Search results

Regarding the Hadamard transform accuracy drop for nvfp4 with RTN in the paper

Status: Open.

#19 In IST-DASLab/FP-Quant;

· TheTinyTeddy opened on Dec 4, 2025

When using transformers.set_seed() instead of src.utils.common_utils.fix_seed(), W4A4 openllm chat-template–related task accuracies drop to 0, but bf16 accuracies are fine.

Status: Open.

#17 In IST-DASLab/FP-Quant;

· WeiweiZhang1 opened on Nov 25, 2025

BUG: Duplicate bias addition in PseudoQuant4x16NoMasterFn causes incorrect inference output

Status: Open.

#16 In IST-DASLab/FP-Quant;

· manyizhang opened on Nov 22, 2025

Realquant checkpoint vllm compatibility bug

Status: Open.

#15 In IST-DASLab/FP-Quant;

· songsm921 opened on Nov 20, 2025

Evaluation Extremely Slow After Quantization

Status: Open.

#8 In IST-DASLab/FP-Quant;

· slchenchn opened on Sep 22, 2025

Environmens

Status: Open.

#4 In IST-DASLab/FP-Quant;

· Aaronhuang-778 opened on Aug 18, 2025

Environment & Dependency Information

Status: Open.

#2 In IST-DASLab/FP-Quant;

· Superone77 opened on Jul 26, 2025