Search results

Why not just surpass 5.2 directly? Let's add the LiE enhancement plugin.

Status: Open.

#686 In karpathy/nanoGPT;

· ZengHaohei opened on Feb 18, 2026

Meet TrueNanoGPT

Status: Open.

#685 In karpathy/nanoGPT;

· nanocreator2 opened on Feb 15, 2026

Downstream works

Status: Open.

#677 In karpathy/nanoGPT;

· Mec-iS opened on Jan 6, 2026

CustomV2: A New Activation Function That Wins Under Early Stopping

Status: Open.

#675 In karpathy/nanoGPT;

· vbepipe opened on Jan 3, 2026

Feature request: Would it be possible to make a nano example of MoE models?

Status: Open.

#665 In karpathy/nanoGPT;

· jzju opened on Nov 18, 2025

[Security] Code Execution via unsafe deserialization

Status: Open.

#661 In karpathy/nanoGPT;

· edoardottt opened on Nov 13, 2025

Dataloader segmentation fault

Status: Open.

#656 In karpathy/nanoGPT;

· mtodescato opened on Oct 31, 2025

MPS + float16: GradScaler / autocast behavior in train.py

Status: Open.

#654 In karpathy/nanoGPT;

· v659 opened on Oct 23, 2025

Feature request: semantic audit during training loop

Status: Open.

#653 In karpathy/nanoGPT;

· elly99-AI opened on Oct 22, 2025

Enable Private Security Reporting

Status: Open.

#646 In karpathy/nanoGPT;

· edoardottt opened on Oct 16, 2025

Loss Distribution at 400k and 600k steps - OWT - 124M model

Status: Open.

#644 In karpathy/nanoGPT;

· MarkoKarbevski opened on Oct 11, 2025

Use Nsight (nsys profile) to profile the training in a single multi-GPU node

Status: Open.

#642 In karpathy/nanoGPT;

· Minyuluck opened on Oct 1, 2025