Method pack

models/quantization.py:315–363

(self, linear, scales, zeros, g_idx=None)

            self.bias = None

    def pack(self, linear, scales, zeros, g_idx=None):
        self.g_idx = g_idx.clone() if g_idx is not None else self.g_idx

        scales = scales.t().contiguous()
        zeros = zeros.t().contiguous()
        scale_zeros = zeros * scales
        self.scales = scales.clone().half()
        if linear.bias is not None:
            self.bias = linear.bias.clone().half()

        intweight = []
        for idx in range(self.infeatures):
            intweight.append(torch.round(
                (linear.weight.data[:, idx] + scale_zeros[self.g_idx[idx]]) / self.scales[self.g_idx[idx]]).to(
                torch.int)[:, None])
        intweight = torch.cat(intweight, dim=1)
        intweight = intweight.t().contiguous()
        intweight = intweight.numpy().astype(np.uint32)
        qweight = np.zeros((intweight.shape[0] // 32 * self.bits, intweight.shape[1]), dtype=np.uint32)
        i = 0
        row = 0
        while row < qweight.shape[0]:
            if self.bits in [2, 4, 8]:
                for j in range(i, i + (32 // self.bits)):
                    qweight[row] |= intweight[j] << (self.bits * (j - i))
                i += 32 // self.bits
                row += 1
            else:
                raise NotImplementedError("Only 2,4,8 bits are supported.")

        qweight = qweight.astype(np.int32)
        self.qweight = torch.from_numpy(qweight)

        zeros -= 1
        zeros = zeros.numpy().astype(np.uint32)
        qzeros = np.zeros((zeros.shape[0], zeros.shape[1] // 32 * self.bits), dtype=np.uint32)
        i = 0
        col = 0
        while col < qzeros.shape[1]:
            if self.bits in [2, 4, 8]:
                for j in range(i, i + (32 // self.bits)):
                    qzeros[:, col] |= zeros[:, j] << (self.bits * (j - i))
                i += 32 // self.bits
                col += 1
            else:
                raise NotImplementedError("Only 2,4,8 bits are supported.")

        qzeros = qzeros.astype(np.int32)
        self.qzeros = torch.from_numpy(qzeros)

    def forward(self, x):
        out_shape = x.shape[:-1] + (self.outfeatures,)