Problemsby GlyphMistral GPU access and drivers are painful
2 posts from 2 people2d activeFading, +0 this weekunverified
Posts per day
The posts behind it
[Bug]: NCCL weight sync: packed transfer fails to unpack misaligned tensors (pack_tensors has no alignment)
### Your current environment Reproduces on CPU from current main (`vllm/distributed/weight_transfer/packed_tensor.py`); no GPU or model needed. ### 🐛 Describe the bug The packed NCCL weight sync (and the packed IPC path) can fail to unpack a buffer with: `pack_tensors` flattens...
Trainer scales the loss of heads that take `**kwargs` but return a mean loss by `world_size * gradient_accumulation_steps` (BERT family, BART/T5, ...)
Hi! I was checking DDP steps of the `Trainer` against a single process on the same global batch, and BERT sequence classification came out with `world_size` times the step of one process, on 5.20.2 (I also read the same code paths on main, `b3d3a61611`, but only ran 5.20.2). My...
Companies and products named
About this problem
Evidence
2 posts from 2 people in 2 places, about 2 a week over 7 days. Mostly on huggingface/transformers, vllm-project/vllm. Tools named alongside: Qwen, Hugging Face, vLLM, GitHub.
Frustration Frustration 0 of 3· Seen on GitHub issues
How it was grouped
Posts that state a pain and match the "gpu-access" rule about Mistral. First post Oct 26, 2026, latest Nov 1, 2026. Unverified: fewer than 3 posts, or one person or place only. Method
The same problem elsewhere
PyTorch GPU access and drivers are painful
14 posts from 14 people on GitHub issues.
NVIDIA GPU access and drivers are painful
12 posts from 12 people on GitHub issues, Mastodon, Reddit.
GPU access, drivers and CUDA versions are painful
7 posts from 7 people on GitHub issues, Reddit.
DeepSeek GPU access and drivers are painful
7 posts from 6 people on GitHub issues.
vLLM GPU access and drivers are painful
5 posts from 3 people on GitHub issues.
Claude Code GPU access and drivers are painful
3 posts from 3 people on GitHub issues.
Ollama GPU access and drivers are painful
3 posts from 3 people on Hacker News, Reddit, GitHub issues.
Chroma GPU access and drivers are painful
2 posts from 1 person on GitHub issues. Unverified.
Other Mistral problems
Mistral users keep filing bugs
Top post: "RoPE corruption due to TF32 on B200 (possibly others) when using `torch.set_float32_matmul_precision(high)`", one of 21 posts from 20 people on GitHub issues.
Mistral upgrades break working setups
9 posts from 9 people on GitHub issues.
Trouble with Mistral vibe
Top post: "Mistral Vibe CLI does not update to 3.0.0", one of 8 posts from 8 people on GitHub issues.
Mistral writes buggy code that is hard to review
8 posts from 6 people on GitHub issues.
Trouble with Mistral session
Top post: "/status session cost ignores prompt-cache discount on the Unified Harness (cached tokens billed at full input", one of 6 posts from 6 people on GitHub issues.
Mistral runs out of context on real work
4 posts from 4 people on GitHub issues.
Mistral rate limits and usage caps block work
4 posts from 4 people on GitHub issues.
Trouble with Mistral system
Top post: "Installation fails, sound system dependency?", one of 3 posts from 3 people on GitHub issues.
Trouble with Mistral harness
Top post: "`enable_otel` exports no traces under the default harness, tracing is only wired into the legacy", one of 3 posts from 3 people on GitHub issues.
Trouble with Mistral blocks
Top post: "oversized MCP image results lose their image blocks", one of 3 posts from 3 people on GitHub issues.
Trouble with Mistral return
Top post: "apply_chat_template ignores return_tensors=None for processors that default to tensors (e.g. Pixtral)", one of 3 posts from 3 people on GitHub issues.
Mistral fails to return valid structured output
2 posts from 2 people on GitHub issues. Unverified.
History