# \#tf-lite
**URL:** https://discuss.ai.google.dev/tag/tf-lite/131.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)
---
## [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…
---
## [TensorFlow Lite and 16 KB Page Size Support](https://discuss.ai.google.dev/t/tensorflow-lite-and-16-kb-page-size-support/93052)
**Author:** [@kevin\_santoki](https://discuss.ai.google.dev/u/kevin_santoki)\
**Replies:** 5\
**Last updated:** [February 10, 2026, 2:23am UTC](https://discuss.ai.google.dev/t/tensorflow-lite-and-16-kb-page-size-support/93052 "2026-02-10T02:23:03Z")
Issue: We’re using libtensorflowlite\_jni.so (from org.tensorflow:tensorflow-lite:2.17.0) in an Android project. We’re targeting SDK 35+ on Android devices that require 16 KB page size. But when inspecting the .so using r…
---
## [Issue Replicating TF-Lite Conv2D Quantized Inference Output](https://discuss.ai.google.dev/t/issue-replicating-tf-lite-conv2d-quantized-inference-output/109636)
**Author:** [@Jom\_Kuriakose](https://discuss.ai.google.dev/u/Jom_Kuriakose)\
**Replies:** 1\
**Last updated:** [January 12, 2026, 1:40pm UTC](https://discuss.ai.google.dev/t/issue-replicating-tf-lite-conv2d-quantized-inference-output/109636 "2026-01-12T13:40:57Z")
Issue Replicating TF-Lite Conv2D Quantized Inference Output I am trying to reproduce the exact layer-wise output of a quantized EfficientNet model (TFLite model, TensorFlow 2.17) by re-implementing Conv2D, DepthwiseConv2…
---
## [On-device training of simple autoencoder fails because of Select Ops](https://discuss.ai.google.dev/t/on-device-training-of-simple-autoencoder-fails-because-of-select-ops/103079)
**Author:** [@JPat\_MC](https://discuss.ai.google.dev/u/JPat_MC)\
**Replies:** 1\
**Last updated:** [November 10, 2025, 8:13pm UTC](https://discuss.ai.google.dev/t/on-device-training-of-simple-autoencoder-fails-because-of-select-ops/103079 "2025-11-10T20:13:47Z")
Hi, I am building and pre-training a basic autoencoder in Python with TF + Keras, then converting it to LiteRT “.tflite” model in order to deploy it into an ARM 32-bit Cortex-A7. Here is the model in question: NUM\_FEA…
---
## [TypeError: Negative indices with dynamic shapes on ResNet18 using ai\_edge\_torch](https://discuss.ai.google.dev/t/typeerror-negative-indices-with-dynamic-shapes-on-resnet18-using-ai-edge-torch/85510)
**Author:** [@Pradana\_Adnyana](https://discuss.ai.google.dev/u/Pradana_Adnyana)\
**Replies:** 1\
**Last updated:** [October 7, 2025, 4:21pm UTC](https://discuss.ai.google.dev/t/typeerror-negative-indices-with-dynamic-shapes-on-resnet18-using-ai-edge-torch/85510 "2025-10-07T16:21:25Z")
I am attempting to convert a torchvision.models.resnet18 model to TFLite using ai-edge-torch with dynamic batch size enabled via the dynamic\_shapes argument. But, it raised the error below: TypeError: Only non-negative…
---
## [On-Device training for LSTM or GRU Model](https://discuss.ai.google.dev/t/on-device-training-for-lstm-or-gru-model/28005)
**Author:** [@yycheng](https://discuss.ai.google.dev/u/yycheng)\
**Replies:** 1\
**Last updated:** [September 22, 2024, 6:47pm UTC](https://discuss.ai.google.dev/t/on-device-training-for-lstm-or-gru-model/28005 "2024-09-22T18:47:28Z")
Hi I’m new to tensorflow and I’m trying to make LSTM or GRU model to be enable to re-train on-device(Android) with tabular data (mostly customer interaction). I’m referencing these examples On-Device Training This is…