Repository navigation

Support for dynamic parameteric inputs in inference on heterogeneous architectures - #45

Merged

sanjibansg merged 32 commits into

ML4EP:gpu/alpakafrom

harz05:feat/dynamic-input-alpaka

Sep 15, 2026

Merged

Support for dynamic parameteric inputs in inference on heterogeneous architectures#45
sanjibansg merged 32 commits into
ML4EP:gpu/alpakafrom
harz05:feat/dynamic-input-alpaka

Commits on Sep 15, 2026
- dynamic tensor allocation
  
  harz05
  authored and
  sanjibansg
  committed
  a476d2b
- testing: dynamic support for Tile
  
  harz05
  authored and
  sanjibansg
  committed
  c57dabe
- test added for dynamic input
  
  harz05
  authored and
  sanjibansg
  committed
  671b498
- dynamic shape support for comparision op
  
  harz05
  authored and
  sanjibansg
  committed
  d533224
- onnx models added and regression tests for dynamic Tile and Equal
  
  harz05
  authored and
  sanjibansg
  committed
  aaf491d
- dynamic input for batchnorm op and test cases
  
  harz05
  authored and
  sanjibansg
  committed
  fd97bf2
- dynamic input for conv op and tests
  
  harz05
  authored and
  sanjibansg
  committed
  9b909dd
- emit dynamic session ctor params in declaration order to match infer
  
  harz05
  authored and
  sanjibansg
  committed
  ea99e28
- dynamic input for binary, gather, transpose, concat and slice ops with tests
  
  harz05
  authored and
  sanjibansg
  committed
  065d7c4
- dynamic support for fused kernels and tests for gemm relu fusion and no bias conv
  
  harz05
  authored and
  sanjibansg
  committed
  4d274d7
- dynamic input for reduce op and test cases
  
  harz05
  authored and
  sanjibansg
  committed
  3a1b9af
- add diagnostic logging
  
  harz05
  authored and
  sanjibansg
  committed
  b861d1f
- chore: remove init-fail diagnostic logging
  
  harz05
  authored and
  sanjibansg
  committed
  bcbea58
- feat: dynamic-input GPU codegen for Shape/Range/BasicNary and the particle-net op set
  
  harz05
  authored and
  sanjibansg
  committed
  b36b34e
- test: add dynamic-input GPU operator tests
  
  harz05
  authored and
  sanjibansg
  committed
  35066ad
- chore: in line comments cleanup in the code
  
  harz05
  authored and
  sanjibansg
  committed
  5f7ecaf
- fix: add missing reference headers for softmax and TopK tests
  
  harz05
  authored and
  sanjibansg
  committed
  21f691b
- feat: broadcast support in BasicNary GPU codegen
  
  harz05
  authored and
  sanjibansg
  committed
  fff0914
- refactor: address review comments on naming, launch style and comments
  
  harz05
  authored and
  sanjibansg
  committed
  c1af5ec
- refactor: modifying dynamic input codegen, share shape helpers and fixing codegen bugs
  
  harz05
  authored and
  sanjibansg
  committed
  4a7a8ff
- fix: guard dynamic sizes at infer and finish the run-time Range path on GPU
  
  harz05
  authored and
  sanjibansg
  committed
  7197e6b
- style: drop alpaka::math from exp and log in the Softmax kernel
  
  harz05
  authored and
  sanjibansg
  committed
  7b224bf
- refactor: pass dyn params to the gpu emitters and rename the Dim shape helpers
  
  harz05
  authored and
  sanjibansg
  committed
  d4f2108
- feat: pass the operation epilogue when registering blas configs
  
  harz05
  authored and
  sanjibansg
  committed
  b19fafb
- feat: special case of reduction operator for the last and first axis
  sanjibansg
  committed
  fbf986b
- feat: dynamic configuration of resource allocation for reduce operator
  sanjibansg
  committed
  6ba406d
- feat: avoid D-H movements by using session initialized values in range operator
  sanjibansg
  committed
  c02389a
- feat: grid-strides to better parallelize the layernorm kernel, remove math define directives
  sanjibansg
  committed
  a971390
- feat: use shared memory and registers for hierarchial search of top-K
  sanjibansg
  committed
  939acf5
- feat: support for inference on dynmainc tensors for GatherND, ScatterND, Split, Trilu, Expand and Pool
  sanjibansg
  committed
  bdecc24
- feat: minor improvments in naming conventions, and redundant code removal
  sanjibansg
  committed
  788c54f
- feat: rename lengthExpr for RModel to iterationLengthExpr to imply iteration space range
  sanjibansg
  committed
  1f8472c

Commits

Commits on Sep 15, 2026