All pull requests

Pull requests

• Add libneuronusdk_adapter.8.mtk.so to load symbols if present.

  #11446·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• litert: Add --use_host_memory_buffers, --host_memory_buffer_alignment, and --simulate_cpu_readback flags to benchmark_litert_model.

  #11445·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• fix(tflite): handle non-positive sizes in portable float quantization

  #11444·professor-moody opened Nov 2, 2026·

  ·

• fix(tflite): validate subgraph tensor indices in kernel_util lookups

  #11443·professor-moody opened Nov 2, 2026·

  ·

• fix(tflite): validate node and tensor indices in graph partitioning

  #11442·professor-moody opened Nov 2, 2026·

  ·

• litert: Map synthetic quantization tensors to signature tensors in DelegateKernelLiteRt.

  #11441·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• Automated Code Change

  #11439·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• Automated Code Change

  #11438·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• Populate LiteRT version during on-device compilation on Google Tensor.

  #11437·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• Use a patch file for dawn C++17 compat fixes

  #11436·JeevakaPrabu opened Nov 2, 2026Contributor·

  ·

• Update XNNPACK dependencies in XLA

  #11434·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• Clamp blockwise quantization scales to float16 normal minimum and apply min_bound after clipping in tensor_zp_scale_from_min_max.

  #11433·copybara-service[bot] opened Nov 2, 2026Bot·

  ·

• benchmark_model: add --print_partition_report.

  #11432·copybara-service[bot] opened Nov 1, 2026Bot·

  ·

• Automated Code Change

  #11431·copybara-service[bot] opened Nov 1, 2026Bot·

  ·

• Automated Code Change

  #11430·copybara-service[bot] opened Nov 1, 2026Bot·

  ·

• Automated Code Change

  #11429·copybara-service[bot] opened Nov 1, 2026Bot·

  ·

• Deduplicate composite decomposition subgraph indices during model partitioning.

  #11427·copybara-service[bot] opened Nov 1, 2026Bot·

  ·

• Add a !tfl.blockwise_quant element type for N-D blockwise quantization.

  #11426·copybara-service[bot] opened Oct 31, 2026Bot·

  ·

• [OpenVINO] Serve the NPU weights pool from the model file, no staging

  #11425·akodanka opened Oct 31, 2026Contributor·

  ·

• Automated Code Change

  #11424·copybara-service[bot] opened Oct 31, 2026Bot·

  ·

• Automated Code Change

  #11423·copybara-service[bot] opened Oct 31, 2026Bot·

  ·

• Automated Code Change

  #11421·copybara-service[bot] opened Oct 31, 2026Bot·

  ·

• Automated Code Change

  #11419·copybara-service[bot] opened Oct 31, 2026Bot·

  ·

• The embedding row prefetch (run option SetPrefetchEmbeddingTokenIds)

  #11418·copybara-service[bot] opened Oct 31, 2026Bot·

  ·

• On POSIX, LiteRT-LM maps every section of a .litertlm file

  #11417·copybara-service[bot] opened Oct 31, 2026Bot·

  ·