Class EEmbTransferGradOp

en_ops/e_emb_transfer.cc:85–125

REGISTER_KERNEL_BUILDER(Name("EEmbTransfer").Device(DEVICE_CPU), EEmbTransferOp);

class EEmbTransferGradOp : public OpKernel
{
public:
    explicit EEmbTransferGradOp(OpKernelConstruction *context) : OpKernel(context)
    {
        OP_REQUIRES_OK(context, context->GetAttr("eid_low", &eid_low_));
        OP_REQUIRES_OK(context, context->GetAttr("eid_high", &eid_high_));
        lib = dlopen("/home/zhangyan/jhrsgx/privacy_test/privacy_tf/sgx_tf_ops/sgx.so", RTLD_LAZY);
        OP_REQUIRES(context, lib != NULL, errors::Unknown("Unable to load sgx.so!"));
    }
    void Compute(OpKernelContext *context) override
    {

        const Tensor &grad = context->input(0);
        auto grad_flat = grad.flat<float>();
        const TensorShape &input_shape = grad.shape();

        TensorShape output_shape;
        output_shape.AddDim(input_shape.dim_size(0) / 10);
        output_shape.AddDim(input_shape.dim_size(1) + 1);
        Tensor *output = NULL;
        OP_REQUIRES_OK(context, context->allocate_output(0, output_shape, &output));
        auto output_flat = output->flat<float>();

        int M = input_shape.dim_size(0) / 10;
        int C = input_shape.dim_size(1);

        unsigned long int eid_ = (eid_high_ << 32) + eid_low_;
        typedef void (*function)(unsigned long int eid, float *grad, float *output, int M, int C);
        dlerror();
        function emb_transfer_grad_kernel = (function)dlsym(lib, "emb_transfer_grad");
        const char *dlsym_error = dlerror();
        OP_REQUIRES(context, !dlsym_error, errors::Unknown("loading of relu_grad failed: ", dlsym_error));
        emb_transfer_grad_kernel(eid_, (float *)grad_flat.data(), (float *)output_flat.data(), M, C);
    };

private:
    void *lib;
    int64 eid_low_;
    int64 eid_high_;
};
REGISTER_KERNEL_BUILDER(Name("EEmbTransferGrad").Device(DEVICE_CPU), EEmbTransferGradOp);