Issues

• [Accuracy] Qwen-Image text encoder assembles the M-RoPE table with contiguous sections instead of the Qwen2.5-VL interleave

  bug
  NVIDIA/TensorRT-Model-Connect#1759 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] Qwen-Image text encoder computes attention scores in BF16 and loses the softmax for large logits

  bug
  NVIDIA/TensorRT-Model-Connect#1758 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] Qwen-Image-Edit sends raster-order pixels to a vision engine that expects merge-group patch order

  bug
  NVIDIA/TensorRT-Model-Connect#1757 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] Qwen-Image-Edit resizes input images with different filters than the reference processors

  bug
  NVIDIA/TensorRT-Model-Connect#1756 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] Qwen-Image-Edit ignores zero_cond_t and modulates condition tokens with the denoising timestep

  bug
  NVIDIA/TensorRT-Model-Connect#1755 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] PersonaPlex FP16 bundles decode with an FP16 Mimi decoder instead of the reference FP32 codec

  bug
  NVIDIA/TensorRT-Model-Connect#1750 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] PersonaPlex resamples input audio differently from the reference loader and changes the Mimi input codes

  bug
  NVIDIA/TensorRT-Model-Connect#1749 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] DINOv3 FP16 ViT patch embedding accumulates in FP16 and misses the feature-parity gate

  bug
  NVIDIA/TensorRT-Model-Connect#1747 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] InternVL3 runtime encodes one image tile and a different prompt than InternVLProcessor

  bug
  NVIDIA/TensorRT-Model-Connect#1745 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] GLM-ASR long clips return empty or degenerate transcripts

  bug
  NVIDIA/TensorRT-Model-Connect#1739 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] MoGe focal/shift recovery stops far from the optimum on shallow-depth images

  bug
  NVIDIA/TensorRT-Model-Connect#1737 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] SAM3 text-only image prompts attend to an extra empty-geometry token

  bug
  NVIDIA/TensorRT-Model-Connect#1735 · Filter by author chaofengw-nv opened on Oct 16, 2026
  

• [Accuracy] MoGe selected convolutions replace Native replicate padding with zeros

  bug
  NVIDIA/TensorRT-Model-Connect#1727 · Filter by author chaofengw-nv opened on Oct 15, 2026
  

• Feature: Support fishaudio/drama-3 (multi-speaker TTS, 35B-A3B MoE)

  NVIDIA/TensorRT-Model-Connect#1700 · Filter by author jkzhang7 opened on Oct 14, 2026
  4

• AIPerf text stage F: qualify a speculative-decoding prompt corpus

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1683 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf text stage E: qualify stateless Responses API

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1682 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf text stage D: qualify controlled shared-prefix workloads

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1681 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf text stage C: qualify production trace replay

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1680 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf text stage B: qualify multi-turn chat

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1679 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf text stage A: qualify exact real-prompt inputs

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1678 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf stage 10: qualify all TRTMC video-language question-answering models

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1677 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf stage 9: qualify all TRTMC audio-language question-answering models

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1676 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf stage 8: qualify all TRTMC text-to-video models

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1675 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf stage 7: qualify all TRTMC reference-image editing models

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1674 · Filter by author yizhuoz004 opened on Oct 13, 2026
  

• AIPerf stage 6: qualify all TRTMC text-to-image models

  enhancement
  Task

  NVIDIA/TensorRT-Model-Connect#1673 · Filter by author yizhuoz004 opened on Oct 13, 2026