All pull requests

Pull requests

• Align Gemma4 VLM precision and gate custom CUDA runtime

  #787·titaiwangms opened Oct 23, 2026Contributor·

  ·
  5

• docs: investigate external Bolt Mamba precision boundaries

  #786·rui-ren opened Oct 22, 2026Contributor·

  ·
  1

• Add opt-in sparse IQ2 MoE lowering for Nemotron-H Omni

  #785·rui-ren opened Oct 22, 2026Contributor·

  ·
  3

• Allow canonical Olive INT3 expert overrides in native QMoE export

  #783·jiafatom opened Oct 21, 2026·

  ·
  1

• Implement opt-in Qwen4-Exp ngram/main split

  #780·titaiwangms opened Oct 19, 2026Contributor·

  ·
  4

• Add CUDA Nemotron 3.5 Bolt Omni export with native mixed quantization

  #779·rui-ren opened Oct 15, 2026Contributor·

  ·
  24

• Document Qwen4-Exp MTP forward contract

  #777·titaiwangms with CopilotAI opened Oct 14, 2026·

  ·
  1

  6

  

• Propose opt-in Qwen4-Exp ngram/main split contract

  #776·titaiwangms with CopilotAI opened Oct 14, 2026·

  ·
  1

  8

  

• Add K2 Horizon model (llama.cpp dense + MoVA support)

  #763·gramalingam opened Oct 8, 2026Collaborator·

  ·
  18

• Add Nemotron Lightning target/MTP graph exports and explicit BF16 reconstruction

  #761·rui-ren opened Oct 7, 2026Contributor·

  ·
  36

• Add Clef-Flash joint schema decision model support

  #760·gramalingam opened Oct 7, 2026Collaborator·

  ·
  14

• Align Qwen4-Exp CUDA math and bound sparse-indexer masks

  #757·tianleiwu opened Oct 5, 2026·2 of 4 tasks·

  ·
  2

• [Atlas] onboard CLM-v0.1-8B

  #756·rui-ren opened Oct 4, 2026Contributor·

  ·
  13

• Pack compatible INT4 QKV projections for CUDA exports

  #754·titaiwangms opened Oct 2, 2026Contributor·3 of 9 tasks·

  ·
  6

• Add dense Qwen PagedAttention export for ORT GenAI

  #753·titaiwangms opened Oct 1, 2026Contributor·

  ·
  7

• Standardize symbolic dimensions in common ORT GenAI decoder exports

  #751·tadani3 opened Oct 1, 2026Contributor·

  ·
  3

• Add multi-component export support for CLM and KEV

  #749·apsonawane opened Sep 29, 2026Contributor·

  ·
  29

• Add gemma4 OpenVINO NPU support

  #746·xiaoyu-work opened Sep 23, 2026Member·

  ·
  14

• Separate target build contracts from graph rewrites

  #743·xiaoyu-work opened Sep 21, 2026Member·

  ·
  4

• [TensorRT] Add static-cache export support and numerical runtime coverage

  #741·rui-ren opened Sep 17, 2026Contributor·

  ·
  7

• Add experimental Qwen3.8 packed hybrid export for ORT GenAI Engine

  #738·titaiwangms with CopilotAI opened Sep 16, 2026·

  ·
  1

  14

  

• Bump ruff from 0.16.2 to 0.16.6 in /requirements/lintrunner

  dependenciespython
  #734·dependabot[bot] opened Sep 11, 2026Bot·

  ·
  3

• Add MatterGen crystal diffusion score-core export

  #731·justinchuby opened Sep 4, 2026Member·

  ·
  22

• Add TimesFM 3 multivariate forecasting support

  #718·justinchuby opened Sep 1, 2026Member·

  ·
  8

• Add GLM-5.3 Flash multimodal support

  automated
  #713·justinchuby opened Aug 31, 2026Member·

  ·
  5