Search results

RuntimeError with Assertion failed: driver_result == cudaDriverEntryPointSuccess.

bugSomething isn't workingSomething isn't working

Status: Open.

#2258 In NVIDIA/TransformerEngine;

· Salieri0515 opened on Oct 10, 2025

Question about the construction of cu_seqlens_q in ring attention unit test

Status: Open.

#2257 In NVIDIA/TransformerEngine;

· Kangkang-wky opened on Oct 10, 2025

is NVFP4 not supported for rtx 50 series?

Status: Open.

#2255 In NVIDIA/TransformerEngine;

· yash3056 opened on Oct 9, 2025

Pre-built wheels

buildBuild systemBuild system

Status: Open.

#2249 In NVIDIA/TransformerEngine;

· bhatuzdaname opened on Oct 8, 2025

Discuss release of NVFP4-trained LLM on Hugging Face

Status: Open.

#2224 In NVIDIA/TransformerEngine;

· NielsRogge opened on Oct 1, 2025

[Colab Py3.12 / torch 2.8.0+cu126 / CUDA 12.6] transformer_engine.pytorch .so missing in 2.5.0 wheel (blocks Evo2 1B Base)

bugSomething isn't workingSomething isn't working

Status: Open.

#2209 In NVIDIA/TransformerEngine;

· nakane-scc opened on Sep 29, 2025

thd format is not supported with hierarchical CP implementation yet

Status: Open.

#2208 In NVIDIA/TransformerEngine;

· stormchasingg opened on Sep 28, 2025

Potential Performance Regression in v2.9.0 due to CUTLASS kernels GroupedGemm on large token sizes

bugSomething isn't workingSomething isn't working

Status: Open.

#2203 In NVIDIA/TransformerEngine;

· Skylion007 opened on Sep 25, 2025

Increase MAX_TENSOR_NUM

Status: Open.

#2189 In NVIDIA/TransformerEngine;

· mccatec opened on Sep 19, 2025

Fused attention with THD format + CP may cause a bug in backward pass, leading to NaN values during training LLM.

bugSomething isn't workingSomething isn't working

Status: Open.

#2186 In NVIDIA/TransformerEngine;

· rockychan724 opened on Sep 18, 2025

Regarding E5M2

questionFurther information is requestedFurther information is requested

Status: Open.

#2172 In NVIDIA/TransformerEngine;

· hailo231 opened on Sep 15, 2025

Questions about class AttnFuncWithCPAndQKVOA2A (torch. auto grade. Function)

Status: Open.

#2170 In NVIDIA/TransformerEngine;

· zhaoyizhan opened on Sep 11, 2025