[ExecuTorch][WebGPU] Add reduce op (sum/mean) to the WebGPU backend#21076

Merged

JCNTH merged 25 commits into

gh/JCNTH/61/origfrom

gh/JCNTH/62/orig

Jul 21, 2026

Commits on Jul 21, 2026

[ExecuTorch][WebGPU] Add reduce op (sum/mean) to the WebGPU backend
JCNTH
committed
[ExecuTorch][WebGPU] Tests for the reduce WebGPU ops
JCNTH
committed
[ExecuTorch][WebGPU] Add select/boolean ops (where, scalar compares, logical_not) to the WebGPU backend
JCNTH
committed
[ExecuTorch][WebGPU] Tests for the select_bool WebGPU ops
JCNTH
committed
[ExecuTorch][WebGPU] Add gather and embedding ops to the WebGPU backend
JCNTH
committed
[ExecuTorch][WebGPU] Tests for the gather WebGPU ops
JCNTH
committed
[ExecuTorch][WebGPU] Add shape/copy ops (dim_order, expand_copy, fill) to the WebGPU backend
JCNTH
committed
[ExecuTorch][WebGPU] Tests for the shape_copy WebGPU ops
JCNTH
committed
[ExecuTorch][WebGPU] Add q4gsw training kernels (backward, weight-grad, requant) to the WebGPU backend
JCNTH
committed
[ExecuTorch][WebGPU] Tests for the q4gsw_train WebGPU ops
JCNTH
committed
[ExecuTorch][WebGPU] Add et_vk.fused_ce (fused cross-entropy) to the WebGPU backend
JCNTH
committed
[ExecuTorch][WebGPU] Tests for the fused_ce WebGPU ops
JCNTH
committed
[ExecuTorch][WebGPU] Add et_vk.adamw_step (on-GPU AdamW optimizer step) to the WebGPU backend
JCNTH
committed
[ExecuTorch][WebGPU] Tests for the adamw WebGPU ops
JCNTH
committed
[ExecuTorch][Vulkan] Add et_vk.adamw_step kernel (on-GPU AdamW optimizer step)
JCNTH
committed
[ExecuTorch][Vulkan] Tests for et_vk.adamw_step
JCNTH
committed
[ExecuTorch][Vulkan] Add et_vk.linear_dW kernel (dense fp32 weight-grad GEMM)
JCNTH
committed
[ExecuTorch][Vulkan] Tests for et_vk.linear_dW
JCNTH
committed
[ExecuTorch][Vulkan] Add et_vk.fused_ce kernels (fused cross-entropy loss + dlogits)
JCNTH
committed
[ExecuTorch][Vulkan] Tests for et_vk.fused_ce
JCNTH
committed
[ExecuTorch][Vulkan] Add et_vk.linear_q4gsw_backward kernel (4-bit input-grad)
JCNTH
committed
[ExecuTorch][Vulkan] Tests for et_vk.linear_q4gsw_backward
JCNTH
committed
[ExecuTorch][Vulkan] Add et_vk.q4gsw_requant kernel (STE re-quant to W_4X8)
JCNTH
committed
[ExecuTorch][Vulkan] Tests for et_vk.q4gsw_requant
JCNTH
committed
[ExecuTorch][WebGPU] Glob runtime/ops sources in CMakeLists
JCNTH
committed