Issues

• LiteRT DENSIFY sparse FormatConverter: heap OOB write/read on crafted sparsity metadata, including a model that passes tflite::Verify

  google-ai-edge/LiteRT#11018 · Filter by author professor-moody opened on Oct 19, 2026

• Python GIL held by CompiledModel API?

  status:awaiting user responsetype:bug
  google-ai-edge/LiteRT#10994 · Filter by author dreiss opened on Oct 18, 2026
  1

  

• Getting started with LiteRT for mobile inference profiling

  type:support
  google-ai-edge/LiteRT#10977 · Filter by author zehnderdev opened on Oct 16, 2026
  1

  

• GPU accelerator fails on FULLY_CONNECTED with int8 weights + DEQUANTIZE ("Shape mismatch: {4096, 1, 1, 1024} vs {1, 1, 4096, 1024}")

  type:bug
  google-ai-edge/LiteRT#10934 · Filter by author ductai199x opened on Oct 15, 2026
  2

  

• [Web] Support multithreading with standard SIMD (without Relaxed SIMD)

  type:feature
  google-ai-edge/LiteRT#10903 · Filter by author bc-lee opened on Oct 15, 2026
  1

  

• UINT8 output remains unchanged after LiteRT OpenCL inference on Android

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#10800 · Filter by author WuJune opened on Oct 13, 2026
  2

  

• Variable kernels can read beyond malformed RESOURCE backing

  status:awaiting PR reviewstatus:awaiting user responsetype:bug
  google-ai-edge/LiteRT#10753 · Filter by author professor-moody opened on Oct 12, 2026
  3

  

• Quantized int8 DIV saturates small quotients to the int8 endpoints

  status:awaiting PR reviewtype:bug
  google-ai-edge/LiteRT#10745 · Filter by author vera-rubin opened on Oct 12, 2026
  2

  

• Out-of-bounds write after non-topological operator preparation leaves a stale output allocation

  status:awaiting user responsetype:bug
  google-ai-edge/LiteRT#10743 · Filter by author professor-moody opened on Oct 12, 2026
  5

  

• how to get the whole SDK for developing custom algorithm based on google pixel 9 pro phone(Tensor G4)

  status:awaiting LiteRTertype:supporttype:tensor sdk
  google-ai-edge/LiteRT#10740 · Filter by author mminrong opened on Oct 12, 2026
  6

  

• [Google Tensor SDK Beta][Tensor G6] AOT compiler returns INTERNAL for the Qwen3-0.6B decoder of Qwen3-ASR-0.6B and Fun-ASR-Nano-2512 (audio encoders compile fully)

  status:awaiting LiteRTertype:bugtype:tensor sdk
  google-ai-edge/LiteRT#10479 · Filter by author nckuot89-aba opened on Oct 2, 2026
  5

  

• [Feature] Support standard shape broadcasting in MLDrift GPU delegate (or automate 4D Reshape insertion for lower/mixed-rank inputs)

  type:acceleratorstype:featuretype:gpu delegate
  Feature

  google-ai-edge/LiteRT#10445 · Filter by author terryheo opened on Oct 1, 2026
  1

  

• [Google Tensor G5] AOT compiler worker crashes on large RESIZE_NEAREST_NEIGHBOR

  status:awaiting LiteRTertype:bugtype:tensor sdk
  google-ai-edge/LiteRT#10305 · Filter by author geneing opened on Sep 27, 2026
  2

  

• GPU accelerator (Adreno OpenCL, LiteRT 2.2.0): CompiledModel creation faults in clEnqueueNDRangeKernel once an unrolled LSTM passes ~85 sequential steps (T=84 compiles, T=88 SIGSEGV; independent of operator count, bytes and memory)

  status:awaiting LiteRTertype:acceleratorstype:bug
  google-ai-edge/LiteRT#10300 · Filter by author john-rocky opened on Sep 26, 2026
  3

  

• GPU accelerator (OpenCL / ML Drift) returns non-deterministic results for a single CONV_2D on Adreno 830

  status:awaiting LiteRTertype:acceleratorstype:bug
  google-ai-edge/LiteRT#10192 · Filter by author g-moschetti opened on Sep 23, 2026
  3

  

• macOS arm64 wheel 2.2.0: CompiledModel.from_file with HardwareAccelerator.GPU (Metal) ends the process with SIGSEGV on a Linear+ReLU graph; the 2.1.6 wheel runs it

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#10131 · Filter by author john-rocky opened on Sep 22, 2026
  2

  

• [Tensor SDK][AVF] Device assignment of the Tensor TPU mailbox context (Pixel 11): will app-launched pVMs get it, and is Tensor G5 in scope?

  status:awaiting LiteRTertype:supporttype:tensor sdk
  google-ai-edge/LiteRT#10081 · Filter by author SteveDeFacto opened on Sep 19, 2026
  3

  

• CompiledModel.from_file on a truncated .tflite ends the process with SIGSEGV (ai-edge-litert 2.2.0, macOS arm64); from_buffer and 2.1.6 raise

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#10080 · Filter by author john-rocky opened on Sep 19, 2026
  3

  

• Add py.typed and type hints to Python package

  type:feature
  google-ai-edge/LiteRT#9973 · Filter by author johnthagen opened on Sep 15, 2026
  1

  

• LiteRT.js / WASM: fp16-quantized PP-OCRv5 recognizer runs 21× slower than its fp32 parent on XNNPACK

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#9663 · Filter by author john-rocky opened on Aug 31, 2026
  3

  

• LiteRT.js / WebGPU: Matcha-TTS flow-matching decoder output diverges from WASM (mean abs 0.34, deterministic, Apple Metal)

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#9662 · Filter by author john-rocky opened on Aug 31, 2026
  2

  

• LiteRT.js / WebGPU: PP-OCRv5 recognizer deterministically misreads text vs WASM (fp32 weights too, Apple Metal)

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#9661 · Filter by author john-rocky opened on Aug 31, 2026
  3

  

• LiteRT.js / WebNN: Failed to execute 'conv2d' on 'MLGraphBuilder' for efficientvit Segmentation demo

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#9548 · Filter by author ibelem opened on Aug 25, 2026
  11

  

• CompiledModel::CreateOutputBuffers creates 4 bytes size buffer for CPU accelerator

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#9518 · Filter by author BohdanPetrykiv opened on Aug 23, 2026
  5

  

• MediaTek MT6878 NPU dispatch: SIGSEGV at first inference after "contradictory buffer requirements" warning for KV-cache index tensors

  status:awaiting LiteRTertype:bug
  google-ai-edge/LiteRT#9500 · Filter by author laknath opened on Aug 21, 2026
  5