Owner

Lightning-AI

13 indexed repositories · View on GitHub

pytorch-lightning
Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.

Python · 31351 stars
litgpt
20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.

Python · 13670 stars
lit-llama
Implementation of the LLaMA language model based on nanoGPT. Supports flash attention, Int8 and GPTQ 4bit quantization, LoRA and LLaMA-Ad...

Python · 6082 stars
LitServe
A minimal Python framework for building custom AI inference servers with full control over logic, batching, and scaling.

Python · 3942 stars
torchmetrics
Machine learning metrics for distributed, scalable PyTorch applications.

Python · 2473 stars
lightning-thunder
PyTorch compiler that accelerates training and inference. Get built-in optimizations for performance, memory, parallelism, and easily wri...

Python · 1473 stars
litData
Speed up model training by fixing data loading.

Python · 614 stars
utilities
Common Python utilities and GitHub Actions in Lightning Ecosystem

Python · 62 stars
litAI
LLM router + minimal agent framework in one. Call any LLM API with OpenAI format. Unified billing, tools, retries, fallback, logging. Bui...

Python · 52 stars
LitModels
Save, load, host, and share AI model checkpoints without slowing down training. Host on Lightning AI or your own cloud with enterprise-gr...

Python · 40 stars
LitLogger
A minimal Python logger that tracks everything you try when building AI - metrics, prompts, models, etc, so you can see what changed and ...

Python · 37 stars
sdk
SDK and CLI for lightning.ai platform. Develop and train AI with ease.

Python · 10 stars
Lightning-multinode-templates
Multinode templates for Pytorch Lightning

Python · 9 stars

Beginner friendly Everything

Open All

bug(jobs): maxRunAttempts=1 reads back as 0 Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

Lightning-AI/sdk#152 · Opened Sep 14, 2026
`fast_queue=True` is silently ignored — the ZMQ transport is never enabled (and would be misconfigured if it were) Open

Difficulty 3/5 1-2 days Newbie friendliness 76/100

Lightning-AI/LitServe#751 · Opened Sep 14, 2026
Add an optional ultrafast-pycocotools backend for COCO mean average precision Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

Lightning-AI/torchmetrics#3499 · 1 comment · Opened Sep 10, 2026
Failed requests leave active_requests permanently incremented Open

Difficulty 2/5 1-3 hours Newbie friendliness 78/100

Lightning-AI/LitServe#747 · Opened Sep 10, 2026
OpenAISpec rejects standard required and named function tool_choice values Open

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

Lightning-AI/LitServe#745 · Opened Sep 10, 2026
First epoch after resuming from checkpoint uses sampler epoch 0 (regression from #20775, num_workers > 0) Open

bug needs triage ver: 2.6.x

Difficulty 3/5 1-2 days Newbie friendliness 76/100

Lightning-AI/pytorch-lightning#21938 · Opened Sep 8, 2026
AdjustedMutualInfoScore mishandles perfect-match limit cases Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

Lightning-AI/torchmetrics#3495 · Opened Sep 8, 2026
One-sided empty inputs are weighted inconsistently in intersection metrics Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

Lightning-AI/torchmetrics#3493 · Opened Sep 8, 2026
LLM.generate() can silently select a token excluded by top_k on CPU when sampling probabilities are float16 Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

Lightning-AI/litgpt#2323 · Opened Sep 8, 2026
Group fairness metrics can relabel missing group IDs Open

Difficulty 3/5 1-2 days Newbie friendliness 45/100

Lightning-AI/torchmetrics#3489 · Opened Sep 8, 2026
`Metric.forward()` drops previously accumulated state when batch computation raises Open

Difficulty 4/5 3-5 days Newbie friendliness 42/100

Lightning-AI/torchmetrics#3487 · Opened Sep 8, 2026
`Metric.sync_context` leaves the metric synchronized when `compute()` raises Open

Difficulty 3/5 1-2 days Newbie friendliness 65/100

Lightning-AI/torchmetrics#3486 · 1 comment · Opened Sep 8, 2026
V-measure returns 1 for independent nontrivial clusterings Open

Difficulty 2/5 1-3 hours Newbie friendliness 76/100

Lightning-AI/torchmetrics#3484 · 1 comment · Opened Sep 8, 2026
test_tensorboard_with_symlink fails on Windows without symlink privilege Open

bug logger: tensorboard tests ver: 2.7.x

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

Lightning-AI/pytorch-lightning#21932 · 1 comment · Opened Sep 6, 2026
[Fabric] FSDPStrategy ignores error_if_nonfinite in gradient clipping Open

Difficulty 3/5 1-2 days Newbie friendliness 70/100

Lightning-AI/pytorch-lightning#21928 · 1 comment · Opened Sep 3, 2026
litgpt pretrain crashes or deadlocks when a rank's val_dataloader yields zero batches (multi-GPU FSDP) Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

Lightning-AI/litgpt#2319 · 1 comment · Opened Sep 2, 2026
Publish a LitLogger release containing #66; PyPI 2026.6.25 cannot import with current lightning-sdk Open

Difficulty 3/5 1-2 days Newbie friendliness 52/100

Lightning-AI/LitLogger#70 · Opened Aug 27, 2026
Sync streaming loop fires AFTER_PREDICT twice and skips decode-request callback events Open

Difficulty 3/5 1-2 days Newbie friendliness 84/100

Lightning-AI/LitServe#735 · Opened Aug 26, 2026
Speculative decoding drops the first generated token Open

