# TensorFlow **URL:** https://discuss.ai.google.dev/c/tensorflow/21.md [Latest](https://discuss.ai.google.dev/latest.md) · [Categories](https://discuss.ai.google.dev/categories.md) · [Tags](https://discuss.ai.google.dev/tags.md) --- ## [TensorFlow dataset pipeline not reading some downloaded PPT/PDF files correctly](https://discuss.ai.google.dev/t/tensorflow-dataset-pipeline-not-reading-some-downloaded-ppt-pdf-files-correctly/187112)
**Author:** [@JohnJennifer](https://discuss.ai.google.dev/u/JohnJennifer)\ **Replies:** 1\ **Last updated:** [October 9, 2026, 8:54pm UTC](https://discuss.ai.google.dev/t/tensorflow-dataset-pipeline-not-reading-some-downloaded-ppt-pdf-files-correctly/187112 "2026-10-09T20:54:56Z")
Hi everyone, I am working on a small document dataset for a TensorFlow project. I have some presentation files in PPT/PDF format and I am trying to extract the text from them before putting the data into my TensorFlow p… --- ## [Best Practices for Deploying TensorFlow Models in Production Environments](https://discuss.ai.google.dev/t/best-practices-for-deploying-tensorflow-models-in-production-environments/183692)
**Author:** [@codeautomation](https://discuss.ai.google.dev/u/codeautomation)\ **Replies:** 0\ **Last updated:** [September 19, 2026, 10:37am UTC](https://discuss.ai.google.dev/t/best-practices-for-deploying-tensorflow-models-in-production-environments/183692 "2026-09-19T10:37:59Z")
Hi TensorFlow community, I have been exploring different approaches for building and deploying machine learning models with TensorFlow and wanted to learn from developers who are working with production ML systems. Bui… --- ## [TFRecord Writer writing patches in unexpected ways](https://discuss.ai.google.dev/t/tfrecord-writer-writing-patches-in-unexpected-ways/29894)
**Author:** [@SlimThugg](https://discuss.ai.google.dev/u/SlimThugg)\ **Replies:** 1\ **Last updated:** [September 10, 2026, 12:34pm UTC](https://discuss.ai.google.dev/t/tfrecord-writer-writing-patches-in-unexpected-ways/29894 "2026-09-10T12:34:38Z")
I have implemented this example notebook -a methodology to segment satellite imagery using Google Earth Engine (GEE) and Tensorflow- on my specific use case. The final step in this pipeline is uploading inference imager… --- ## [CPython 3.14 Free-Threaded TensorFlow Support — Upstream PR Series and Validation](https://discuss.ai.google.dev/t/cpython-3-14-free-threaded-tensorflow-support-upstream-pr-series-and-validation/180353)
**Author:** [@Mina\_Ibrahim](https://discuss.ai.google.dev/u/Mina_Ibrahim)\ **Replies:** 0\ **Last updated:** [September 1, 2026, 4:11am UTC](https://discuss.ai.google.dev/t/cpython-3-14-free-threaded-tensorflow-support-upstream-pr-series-and-validation/180353 "2026-09-01T04:11:41Z")
CPython 3.14 Free-Threaded TensorFlow Support — Upstream PR Series and Validation Hi everyone, I have been working on making TensorFlow and parts of its Python runtime ecosystem compatible with CPython 3.14 free-threade… --- ## [TF on Cuda 13.2 -- Possible?](https://discuss.ai.google.dev/t/tf-on-cuda-13-2-possible/177185)
**Author:** [@concho](https://discuss.ai.google.dev/u/concho)\ **Replies:** 1\ **Last updated:** [August 15, 2026, 9:32pm UTC](https://discuss.ai.google.dev/t/tf-on-cuda-13-2-possible/177185 "2026-08-15T21:32:46Z")
I’m wanting to deploy TF on a newer system which defaults to Cudatoolkit 13.2. Wanted to see the possibility of this? --- ## [Improving TensorFlow performance by hinting the Python GC about unreachable cycles](https://discuss.ai.google.dev/t/improving-tensorflow-performance-by-hinting-the-python-gc-about-unreachable-cycles/177931)
