// What we want to calculate is: R = matmul(P-z, Q-w).
// Then r_ij = sigma(k) (p_ik - z) * (q_kj - w)
//           = sigma(k)(p_ik * q_kj) - w * sigma(k)p_ik - z * sigma(k)q_kj
//             + sigma(k)z*w.
// zp_input_contribution = z * sigma(k)q_kj
// zp_weight_contribution = w * sigma(k)p_ik
// In case z != 0 and w != 0, we need to additionally calculate sigma(k)z*w,
// which is: # of reduced dim(n in this case) * input_zp * weight_zp

        entities.put("Nu", "\u039d");
        entities.put("Xi", "\u039e");
        entities.put("Omicron", "\u039f");
        entities.put("Pi", "\u03a0");
        entities.put("Rho", "\u03a1");
        entities.put("Sigma", "\u03a3");
        entities.put("Tau", "\u03a4");
        entities.put("Upsilon", "\u03a5");
        entities.put("Phi", "\u03a6");
        entities.put("Chi", "\u03a7");
        entities.put("Psi", "\u03a8");

	"Kappa":    "\u039A",
	"Lambda":   "\u039B",
	"Mu":       "\u039C",
	"Nu":       "\u039D",
	"Xi":       "\u039E",
	"Omicron":  "\u039F",
	"Pi":       "\u03A0",
	"Rho":      "\u03A1",
	"Sigma":    "\u03A3",
	"Tau":      "\u03A4",
	"Upsilon":  "\u03A5",
	"Phi":      "\u03A6",
	"Chi":      "\u03A7",
	"Psi":      "\u03A8",
	"Omega":    "\u03A9",
	"alpha":    "\u03B1",
	"beta":     "\u03B2",

// Analysis series B. Vol. 1:
// lgamma(z + 1) = (log(2) + log(pi)) / 2 + (z + 1/2) * log(t(z)) - t(z) + A(z)
// t(z) = z + kLanczosGamma + 1/2
// A(z) = kBaseLanczosCoeff
//       + sigma(k = 1, n, kLanczosCoefficients[i] / (z +  k))
//
// Coefficients for the Lanczos approximation of the gamma function. The
// coefficients are uniquely determined by the choice of g and n

	{as: AVSBOX, a1: C_VREG, a6: C_VREG, type_: 82, size: 4},              /* vector AES subbytes, vx-form */

	/* Vector SHA */
	{as: AVSHASIGMA, a1: C_U16CON, a2: C_VREG, a3: C_U16CON, a6: C_VREG, type_: 82, size: 4}, /* vector SHA sigma, vx-form */

	/* VSX vector load */
	{as: ALXVD2X, a1: C_XOREG, a6: C_VSREG, type_: 87, size: 4},        /* vsx vector load, xx1-form */

boxes based on score.}]>:$score_threshold,
    Arg<TensorOf<[TF_Float16, TF_Float32]>, [{A 0-D float tensor representing the sigma parameter for Soft NMS; see Bodla et
al (c.f. https://arxiv.org/abs/1704.04503).  When `soft_nms_sigma=0.0` (which
is default), we fall back to standard (hard) NMS.}]>:$soft_nms_sigma,