Skip to content

Pull requests: karpathy/nanochat

New pull request New
44 Open 307 Closed
44 Open 307 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Small fixes docs Improvements or additions to documentation
#595 opened Mar 6, 2026 by svlandeg • Draft
Add upload to hf hub at the end of training feature New feature or request
#588 opened Mar 5, 2026 by Randl Loading…
2
fix: correct CSV extraction in scaling_laws.sh potential_bug Needs investigation/confirmation whether or not it's a bug scripts Edits in the bash scripts
#580 opened Feb 28, 2026 by geopti Loading…
fix: guard target-param-data-ratio against zero to avoid ZeroDivision… code robustness
#579 opened Feb 28, 2026 by suraj-self Loading…
1
1
Validate --source so invalid values don’t raise KeyError code robustness
#576 opened Feb 28, 2026 by dhunganapramod9 Loading…
1
fix(tokenizer): Avoid IndexError when conversation has only a system message code robustness waiting Waiting for user feedback/action
#574 opened Feb 28, 2026 by dhunganapramod9 Loading…
1
fix(chat_web): validate ,--num-gpus against available GPUs for clear s… UX
#573 opened Feb 28, 2026 by dhunganapramod9 Loading…
1
Remove inherited parameters from scripts code robustness scripts Edits in the bash scripts suggest/merge
#568 opened Feb 26, 2026 by svlandeg Loading…
fix: add retry with exponential backoff to download_file_with_lock code robustness
#555 opened Feb 22, 2026 by icenfly Loading…
1
fix: clamp few-shot sample size to prevent ValueError in core_eval code robustness suggest/review
#551 opened Feb 21, 2026 by suraj-self Loading…
1
2
dataloader: reuse cropped remainders to reduce token waste ~35% → ~23% improvement waiting Waiting for user feedback/action
#544 opened Feb 19, 2026 by handsomesun Loading…
4
MoE feature New feature or request
#543 opened Feb 19, 2026 by karpathy • Draft
Multi platform CI - Github Workflow tests v4 tests
#536 opened Feb 16, 2026 by SermetPekin Loading…
10
Improve error handling for batch size alignment suggest/merge UX
#533 opened Feb 15, 2026 by suraj-self Loading…
1
3
Speed up CORE metric evaluation with batched GPU inference and cross-call caching feature New feature or request
#526 opened Feb 13, 2026 by AutomaticHourglass Loading…
feat: Enhanced platform detection (ARM64, RISC-V, device models) feature New feature or request
#522 opened Feb 11, 2026 by RalphBigBear Loading…
docs: Add Q-Lite deployment guide + edge device integration docs Improvements or additions to documentation
#521 opened Feb 11, 2026 by RalphBigBear Loading…
Fix hellaswag memory leak and extract cleanup config improvement waiting Waiting for user feedback/action
#519 opened Feb 9, 2026 by haltingstate Loading…
4 tasks
1
@svlandeg
2
use 1-sqrt warmdown shape for LR schedule potential_improvement
#513 opened Feb 8, 2026 by spjosyula Loading…
2
add multi-node training script for distributed training setup feature New feature or request scripts Edits in the bash scripts
#511 opened Feb 6, 2026 by kunwar-vikrant Loading…
2
Change default sliding window pattern to the recommended "L" when FA3 is not available suggest/merge UX
#509 opened Feb 6, 2026 by ddudek Loading…
1
LRU Cache for _get_sliding_window_mask and convert tensors once in dataloader improvement suggest/review
#501 opened Feb 4, 2026 by Narden91 Loading…
Speedrun Submission: MuonH Optimizer improvement leaderboard candidate suggest/review
#498 opened Feb 4, 2026 by dangxingyu Loading…
8
Add TPU/XLA support and local smoke tests feature New feature or request waiting Waiting for user feedback/action
#495 opened Feb 3, 2026 by caseyonit Loading…
1
Fix rotation calculations in apply_rotary_emb function improvement suggest/merge
#489 opened Feb 1, 2026 by qdrk Loading…
3
ProTip! no:milestone will show everything without a milestone.