**Author:** [@Omri\_Palmon](https://discuss.ai.google.dev/u/Omri_Palmon)\ **Replies:** 0\ **Last updated:** [August 12, 2026, 5:55am UTC](https://discuss.ai.google.dev/t/improving-tensorflow-performance-by-hinting-the-python-gc-about-unreachable-cycles/177931 "2026-08-12T05:55:06Z")
I have recently tried to convince the Python community to add a feature to the Python GC which would accept hints that a certain group of objects is unreachable and can be deallocated. Such a feature would reduce the amo… --- ## [Will the ARM Mali GPU accelerate a Tensor Flow Lite model](https://discuss.ai.google.dev/t/will-the-arm-mali-gpu-accelerate-a-tensor-flow-lite-model/32079)
**Author:** [@Charry2014](https://discuss.ai.google.dev/u/Charry2014)\ **Replies:** 5\ **Last updated:** [July 31, 2026, 8:19pm UTC](https://discuss.ai.google.dev/t/will-the-arm-mali-gpu-accelerate-a-tensor-flow-lite-model/32079 "2026-07-31T20:19:04Z")
Hello - I am developing an object detection model in Python closely following Google’s examples. This runs on a Raspberry Pi 4 and works quite well. There are many other SBCs and many have the ARM Mali G610 GPU and other… --- ## [Gradients for Encoder in my BiGAN network are returning None values. Can someone with experience in this field identify the issue?](https://discuss.ai.google.dev/t/gradients-for-encoder-in-my-bigan-network-are-returning-none-values-can-someone-with-experience-in-this-field-identify-the-issue/27970)
**Author:** [@PaulD153](https://discuss.ai.google.dev/u/PaulD153)\ **Replies:** 1\ **Last updated:** [July 24, 2026, 11:17am UTC](https://discuss.ai.google.dev/t/gradients-for-encoder-in-my-bigan-network-are-returning-none-values-can-someone-with-experience-in-this-field-identify-the-issue/27970 "2026-07-24T11:17:56Z")
I’m implementing a BiGAN network for multivariate time-series data and I’m running into problems with training my Encoder. When calling tape.gradient(e\_loss, encoder.trainable\_variables) I’m receiving only None values. C… --- ## [Normalization And Categorical Encoding](https://discuss.ai.google.dev/t/normalization-and-categorical-encoding/175360)
**Author:** [@Charlie\_Barber](https://discuss.ai.google.dev/u/Charlie_Barber)\ **Replies:** 0\ **Last updated:** [July 20, 2026, 12:55am UTC](https://discuss.ai.google.dev/t/normalization-and-categorical-encoding/175360 "2026-07-20T00:55:39Z")
I am trying to really understand pre-processing layers as discussed in this tutorial . In the tutorial it is using a pet-finder dataset. And when it goes to normalize the PhotoAmt it creates the get\_normalization\_laye… --- ## [Getting nan loss while training ArcFace model](https://discuss.ai.google.dev/t/getting-nan-loss-while-training-arcface-model/41032)
**Author:** [@Pritam\_Solanki](https://discuss.ai.google.dev/u/Pritam_Solanki)\ **Replies:** 1\ **Last updated:** [July 18, 2026, 6:04pm UTC](https://discuss.ai.google.dev/t/getting-nan-loss-while-training-arcface-model/41032 "2026-07-18T18:04:47Z")
Dear Team, I am getting issue of loss nan while training arcface with pre-trained facenet model and also with resnet50 model. please suggest what changes can resolve the nan issue. Code Below: import os import tensor… --- ## [Deformable convolution and other custom ops](https://discuss.ai.google.dev/t/deformable-convolution-and-other-custom-ops/32546)
**Author:** [@Bhack](https://discuss.ai.google.dev/u/Bhack)\ **Replies:** 27\ **Last updated:** [July 11, 2026, 10:07pm UTC](https://discuss.ai.google.dev/t/deformable-convolution-and-other-custom-ops/32546 "2026-07-11T22:07:24Z")
Recently we had a refresh over a Deformable convloution WIP PR in Addons. I’ve cherry-picked this as an example as this requires us to maintain almost 3k lines of new code in the repository. This is maintainer-ship ove… --- ## [Tensorflow And Ubuntu 26](https://discuss.ai.google.dev/t/tensorflow-and-ubuntu-26/171573)
