Issues

• perf: residual batched decode gaps left by the Metal acceptance of #2566

  platform:macospriority:mediumstatus:readytype:performance
  lablup/mlxcel#2817 · Filter by author inureyes opened on Nov 1, 2026

• [nightly-verify] main is red

  priority:highstatus:readytype:bug
  lablup/mlxcel#2806 · Filter by author github-actions opened on Oct 31, 2026

• perf(cuda): MoE batched decode reads top_k * B expert matrices per gather

  area:corepriority:mediumstatus:in-progresstype:performance
  lablup/mlxcel#2784 · Filter by author inureyes opened on Oct 31, 2026
  1

• perf(cuda): single-kernel tanh GELU so tanh-switched families match erf prefill cost

  area:coreplatform:linuxpriority:lowstatus:readytype:performance
  lablup/mlxcel#2780 · Filter by author inureyes opened on Oct 31, 2026

• Explain the temperature 0 flip between plain and block-built KV history

  area:corepriority:mediumstatus:readytype:bug
  lablup/mlxcel#2778 · Filter by author inureyes opened on Oct 30, 2026
  1

• fix: join side-model workers and finish remaining exit-path gaps after #2738

  area:inferencepriority:lowstatus:readytype:bug
  lablup/mlxcel#2745 · Filter by author inureyes opened on Oct 30, 2026
  1

• fix(logging): close the remaining lost-diagnostic gaps after #2724

  priority:lowstatus:readytype:bug
  lablup/mlxcel#2725 · Filter by author inureyes opened on Oct 29, 2026

• test(rocm): validate NVFP4 kernels on a wave64 (CDNA) GPU

  platform:linuxpriority:lowstatus:blockedtype:test
  lablup/mlxcel#2699 · Filter by author inureyes opened on Oct 28, 2026
  1

• perf(cuda): re-measure the fused-MoE Dff bound and GB10 fused-MoE defaults without the JIT compile in the window

  arch:moearea:benchmarkplatform:linuxpriority:lowstatus:readytype:performance
  lablup/mlxcel#2540 · Filter by author inureyes opened on Oct 22, 2026

• [nightly-verify] the run did not finish

  priority:highstatus:readytype:bug
  lablup/mlxcel#2529 · Filter by author github-actions opened on Oct 21, 2026
  8

• feat: Extend conversation compaction to Chat Completions, compaction_trigger and multi-pass summaries

  area:inferencepriority:lowstatus:readytype:enhancement
  lablup/mlxcel#2523 · Filter by author adrysn opened on Oct 21, 2026

• perf(moe): measure the switch gate/up kernel at batch decode

  arch:moearea:benchmarkpriority:mediumstatus:readytype:performance
  lablup/mlxcel#2515 · Filter by author inureyes opened on Oct 21, 2026
  1

• perf(speculative): close batched DFlash gaps left by the uniform clamp

  area:inferencepriority:mediumstatus:readytype:enhancement
  lablup/mlxcel#2442 · Filter by author inureyes opened on Oct 18, 2026
  2

• investigate: gelu_approx computes erf GELU where upstream models use the tanh approximation

  area:corearea:modelspriority:mediumstatus:investigationtype:bug
  lablup/mlxcel#2423 · Filter by author inureyes opened on Oct 16, 2026
  2

• feat(models): Support IndicOCR (indic_ocr) layout plus Qwen3.5 OCR pipeline

  arch:hybridarea:modelsmodelsize:smallmodeltype:vlmpriority:lowstatus:blockedtype:enhancement
  lablup/mlxcel#2414 · Filter by author inureyes opened on Oct 16, 2026

• feat(detect): Add YOLO11 detector for OmniParser v2 icon detection

  area:cliarea:modelsmodelsize:smallmodeltype:vlmpriority:mediumstatus:readytype:enhancement
  lablup/mlxcel#2413 · Filter by author inureyes opened on Oct 16, 2026

• feat: Load compressed-tensors float-quantized FP8 checkpoints

  area:modelspriority:lowstatus:readytype:enhancement
  lablup/mlxcel#2411 · Filter by author inureyes opened on Oct 16, 2026

• perf: Trim the Gemma 4 KV-shared tail to the sampled row in prefill

  area:modelspriority:lowstatus:readytype:performance
  lablup/mlxcel#2410 · Filter by author inureyes opened on Oct 16, 2026

• feat(vlm): add MiMo-V2.6 (mimo_v2) support

  arch:hybridarea:modelsmodelsize:xlargemodeltype:omnipriority:mediumstatus:blockedtype:enhancement
  lablup/mlxcel#2409 · Filter by author inureyes opened on Oct 16, 2026

• perf: Use absorbed MLA attention for multi-token steps below crossover

  area:modelspriority:lowstatus:readytype:performance
  lablup/mlxcel#2408 · Filter by author inureyes opened on Oct 16, 2026

• perf: Add the packed Dk=128 gated-delta Metal kernel for prefill

  arch:hybridarea:coreplatform:macospriority:lowstatus:readytype:performance
  lablup/mlxcel#2407 · Filter by author inureyes opened on Oct 16, 2026

• feat: Add tensor-parallel support for Qwen3 MoE

  arch:moearea:modelsmodeltype:textpriority:lowstatus:readytype:enhancement
  lablup/mlxcel#2405 · Filter by author inureyes opened on Oct 16, 2026

• feat: Add pipeline stage executors for Qwen3 MoE, Qwen2, DeepSeek-V3.2

  area:modelsmodeltype:textpriority:lowstatus:readytype:enhancement
  lablup/mlxcel#2404 · Filter by author inureyes opened on Oct 16, 2026

• feat(vlm): add GLM-5 Next (glm5_next) support

  arch:hybridarea:modelsmodelsize:xlargemodeltype:vlmpriority:mediumstatus:blockedtype:enhancement
  lablup/mlxcel#2403 · Filter by author inureyes opened on Oct 16, 2026

• feat: Render per-group video timestamps for the Qwen3-VL family

  area:modelsmodeltype:vlmpriority:mediumstatus:readytype:enhancement
  lablup/mlxcel#2402 · Filter by author inureyes opened on Oct 16, 2026