[tmva][sofie] Add support for optimal memory allocation of dynamic tensors #20434

Open

lmoneta wants to merge 10 commits intoroot-project:masterfrom

lmoneta:tmva_sofie_dynamic_tensor

Commits on Jan 6, 2026

[tmva][sofie] Apply fixes for supporting Dynamic tensors
lmoneta
committed
[tmva][sofie] Remove special case handling bool outputs
lmoneta
committed
[tmva][sofie] Add support for greedy memory allocation for dynammic tensors
lmoneta
committed
[tmva][sofie] Fix an issue in genereting code for dynamic tensor when broadcasting
lmoneta
committed
[tmva][sofie] Fix stacked MatMul and speedup LayerNorm
lmoneta
committed
[tmva][sofie] Fix issue with order execution of tensors in GNN models
lmoneta
committed
[tmva][sofie] Apply fixes for the TestCustiomModelsFrom ROOT
lmoneta
committed

Commits on Jan 7, 2026

[tmva][sofie] Sort in alphabetical order the shape parameter in Session ctor
lmoneta
committed

Commits on Jan 8, 2026

[tmva][sofie] Do not permor broadcast of bias tensor in Gemm in Session ctor
lmoneta
committed

Commits on Jan 9, 2026

[tmva][sofie] Add alias tensors
lmoneta
committed