**Author:** [@Charlie\_Barber](https://discuss.ai.google.dev/u/Charlie_Barber)\ **Replies:** 1\ **Last updated:** [July 10, 2026, 4:02pm UTC](https://discuss.ai.google.dev/t/tensorflow-and-ubuntu-26/171573 "2026-07-10T16:02:34Z")
I am having a lot of issues trying to install Tensorflow on Ubuntu 26. It won’t work with the simple pip installl tensorflow. So I tried to install it from source and got to the very end of the bazel install and it fail… --- ## [Face Detector in MediaPipe model maker](https://discuss.ai.google.dev/t/face-detector-in-mediapipe-model-maker/39959)
**Author:** [@Jason\_Xie](https://discuss.ai.google.dev/u/Jason_Xie)\ **Replies:** 5\ **Last updated:** [June 15, 2026, 7:47am UTC](https://discuss.ai.google.dev/t/face-detector-in-mediapipe-model-maker/39959 "2026-06-15T07:47:51Z")
I see that face detection is missing in MediaPipe Model Maker and I was wondering whether it will be supported any time soon. Is there a roadmap for MediaPipe Model Maker that can be shared? --- ## [What is the maximum model size to run on Google Coral Dev Board Micro? or How to run FullSubNet on Coral mirco?](https://discuss.ai.google.dev/t/what-is-the-maximum-model-size-to-run-on-google-coral-dev-board-micro-or-how-to-run-fullsubnet-on-coral-mirco/147283)
**Author:** [@Alvis1](https://discuss.ai.google.dev/u/Alvis1)\ **Replies:** 0\ **Last updated:** [May 22, 2026, 8:18am UTC](https://discuss.ai.google.dev/t/what-is-the-maximum-model-size-to-run-on-google-coral-dev-board-micro-or-how-to-run-fullsubnet-on-coral-mirco/147283 "2026-05-22T08:18:29Z")
I am trying to run a FullSubNet model on the micro coral board, but it just stuck on the running interface. I converted the FullSubNet from pytorch to TensorFlow Lite by using Nobuco and TensorFlow Lite converter, result… --- ## [Mixed rounding behavior in quantized model using TFL Interpreter](https://discuss.ai.google.dev/t/mixed-rounding-behavior-in-quantized-model-using-tfl-interpreter/138662)
**Author:** [@Brian\_Provost](https://discuss.ai.google.dev/u/Brian_Provost)\ **Replies:** 2\ **Last updated:** [May 21, 2026, 5:58pm UTC](https://discuss.ai.google.dev/t/mixed-rounding-behavior-in-quantized-model-using-tfl-interpreter/138662 "2026-05-21T17:58:17Z")
Hello, I’m comparing TensorFlow Lite reference kernels (tf.lite.Interpreter with OpResolverType.BUILTIN\_REF) to CMSIS-NN int8 implementations. CMSIS-NN says it is bit-exact relative to TFL/TFLM for many operators. I wr… --- ## [Import tensorflow "Could not find cuda drivers on your machine, GPU will not be used"](https://discuss.ai.google.dev/t/import-tensorflow-could-not-find-cuda-drivers-on-your-machine-gpu-will-not-be-used/143227)
**Author:** [@Graham\_Smith](https://discuss.ai.google.dev/u/Graham_Smith)\ **Replies:** 3\ **Last updated:** [May 5, 2026, 9:12am UTC](https://discuss.ai.google.dev/t/import-tensorflow-could-not-find-cuda-drivers-on-your-machine-gpu-will-not-be-used/143227 "2026-05-05T09:12:55Z")
I’m trying to understand why tensorflow can’t find cuda drivers. python3 -c "import tensorflow" WARNING: All log messages before absl::InitializeLog() is called are written to STDERR I0000 00:00:1777752935.272880 6100… --- ## [Can tensorflow lite object detection model be trained with AMD grpahics card](https://discuss.ai.google.dev/t/can-tensorflow-lite-object-detection-model-be-trained-with-amd-grpahics-card/142512)
**Author:** [@Joe\_Smith3](https://discuss.ai.google.dev/u/Joe_Smith3)\ **Replies:** 1\ **Last updated:** [May 3, 2026, 6:17pm UTC](https://discuss.ai.google.dev/t/can-tensorflow-lite-object-detection-model-be-trained-with-amd-grpahics-card/142512 "2026-05-03T18:17:19Z")
I have an AMD graphics card, was wondering if tensorflow lite object detection is supported by AMD graphics card? If so is it just via ROCm? --- ## [Getting 0 nodes delegated while using TFLite C Library with TFLite Flex Delegate](https://discuss.ai.google.dev/t/getting-0-nodes-delegated-while-using-tflite-c-library-with-tflite-flex-delegate/72789)
