All pull requests

Pull requests

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

  #1396·eplatero97 opened Oct 1, 2026Contributor·

  ·

• Set logger level in create_logger so INFO messages are emitted

  #1395·qaic-mka opened Oct 1, 2026·

  ·

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

  #1394·quic-dhirajku opened Oct 1, 2026Contributor·

  ·

• Fix GPT-OSS CCL subfunction compilation

  #1393·abhishek-singh591 opened Oct 1, 2026Contributor·

  ·
  1

• Vlm disagg WF Infra and Qwen3vlmoe VLM with WF

  #1392·quic-akuruvil opened Oct 1, 2026Contributor·

  ·

• Fix prefill_qkv query and head chunking for MDP compilation

  #1391·tv-karthikeya opened Oct 1, 2026Contributor·

  ·
  2

• Skip scatter gather for prefill when bsz=fbz=1

  #1390·quic-mkshirsa opened Oct 1, 2026·

  ·
  6

• Add DFlash support for Qwen3.5 hybrid VLMs

  #1389·vjanfaza opened Oct 1, 2026Contributor·

  ·

• Vlm disagg WF Infra and Qwen3vlmoe VLM with WF

  #1387·quic-akuruvil opened Sep 30, 2026Contributor·

  ·

• Fix Gemma4 26b junk output issue

  #1385·nsivalin opened Sep 30, 2026Contributor·

  ·

• Gemma4 weight_free

  #1384·tchawada opened Sep 29, 2026Contributor·

  ·

• [WIP] Enabling MDP support for MQTH with p2p

  #1382·tv-karthikeya opened Sep 29, 2026Contributor·

  ·

• Set logger level in create_logger so INFO messages are emitted

  #1379·qaic-mka opened Sep 29, 2026·

  ·
  2

• Added weight-free and dynamo support for ASR models

  #1378·asmigosw opened Sep 29, 2026Contributor·

  ·
  2

• Enabled KV, QKV, HQKV, BHQKV, KV_HEADPAR, KV_BATCH_FOLD* blocking support for weightfree in causallm models

  #1373·quic-abhamidi opened Sep 28, 2026Contributor·

  ·
  2

• Fix for Gemma4 Junk Output Issue

  #1372·nsivalin opened Sep 28, 2026Contributor·

  ·

• Fix: Replace dynamic Slice in Llama4 CCL attention mask for QAIC export

  #1371·quic-vishali opened Sep 25, 2026Contributor·

  ·

• Enable blocking for weightfree

  #1370·quic-abhamidi opened Sep 25, 2026Contributor·

  ·
  1

• Mdp Fix with Full Batch Size

  #1369·mohiso22 opened Sep 24, 2026Contributor·

  ·

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

  1.23bugfix
  #1368·vbaddi opened Sep 24, 2026Contributor·

  ·
  4

  

• [vLLM-Pipeline]: Vllm benchmarking Pipeline for QEFF Models

  #1367·abukhoy opened Sep 24, 2026Contributor·

  ·

• Fix GPT-OSS FP16 export by replacing unsupported ONNX Clip ops

  #1365·quic-vishali opened Sep 24, 2026Contributor·

  ·
  4

• Add artifact-only compile and generation bundles for weight free export

  ci_ready
  #1363·quic-mamta opened Sep 23, 2026Contributor·

  ·
  21

  

• Enabling weight-free for embedding models

  #1362·asmigosw opened Sep 23, 2026Contributor·

  ·
  12

• [github docs]: Release/v1.23.0 and Release/v1.22.0 added in gh-page

  #1361·abukhoy opened Sep 23, 2026Contributor·

  ·
  1