Searched refs:bias_scale (Results 1 – 2 of 2) sorted by relevance
263 const double bias_scale = biasShape.scale; in GetQuantizedConvolutionMultipler() local266 NN_OPS_CHECK(std::abs(input_product_scale - bias_scale) <= in GetQuantizedConvolutionMultipler()267 1e-6 * std::min(input_product_scale, bias_scale)); in GetQuantizedConvolutionMultipler()
1033 * with zeroPoint of 0 and bias_scale == input_scale * filter_scale.4206 * of 0 and bias_scale == input_scale * filter_scale. For filter tensor4209 * 0 and bias_scale of 0. The actual scale of each value 'i' is equal to4210 * bias_scale[i] = input_scale * filter_scale[i].4251 * of 0 and bias_scale == input_scale * filter_scale. For filter tensor4254 * 0 and bias_scale of 0. The actual scale of each value 'i' is equal to4255 * bias_scale[i] = input_scale * filter_scale[i].5640 * with zeroPoint of 0 and bias_scale == input_scale * filter_scale.5643 * and bias_scale of 0. The actual scale of each value 'i' is equal to5644 * bias_scale[i] = input_scale * filter_scale[i].[all …]