**Author:** [@l1234567j222](https://discuss.ai.google.dev/u/l1234567j222)\ **Replies:** 1\ **Last updated:** [April 23, 2026, 3:27am UTC](https://discuss.ai.google.dev/t/getting-0-nodes-delegated-while-using-tflite-c-library-with-tflite-flex-delegate/72789 "2026-04-23T03:27:47Z")
I am using the TFLite C library to generate model inference C code for an edge device. In my use case, I require certain operations, such as tf.unpack with shape (0, 2), which are supported by TensorFlow but not current… --- ## [Does not contain symbol 'cc\_autoconf\_impl'](https://discuss.ai.google.dev/t/does-not-contain-symbol-cc-autoconf-impl/139439)
**Author:** [@brokencog](https://discuss.ai.google.dev/u/brokencog)\ **Replies:** 2\ **Last updated:** [April 20, 2026, 2:43pm UTC](https://discuss.ai.google.dev/t/does-not-contain-symbol-cc-autoconf-impl/139439 "2026-04-20T14:43:56Z")
I’m trying to build tensorflow locally, because that was given as a suggestion for resolving the error resulting from installing via pip. I’ll list the error from the pip install first, then the compile error. After do… --- ## [Regression in a dataset with flattened images and target reals](https://discuss.ai.google.dev/t/regression-in-a-dataset-with-flattened-images-and-target-reals/28301)
**Author:** [@Rafael\_Scatena](https://discuss.ai.google.dev/u/Rafael_Scatena)\ **Replies:** 1\ **Last updated:** [April 13, 2026, 12:03pm UTC](https://discuss.ai.google.dev/t/regression-in-a-dataset-with-flattened-images-and-target-reals/28301 "2026-04-13T12:03:27Z")
Hello, I am trying to set up a dataset with flattened lung images and 2 reals indicating the lung nodule location but I am facing some problems: import tensorflow as tf import tensorflow\_io as tfio import matplotlib.… --- ## [Memory leak with training different models in loops](https://discuss.ai.google.dev/t/memory-leak-with-training-different-models-in-loops/136413)
**Author:** [@DefinitlyNotAModel](https://discuss.ai.google.dev/u/DefinitlyNotAModel)\ **Replies:** 3\ **Last updated:** [April 7, 2026, 7:47am UTC](https://discuss.ai.google.dev/t/memory-leak-with-training-different-models-in-loops/136413 "2026-04-07T07:47:13Z")
:brain: TensorFlow Memory Growth Issue in Training Loop :pushpin: Summary I encountered a progressive memory growth issue when repeatedly creating, training, and deleting a tf.keras.Model inside a loop. Despite explicitl… --- ## [Why tensorflow switches to clang suddenly?](https://discuss.ai.google.dev/t/why-tensorflow-switches-to-clang-suddenly/31149)
**Author:** [@Quan\_Nie](https://discuss.ai.google.dev/u/Quan_Nie)\ **Replies:** 1\ **Last updated:** [April 2, 2026, 12:51pm UTC](https://discuss.ai.google.dev/t/why-tensorflow-switches-to-clang-suddenly/31149 "2026-04-02T12:51:20Z")
According to 從原始碼開始建構 | TensorFlow, tensorflow sets its default compiler as clang-16 since 2.13.0. May I ask the reason? --- ## [Error running tuner component in tfx pipeline](https://discuss.ai.google.dev/t/error-running-tuner-component-in-tfx-pipeline/27711)
**Author:** [@Brempong\_Appiah\_Dank](https://discuss.ai.google.dev/u/Brempong_Appiah_Dank)\ **Replies:** 1\ **Last updated:** [April 2, 2026, 8:24am UTC](https://discuss.ai.google.dev/t/error-running-tuner-component-in-tfx-pipeline/27711 "2026-04-02T08:24:07Z")
I’m trying to create a pipeline for ML project. But whenever I run the the Tuner component, I get this error --------------------------------------------------------------------------- TypeError … --- ## [Transformer for asynchronous multi-stream image time-series with online prediction?](https://discuss.ai.google.dev/t/transformer-for-asynchronous-multi-stream-image-time-series-with-online-prediction/137033)
