e6ce836b2925a677d0924e8859439937ba5808dc,gpflow/conditionals/mo_conditionals.py,,_conditional,#Any#Any#Any#Any#,73

Before Change

    // rmu, rvar = tf.map_fn(lambda x: single_gp_conditional(x[0], x[1], x[2], x[3], x[4]),
    //                       (Kmms, Kmns, Knns,fs, q_sqrts), dtype=dtypes)
    for i in range(Knns.shape[0]):
        single_gp_conditional(Kmms[i, :, :], Kmns[i, :, :], Knns[i, :], fs[i, :, :], q_sqrts[i,
                                                                                      :, :])
    // [P,
    // N, 1], [P, 1, N, N] or [P, N, 1]
    fmu = tf.linalg.transpose(rmu[:, :, 0])  // [N, P]

After Change

        return base_conditional(Kmn, Kmm, Knn, f, full_cov=full_cov, q_sqrt=q_sqrt, white=white)

    // [P, N, 1], [P, 1, N, N] or [P, N, 1]
    rmu, rvar = tf.map_fn(single_gp_conditional,
                          (Kmms, Kmns, Knns, fs, q_sqrts),
                          (default_float(), default_float()))
    fmu = tf.linalg.transpose(rmu[:, :, 0])  // [N, P]

    if full_cov:
        fvar = rvar[:, 0, :, :]  // [P, N, N]

Instances