Repository Issue Activity (beta)

#2481 Your project is on StackMap — a curated map of the AI stack

#1871 Clarify canonical documentation between guides and example READMEs (pruning)

#2456 GLM-5.3 nvfp4 checkpoint

#2442 qdq_to_dq fails on MaxViT due to unsupported Transpose after DequantizeLinear

#2424 calling mtq.quantize causes AttributeError: 'tuple' object has no attribute 'dim'

#2209 Per-expert weight_quantizer._amax fails Megatron validate_sharding_integrity on topology reshard (TEGroupedMLP NVFP4)

#215 INT8 to FP8 scale conversion

#2011 NVFP4 support for Qwen3_5MoeExperts (fused MoE): quantizer registration + export serializer

#1699 [RFC] NVIDIA Model Optimizer — Product Roadmap

#2337 Only 1 sample from AutoCast calibration data set is truly used

#1926 [Bugs] Second batch of verified findings from the unit-coverage initiative (fold_weight crash on Llama-4/GPT-OSS MoE, KD-loss default footgun, and smaller items)

#2330 After applying NVFP4 quantization followed by compression, RealQuantLinear fails to locate the corresponding GEMM implementation during inference.

#2326 CI: onnx example lanes fail with CUDNN_STATUS_SUBLIBRARY_LOADING_FAILED on linux-amd64-gpu-rtxpro6000-latest-1

#1895 warnings.warn(f"RealQuantLinear: No real-quant GEMM found: {self}.")

#2189 megatron_generate drops VLM vision inputs during no-cache decoding

#2002 [QAD] Request for exact training configurations used in arXiv:2601.20088

#1577 generate_context_logits returns truncated logits when KV prefix cache is enabled, causing silently wrong loglikelihood evaluation

#1974 Nvidia Modelopt structured 2:4 weight sparsity showing no speed improvement compared to dense model

#2281 ONNX PTQ: no NVFP4 path for convolutional models

#2131 fold_weight crashes with AttributeError: 'NoneType' object has no attribute 'data' on Megatron models with tied word embeddings

#2230 hf_ptq.py: deprecated --auto_quantize_bits CLI flag silently no-ops instead of enabling AutoQuantize

#2220 [Feature Request] DeepSeek-V4-Flash-0731 NVFP4 checkpoint

#2204 LUT-B Support

#2160 [Bug] init_quantized_weights / --low_memory_mode exports numerically broken NVFP4 (quantizes meta tensors before weights load)

#2158 Can `auto_quant` calculate the score for `kv-cache` separately?