Git Product home page Git Product logo

Comments (3)

advaitjain avatar advaitjain commented on September 26, 2024

Yeah, I have been puzzled by this in the past as well. The not completely satisfying answer is that TfLite and TFLM currently implicitly assume that ints are 32 bit.

In this particular case we have two different functions that expect per_channel_output_shift to be an int* and an int32_t*

tflite-micro/tensorflow/lite/micro/kernels/depthwise_conv.cc

Lines 71 to 73 in 68dd0cf

reference_integer_ops::DepthwiseConvPerChannel(
DepthwiseConvParamsQuantized(params, data),
data.per_channel_output_multiplier, data.per_channel_output_shift,

tflite-micro/tensorflow/lite/kernels/internal/reference/integer_ops/depthwise_conv.h

Lines 22 to 24 in 68dd0cf

inline void DepthwiseConvPerChannel(
const DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const RuntimeShape& input_shape,

tflite-micro/tensorflow/lite/micro/kernels/depthwise_conv_common.cc

Lines 111 to 117 in f583f92

TF_LITE_ENSURE_STATUS(tflite::PopulateConvolutionQuantizationParams(
context, input, filter, bias, output, params.activation,
&data->output_multiplier, &data->output_shift,
&data->output_activation_min, &data->output_activation_max,
data->per_channel_output_multiplier,
reinterpret_cast<int*>(data->per_channel_output_shift),
output_channels));

tflite-micro/tensorflow/lite/kernels/kernel_util.cc

Lines 210 to 215 in 68dd0cf

TfLiteStatus PopulateConvolutionQuantizationParams(
TfLiteContext* context, const TfLiteTensor* input,
const TfLiteTensor* filter, const TfLiteTensor* bias, TfLiteTensor* output,
const TfLiteFusedActivation& activation, int32_t* multiplier, int* shift,
int32_t* output_activation_min, int32_t* output_activation_max,
int32_t* per_channel_multiplier, int* per_channel_shift, int num_channels) {

from tflite-micro.

alanvgreen avatar alanvgreen commented on September 26, 2024

I suppose 32 bit integers are a fairly safe assumption, but it would be nice to make the type names consistent.

Would you be open to a patch to make the PopulateConvolutionQuantizationParams() per_channel_shift parameter an int32_t* , and remove the reinterpret_cast<>s ?

from tflite-micro.

advaitjain avatar advaitjain commented on September 26, 2024

Yeah, that's going to help. The change to PopulateConvolutionQuantizationParams will have to go in the tensorflow/tensorflow repo and it will then get synced to the current repo automatically.

from tflite-micro.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.