**Author:** [@satadru](https://discuss.ai.google.dev/u/satadru)\ **Replies:** 0\ **Last updated:** [March 30, 2026, 12:20pm UTC](https://discuss.ai.google.dev/t/transformer-for-asynchronous-multi-stream-image-time-series-with-online-prediction/137033 "2026-03-30T12:20:10Z")
I have two streams of images, each stream corresponding to a different “channel” (e.g. different sensor modality). The streams are not synchronized — at any given moment, a new image arrives from one stream or the other,… --- ## [Issues Running Sanity Check](https://discuss.ai.google.dev/t/issues-running-sanity-check/25265)
**Author:** [@Mahrukh](https://discuss.ai.google.dev/u/Mahrukh)\ **Replies:** 1\ **Last updated:** [March 26, 2026, 3:27pm UTC](https://discuss.ai.google.dev/t/issues-running-sanity-check/25265 "2026-03-26T15:27:02Z")
Hi I have recently started open-source contributions and am facing issues when performing sanity check using the instructions mentioned in TensorFlow CONTRIBUTING.md guidelines. I have performed Python Coding Style chec… --- ## [Unknown error: Graph execution error](https://discuss.ai.google.dev/t/unknown-error-graph-execution-error/32215)
**Author:** [@Gayathri](https://discuss.ai.google.dev/u/Gayathri)\ **Replies:** 1\ **Last updated:** [March 26, 2026, 12:43pm UTC](https://discuss.ai.google.dev/t/unknown-error-graph-execution-error/32215 "2026-03-26T12:43:08Z")
I’m running into this error while training the model. Couldn’t find any resources online on how to solve it. Link to the complete code - Colab link this is the error message - Epoch 1/100 1/1 \[======================… --- ## [Missing Python wheel package building from source](https://discuss.ai.google.dev/t/missing-python-wheel-package-building-from-source/32204)
**Author:** [@IUSR](https://discuss.ai.google.dev/u/IUSR)\ **Replies:** 1\ **Last updated:** [March 25, 2026, 9:01am UTC](https://discuss.ai.google.dev/t/missing-python-wheel-package-building-from-source/32204 "2026-03-25T09:01:36Z")
Hello, I tried building TensorFlow from source following “Build from source”. I cloned the GitHub repo and checked out the latest 2.16.1 tag. Everything went well with bazelisk managing bazel’s versions, and yet I was m… --- ## [Request for Review: Deterministic Sorting Fix in auto\_control\_deps\_utils.py (PR #110822)](https://discuss.ai.google.dev/t/request-for-review-deterministic-sorting-fix-in-auto-control-deps-utils-py-pr-110822/133472)
**Author:** [@M\_Husnain](https://discuss.ai.google.dev/u/M_Husnain)\ **Replies:** 2\ **Last updated:** [March 19, 2026, 6:36pm UTC](https://discuss.ai.google.dev/t/request-for-review-deterministic-sorting-fix-in-auto-control-deps-utils-py-pr-110822/133472 "2026-03-19T18:36:00Z")
Hello TensorFlow Community, I am reaching out to request a maintainer review for Pull Request #110822, which addresses a non-deterministic resource categorization issue within the core framework. Link to PR: https://gi… --- ## [Building tensorflow from source for RTX5000 GPU series](https://discuss.ai.google.dev/t/building-tensorflow-from-source-for-rtx5000-gpu-series/65171)
**Author:** [@DriaanJ](https://discuss.ai.google.dev/u/DriaanJ)\ **Replies:** 8\ **Last updated:** [March 18, 2026, 12:54pm UTC](https://discuss.ai.google.dev/t/building-tensorflow-from-source-for-rtx5000-gpu-series/65171 "2026-03-18T12:54:01Z")
Nvidia’s new Blackwell GPUs need Cuda 12.8 and cuda compute capability 12.5 with sm\_120 architecture. I’ve installed the latest tensorflow 12.8 with GPU support, but it doesn’t run with the Blackwell architecture. So I … --- ## [Missing Method under c#](https://discuss.ai.google.dev/t/missing-method-under-c/42435)
**Author:** [@Pascal567](https://discuss.ai.google.dev/u/Pascal567)\ **Replies:** 1\ **Last updated:** [March 17, 2026, 6:36am UTC](https://discuss.ai.google.dev/t/missing-method-under-c/42435 "2026-03-17T06:36:31Z")
Hello Forum, i m struggling with KerasNet I use it with c# and install the latest built from nuget. I created an Stateful LSTM and my Problem is, i m unable to reset hidden states to change context. Nowhere i find a m… [Next page](https://discuss.ai.google.dev/c/tensorflow/21.md?page=1)