX inference engine
  • Split c6 computation to reduce peak memory usage #17
  • Fix tracing batch size #18
  • Reduce Gather op when ONNX-exported #20

Others

• Update CI #19

v0.3.0 release

This is the release note of v0.3.0.

Highlights

• use shift for gradient calculation instead of cell #13 (Thank you @So-Takamoto )
  • It includes 1. speed up of stress calculation for batch atoms, and 2. bug fix for stress calculation when cell is skewed.

v0.2.0 release

This is the release note of v0.2.0.

Highlights

• Add PFP citation in README.md #2
• Use pymatgen for pbc neighbor search speed up #3

Bug fixes

• Fixed a bug for inputs with 0 adjacencies. #6 (Thank you @AkihideHayashi )
• Remove RuntimeError on no-cupy environment #8 (Thank you @So-Takamoto )
• Bugfix: batch calculation with abc=True #9 (Thank you @So-Takamoto )

Others

• move pysen to develop dependency #10 (Thank you @So-Takamoto )

v0.1.0 release

First public release of torch-dftd!