Hi, I'm using a tflite model in android which runs without problem w

TFLite runtime with select ops support in windows about tflite-support HOT 4 CLOSED

tensorflow commented on July 19, 2024

TFLite runtime with select ops support in windows

from tflite-support.

Comments (4)

terryheo commented on July 19, 2024

Could you share why you don't want to install TF 2.4 package?
We have TFLite only PIP distribution: https://www.tensorflow.org/lite/guide/python, but you can't use TF ops with this package.

We have an experimental way to build TFLite PIP package with Flex delegate.
I haven't tested it on Windows, but you might be able to use this.
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/pip_package

Comparing this to TF 2.4 package, it's bit lighter but it's less preferable since you need to build the package.

bash CUSTOM_BAZEL_FLAGS=--define=tflite_pip_with_flex=true tensorflow/lite/tools/pip_package/build_pip_package_with_bazel.sh windows

from tflite-support.

pourfard commented on July 19, 2024

Thanks for quick reply. Because its size is huge. TF 2.4 wheel file is about 400mb. I want to deploy a small application for face detection. The model itself is about 1mb but I should embed whole TF package inside my application. I found Tf-Nightly smaller than the main package. Its about 150mb but it still is very large for the application. On android TFlite library with select ops is about 23mb for each architecture (arm, x86,...) but I could not find any small package for that in windows. I think I should compile it from source code. But I still think it size will not be less than tf-nightly?

…

On Wed, Nov 25, 2020, 4:56 AM Terry Heo ***@***.***> wrote: Could you share why you don't want to install TF 2.4 package? We have TFLite only PIP distribution: https://www.tensorflow.org/lite/guide/python, but you can't use TF ops with this package. We have an experimental way to build TFLite PIP package with Flex delegate. I haven't tested it on Windows, but you might be able to use this. https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/pip_package Comparing this to TF 2.4 package, it's bit lighter but it's less preferable since you need to build the package. bash CUSTOM_BAZEL_FLAGS=--define=tflite_pip_with_flex=true tensorflow/lite/tools/pip_package/build_pip_package_with_bazel.sh windows — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#237 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALCEAJ3DJTTCN4DYPVQPLYDSRRMK7ANCNFSM4T57NC7A> .

from tflite-support.

terryheo commented on July 19, 2024

I see. The compiled wheel was less than 50MB. It could be worth to try.

from tflite-support.

pourfard commented on July 19, 2024

Sorry if I ask this question here.

when i run this command :

CUSTOM_BAZEL_FLAGS=--define=tflite_pip_with_flex=true tensorflow/lite/tools/pip_package/build_pip_package_with_bazel.sh windows

there is an error:

ERROR: C:/porsapp/tensorflow/tensorflow/tensorflow/core/common_runtime/BUILD:1651:1: Illegal ambiguous match on configurable attribute "deps" in //tensorflow/core/common_runtime:core_cpu_internal:
//tensorflow:windows
//tensorflow:no_gcp_support
Multiple matches are not allowed unless one is unambiguously more specialized.
ERROR: Analysis of target '//tensorflow/lite/python/interpreter_wrapper:_pywrap_tensorflow_interpreter_wrapper' failed; build aborted:

but when i run this:

tensorflow/lite/tools/pip_package/build_pip_package_with_bazel.sh

there is no error.

I've checked out Tensorflow r.2.4. Do you know what the problem is?

from tflite-support.

TFLite runtime with select ops support in windows about tflite-support HOT 4 CLOSED

Comments (4)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent