Search results

initial dynamic input support for gpu codegen

Status: Open.

#43 In ML4EP/SOFIE;

· harz05 opened on Jun 22, 2026

Fuse conv and ReLU into the GEMM epilogue (alpaka GPU)

Status: Open.

#36 In ML4EP/SOFIE;

· harz05 opened on Jun 14, 2026

Add GTest for Conv GPU SAME_UPPER autopad

Status: Open.

#33 In ML4EP/SOFIE;

· harz05 opened on Jun 8, 2026

Conv generates wrong code for dilation > 1 (GPU silent-wrong, CPU segfault)

Status: Open.

#32 In ML4EP/SOFIE;

· harz05 opened on Jun 8, 2026

[improvement] Batch the non-grouped Conv GEMM with gemmStridedBatched instead of a per sample loop

Status: Open.

#29 In ML4EP/SOFIE;

· harz05 opened on Jun 3, 2026

MaxPool has no GPU implementation

Status: Open.

#26 In ML4EP/SOFIE;

· harz05 opened on May 29, 2026

Move WeightVecKernel out of infer() into session init

Status: Open.

#25 In ML4EP/SOFIE;

· harz05 opened on May 26, 2026

Incorrect output from grouped Conv GPU path

Status: Open.

#23 In ML4EP/SOFIE;

· harz05 opened on May 25, 2026

Add GTests for Conv GPU inference with groups > 1

Status: Open.

#21 In ML4EP/SOFIE;

· harz05 opened on May 24, 2026

Add GTests for Conv GPU inference with batch size > 1

Status: Open.

#18 In ML4EP/SOFIE;

· harz05 opened on May 12, 2026

ReduceMean: loop variable typo in kFirst reduction path produces wrong results

Status: Open.

#17 In ML4EP/SOFIE;

· harz05 opened on Mar 21, 2026

Add error handling and logging to Python model generator scripts

Status: Open.

#15 In ML4EP/SOFIE;

· harz05 opened on Mar 21, 2026