Issues

• Feature: support allenai/Molmo2-8B

  enhancement
  NVIDIA/TensorRT-Model-Connect#2433 · Filter by author willg-nv opened on Nov 13, 2026
  

• Feature: support facebook/musicgen-medium

  enhancement
  NVIDIA/TensorRT-Model-Connect#2432 · Filter by author willg-nv opened on Nov 13, 2026

• Feature: support PaddlePaddle/PP-DocLayoutV3_safetensors

  enhancement
  NVIDIA/TensorRT-Model-Connect#2431 · Filter by author willg-nv opened on Nov 13, 2026
  

• Feature: support llava-hf/llava-1.5-7b-hf

  enhancement
  NVIDIA/TensorRT-Model-Connect#2430 · Filter by author willg-nv opened on Nov 13, 2026
  

• Feature: support Salesforce/blip-image-captioning-base

  enhancement
  NVIDIA/TensorRT-Model-Connect#2426 · Filter by author willg-nv opened on Nov 13, 2026
  

• Feature: support lightonai/LightOnOCR-2-1B

  enhancement
  NVIDIA/TensorRT-Model-Connect#2424 · Filter by author willg-nv opened on Nov 13, 2026

• TensorRT 11.4: FP32 decomposable IAttention between FP16 casts fails to build with an FP32 mask and keeps its scores in FP16 (NaN)

  bugnvbug:myelin
  NVIDIA/TensorRT-Model-Connect#2423 · Filter by author chaofengw-nv opened on Nov 13, 2026
  

• mark_debug on an IAttention output fails the TensorRT build with a Myelin assertion (TensorRT 11.4 and 11.1)

  bugnvbug:myelin
  NVIDIA/TensorRT-Model-Connect#2422 · Filter by author chaofengw-nv opened on Nov 13, 2026
  

• TensorRT 11.4 on GB300: FP8 Q/K/V GEMMs that share one input and feed IAttention return NaN, so FLUX.2 FP8 cannot keep its FP8 GEMMs

  bugnvbug:myelin
  NVIDIA/TensorRT-Model-Connect#2421 · Filter by author chaofengw-nv opened on Nov 13, 2026
  

• Feature: support CohereLabs/cohere-transcribe-arabic-07-2026

  enhancement
  NVIDIA/TensorRT-Model-Connect#2419 · Filter by author willg-nv opened on Nov 13, 2026

• Feature: support HuBERT Large CTC (HubertForCTC)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2375 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support GLM-OCR (GlmOcrForConditionalGeneration)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2372 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support OneFormer (OneFormerForUniversalSegmentation)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2370 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support Apertus 8B Instruct (ApertusForCausalLM)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2369 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support MobileBERT (MobileBertForPreTraining)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2367 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Bug: native qwen tokenizer mis-splits upper-case contractions and apostrophes before newlines

  NVIDIA/TensorRT-Model-Connect#2366 · Filter by author 0z5a opened on Nov 12, 2026

• Feature: support Gemma 3n (Gemma3nForConditionalGeneration)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2364 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support Voxtral Mini 3B (VoxtralForConditionalGeneration)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2362 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support Stable Diffusion 3.5 Medium (StableDiffusion3Pipeline)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2356 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support PaliGemma (PaliGemmaForConditionalGeneration)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2351 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support Nemotron 3 Diarization (Nemotron3DiarizationForAudioFrameClassification)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2344 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support MOSS-Transcribe-Diarize (MossTranscribeDiarizeForConditionalGeneration)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2343 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support LayoutLM document question answering (LayoutLMForQuestionAnswering)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2342 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support Swin Transformer V2 (Swinv2ForImageClassification)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2338 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1

  

• Feature: support TrOCR (VisionEncoderDecoderModel)

  enhancement
  NVIDIA/TensorRT-Model-Connect#2335 · Filter by author qikail-ctrl opened on Nov 12, 2026
  1