Skip to content

feat: QONNX parsing and CPU/GPU lowering#44

Open
shaunlee8 wants to merge 6 commits into
gpu/alpakafrom
quantize
Open

feat: QONNX parsing and CPU/GPU lowering#44
shaunlee8 wants to merge 6 commits into
gpu/alpakafrom
quantize

Conversation

@shaunlee8

@shaunlee8 shaunlee8 commented Jun 30, 2026

Copy link
Copy Markdown
Collaborator

Initial quantization integration to SOFIE. Detects a quantized GEMM region by inspecting local QONNX nodes after the base graph is built. Made semi-independent through hooks into base SOFIE files (moderate changes to core/src/RModel_ALPAKA.cxx).

  • QONNX parsing and simple CPU lowering path for quantized GEMM added.
  • Alpaka/cuBLASLt-optimized int8 quantized GEMM path added.
  • ctest for QONNX to quantized GEMM Alpaka CUDA codegen, direct int8.

@shaunlee8 shaunlee8 changed the title feat: QONNX parsing and basic lowering feat: QONNX parsing and CPU/GPU lowering Jul 2, 2026
@shaunlee8 shaunlee8 marked this pull request as ready for review July 6, 2026 08:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant