Skip to content
Problems.

NVIDIA users keep filing bugs

MLOps6 posts from 6 people6d active+11 post in the last 7 days, 1 the 7 days before (steady)

Posts per day

Posts per day6 posts, Aug 27 to Sep 25

The posts behind it

6, newest first
PostDate
Compile bug: Build error with sm_70 on Windows 11...steps to reproduce Compiling with CUDA sm_70 on Windows 11 fails (V100, CUDA 12.9). The failing target is fattn.cu.obj; the error is reported in fattn-mma-f16.cuh, which is included by fattn.cu. load_ldmatrix has no matching overload for the Volta tile tile . The 5-argument...ggml-org/llama.cpplingyezhixingSep 214 days ago
How to use onnxruntime on same GPU as graphics api i.e opengl...order to prepare frame for rendering. Which leads to application having low frame rate.. Is there a way to tell if a GPU will handle context switching between ML and graphics or is the only option to optimise the underlying model ops to run faster, I'm specifically referring to...Stack Overflow [tensorflow]james kerrSep 169 days ago
`cudagraph_skip_dynamic_graphs=True` still re-records one cudagraph per input size### 🐛 Describe the bug Hi. It seems `torch._inductor.config.triton.cudagraph_skip_dynamic_graphs` does not always do what it advertises for a node that consumes a dynamically-shaped tensor but produces a statically-shaped output. The config comment says: With it set to `True` I...pytorch/pytorchrmcgibboSep 92 weeks ago
[1.36] Kubelet retries DRA health streams for plugins that do not advertise health support...retrying until the plugin re-registers. This is the behavior in Kubernetes 1.37. ### How can we reproduce it (as minimally and precisely as possible)? This was reproduced with a kind v1.36.1 cluster using the head of:...kubernetes/kubernetesharcheSep 43 weeks ago
Bug about transformers==5.16.1 while loading 'nvidia/audio-flamingo-next-hf' model.### System Info ### Who can help? _No response_ ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [ ] My own task or dataset (give details below) ###...huggingface/transformersJinbo-HuAug 284 weeks ago
DFlash2 is not working with `--split-mode tensor`Assertion is failed on llama-server startup: Minimal command options to reproduce assert: It is working with `--split-mode layer` (no assertion on startup) More information: - Windows 10 - Cuda - RTX 5060 Ti 16 Gb x2ggml-org/llama.cppart-denAug 274 weeks ago

Companies and products named

Company or productPosts naming it
NVIDIA6
llama.cpp2
TensorFlow1
Hugging Face1
PyTorch1
GitHub1
About this problem

Evidence

6 posts from 6 people in 5 places, about 2 a week over 26 days. Mostly on ggml-org/llama.cpp, Stack Overflow [tensorflow], huggingface/transformers. Tools named alongside: llama.cpp, TensorFlow, Hugging Face, PyTorch.

Frustration Frustration 0 of 3· Seen on GitHub issues, Stack Overflow

How it was grouped

Posts about NVIDIA that state a pain but match no known issue yet. First post Aug 27, 2026, latest Sep 21, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method

History

  • 2026-09-25 Statement: NVIDIA users keep asking for help to NVIDIA users keep filing bugs
  • 2026-09-25 Added: NVIDIA users keep asking for help (5 posts)

Rising problems by email

Mondays: the problems in data, tech and AI that grew fastest that week.

Double opt-in. Unsubscribe any time.