All pull requests

Pull requests

• Optimize GDN state updates for decode

  #1429·quic-mkshirsa opened Oct 9, 2026Contributor·

  ·

• Dev/weightfree models integration

  #1427·kdulla opened Oct 9, 2026Contributor·

  ·

• docs: migrate README links to Qualcomm URLs

  #1426·Khoi VN (vnk8071) opened Oct 9, 2026Member·6 tasks done·

  ·
  1

• fix(dynamo) : get reusable subfunctions using reuse_hash_fn passed for mixed-layer models

  #1425·smedhe opened Oct 9, 2026Contributor·

  ·

• Enable qwe3.8 model

  #1424·Amit Raj (quic-amitraj) opened Oct 9, 2026Contributor·

  ·

• Fix qaic_config attribute-identity mismatch in VLM DualQPC paged attention

  #1423·Shivani Athavale (athavale-shivani) opened Oct 8, 2026Contributor·

  ·

• vb/minimax_weightfree

  #1422·Vinayak Baddi (vbaddi) opened Oct 8, 2026Contributor·

  ·

• fix(GDN): Restore stale conv/recurrent state reset for all inference (#1402)

  #1420·Rishin Raj (quic-rishinr) opened Oct 8, 2026Contributor·

  ·

• fix(qwen3vlmoe): keep batch-fold KV range ops inside decoder subfunction (#1368)

  #1419·Rishin Raj (quic-rishinr) opened Oct 8, 2026Contributor·

  ·

• Gemma4 weight free

  #1418·Tanisha Chawada (tchawada) opened Oct 8, 2026Contributor·

  ·

• Fix prefill_qkv query and head chunking for MDP compilation (#1391)

  #1417·Karthikeya (tv-karthikeya) opened Oct 8, 2026Contributor·

  ·
  1

• Skip scatter gather for prefill when bsz=fbz=1 (#1390)

  #1416·Rishin Raj (quic-rishinr) opened Oct 8, 2026Contributor·

  ·

• [QEff] : Dump full set of inputs for inference

  #1415·Mamta Singh (quic-mamta) opened Oct 7, 2026Contributor·

  ·
  4

• Removed custom_io check in bf16 dtype case

  #1414·asmigosw opened Oct 7, 2026Contributor·

  ·

• Fix BF16 logits size mismatch for CB in VLMs

  #1412·asmigosw opened Oct 6, 2026Contributor·

  ·

• feat(gpt_oss): wire QEffGptOssMLP into MoEBlockMixin + serialisation fixes

  #1411·Shrubabati7 opened Oct 6, 2026·5 tasks·

  ·
  3

• fix(weightfree): Keep MoE expert parallel weights canonical

  enhancementqeff.weightfree
  #1409·Vinayak Baddi (vbaddi) opened Oct 6, 2026Contributor·

  ·
  

• Add MolmoPoint support

  #1406·samsures-qti opened Oct 6, 2026·

  ·

• Fix Gemma4 Vision output issue

  #1405·Nishanth Sivalingam (nsivalin) opened Oct 5, 2026Contributor·

  ·

• [Fix] Gemma-3 CausalLM test failure

  #1404·Tanisha Chawada (tchawada) opened Oct 5, 2026Contributor·

  ·

• Add blocking support with weightfree and dynamo for CausalLM models

  #1403·Anusha V.S Bhamidipati (quic-abhamidi) opened Oct 5, 2026Contributor·

  ·
  23

• Qwen38 flash decode

  #1398·Onkar Chougule (ochougul) opened Oct 3, 2026Contributor·

  ·

• [Spec Decode] Enable SpD target num_logits_to_keep for MoE models

  #1396·Erick Platero (eplatero97) opened Oct 1, 2026Contributor·

  ·
  2

• Rebased vlm weightfree export for Qwen 2.5, 3 and 3 VL MoE models.

  #1394·Dhiraj Kumar Sah (quic-dhirajku) opened Oct 1, 2026Contributor·

  ·
  5

• Fix GPT-OSS CCL subfunction compilation

  ready for CI
  #1393·Abhishek Kumar Singh (abhishek-singh591) opened Oct 1, 2026Contributor·

  ·
  3