Repository Issue Activity (beta)

#846 match_targets orders overlapping regex targets lexicographically, not by specificity

#847 match_targets orders overlapping regex targets lexicographically, not by specificity

#832 Saved Qwen3-Omni GPTQ W4A16 checkpoints contain NaN/Inf weight_scale tensors

#766 [Performance] Speed up quantization with dedicated kernels

#646 Write a guide for quantizing models using frameworks outside of Compressed-Tensors quantization

#795 [bug] broadcast QuantizationFormat is incorrect for fp8 input-activation only

#645 [Refactor] Remove `g_idx == -1` checks

#796 [Offloading] Better memory capacity estimates for shared cpu mmap

#656 [Offload] Investigate using torch multiprocessing module for distributed offloading

#758 [Bug] Unreachable GROUP+dynamic branch in QuantizationScheme.validate_model_after

#712 ignore= is honored at calibration but NOT at save_pretrained_wrapper — modules excluded from quantization still get RTN-packed at save

#762 [Bug] block_structure accepts non-positive dimensions

#746 Multi-XPU tests intermittently hang or time out across repeated torchrun invocations

#719 Humming RFC

#754 [Bug] block_structure string format skips length validation in QuantizationArgs

#734 CT cast_to_fp4 torch recompile issues

#638 [Bug] Disk Offloading and Deletion of Shared Tensors

#671 [Feature] Distributed model decompression

#647 [Performance] Speed up model loading by desynchronizing sync ops in `from_accelerate`

#652 Kimi-K2.5: AttributeError: 'CompressedLinear' object has no attribute 'weight'

#711 [Bug] `dispatch_with_map` raises `AttributeError: '0' is not an nn.Module` on sharded models (sharded MoE expert ModuleList)

#573 [CI] Have mergify clean up previous messages

#695 match_named_modules O(N×M) causes multi-minute silent load tail on fine-grained MoE checkpoints (Qwen3-30B-A3B-AWQ, ~18K Linear modules)

#677 [convert] Add Converter for AutoAWQ format

#618 Error running Qwen-3-8B-NVFP4