transposeGPU function
Implementation
GPUTensor<Matrix> transposeGPU(GPUTensor<Matrix> a, CommandBuffer tape) {
int M = a.shape[0];
int N = a.shape[1];
GPUTensor<Matrix> out = GPUTensor<Matrix>.empty(<int>[N, M]);
// tmpGrad must match original input shape [M, N]
GPUTensor<Matrix> tmpGrad = GPUTensor<Matrix>.empty(<int>[M, N]);
tape.putInt(OP_TRANSPOSE);
tape.putString(a.id);
tape.putString(out.id);
out.creator = GPUNode(
[a],
(CommandBuffer bTape) {
bTape.putInt(OP_TRANSPOSE);
bTape.putString('${out.id}_grad');
bTape.putString(tmpGrad.id);
bTape.putInt(OP_ADD_INTO);
bTape.putString(tmpGrad.id);
bTape.putString('${a.id}_grad');
},
opName: 'transposeGPU',
extraParams: {'tmpGrad': tmpGrad},
cost: 0,
);
return out;
}