Difficulty 2/5 1-3 hours Newbie friendliness 84/100

Lightning-AI/litgpt#2307 · Opened Aug 25, 2026
Reang Open

question

Difficulty 5/5 Over a week Newbie friendliness 5/100

Lightning-AI/litgpt#2306 · Opened Aug 25, 2026
Missing objects.inv in docs Open

documentation

Difficulty 3/5 1-2 days Newbie friendliness 48/100

Lightning-AI/torchmetrics#3475 · Opened Aug 21, 2026
Python REST client can block indefinitely in TLS handshake when `request_timeout` is omitted Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

Lightning-AI/sdk#132 · Opened Aug 20, 2026
ClasswiseWrapper silently drops values for invalid labels Open

Difficulty 2/5 1-3 hours Newbie friendliness 76/100

Lightning-AI/torchmetrics#3472 · 1 comment · Opened Aug 20, 2026
NMI/AMI return 0.0 instead of 1.0 for two single-cluster labelings Open

Difficulty 2/5 1-3 hours Newbie friendliness 75/100

Lightning-AI/torchmetrics#3469 · 1 comment · Opened Aug 20, 2026
[mobile iOS Safari] Hyperlinks from the main docs page all give "Version not found" Open

docs needs triage

Difficulty 3/5 1-2 days Newbie friendliness 55/100

Lightning-AI/pytorch-lightning#21911 · 2 comments · Opened Aug 19, 2026
Stop sequence matching mishandles overlapping prefixes and buffered tokens Open

Difficulty 3/5 1-2 days Newbie friendliness 78/100

Lightning-AI/litgpt#2302 · Opened Aug 19, 2026
davies_bouldin_score returns a wrong value for off-origin float64 data (float32 buffers) Open

Difficulty 2/5 1-3 hours Newbie friendliness 55/100

Lightning-AI/torchmetrics#3467 · Opened Aug 19, 2026
Pass binary payloads to `decode_request` unchanged Open

enhancement

Difficulty 4/5 3-5 days Newbie friendliness 55/100

Lightning-AI/LitServe#732 · 1 comment · Opened Aug 18, 2026
spearman_corrcoef and kendall_rank_corrcoef silently downcast float64 to float32 Open

Difficulty 3/5 1-2 days Newbie friendliness 75/100

Lightning-AI/torchmetrics#3465 · 2 comments · Opened Aug 18, 2026
Model upload fails when version is omitted because LitLogger explicitly uses "latest" Open

Difficulty 2/5 1-3 hours Newbie friendliness 76/100

Lightning-AI/LitLogger#68 · Opened Aug 18, 2026
Checkpoint uploads truncate output prefix for stable checkpoint filenames Open

bug help wanted

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

Lightning-AI/litData#888 · Opened Aug 18, 2026
Functional nominal metrics raise RuntimeError when category labels do not start at 0 Open

bug / fix help wanted

Difficulty 3/5 1-2 days Newbie friendliness 72/100

Lightning-AI/torchmetrics#3460 · 1 comment · Opened Aug 17, 2026
Unified storage_options support for FSDPStrategy, ModelParallelStrategy, and TorchCheckpointIO Open

feature needs triage

Difficulty 4/5 3-5 days Newbie friendliness 48/100

Lightning-AI/pytorch-lightning#21905 · Opened Aug 17, 2026
Import time: `scipy.signal`, `torchvision` and `matplotlib` are imported eagerly by `import torchmetrics` Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

Lightning-AI/torchmetrics#3457 · 2 comments · Opened Aug 15, 2026
scale_batch_size runs zero batches per trial when started after training has begun Open

ver: 2.7.x

Difficulty 4/5 3-5 days Newbie friendliness 48/100

Lightning-AI/pytorch-lightning#21903 · 1 comment · Opened Aug 14, 2026
BaseFinetuning re-freezes already unfrozen modules when resuming from a checkpoint Open

ver: 2.7.x

Difficulty 3/5 1-2 days Newbie friendliness 25/100

Lightning-AI/pytorch-lightning#21901 · 1 comment · Opened Aug 14, 2026
Many time passes between epochs Open

bug needs triage ver: 2.2.x ver: 2.4.x

Difficulty 4/5 3-5 days Newbie friendliness 35/100

Lightning-AI/pytorch-lightning#21895 · 3 comments · Opened Aug 11, 2026
Concurrency Budget is Static & Estimated From Hardcoded Magic Constants Open

bug help wanted

Difficulty 5/5 Over a week Newbie friendliness 35/100

Lightning-AI/litData#870 · Opened Aug 10, 2026
`CombinedLoader(max_size_cycle)` stops immediately when one iterable is empty Open

ver: 2.7.x

Difficulty 3/5 1-2 days Newbie friendliness 52/100

Lightning-AI/pytorch-lightning#21893 · Opened Aug 9, 2026
Functional nominal metrics fail for non-zero-based category labels Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

Lightning-AI/torchmetrics#3446 · 1 comment · Opened Aug 9, 2026
Prometheus & OpenTelemetry Metrics for LitServe Open

enhancement

Difficulty 5/5 Over a week Newbie friendliness 35/100

Lightning-AI/LitServe#724 · 1 comment · Opened Aug 4, 2026
Optimize remote checkpoint loading with parallel multiprocess downloads and zero-copy mmap Open

feature needs triage

Difficulty 5/5 Over a week Newbie friendliness 35/100

Lightning-AI/pytorch-lightning#21868 · Opened Jul 30, 2026
Dependency resolving to the latest torchmetrics Open

bug / fix help wanted

Difficulty 3/5 1-2 days Newbie friendliness 50/100

Lightning-AI/torchmetrics#3439 · 1 comment · Opened Jul 30, 2026
compatibility with rich 15.0.0 Open

bug progress bar: rich ver: 2.6.x

Difficulty 3/5 1-2 days Newbie friendliness 58/100

Lightning-AI/pytorch-lightning#21865 · Opened Jul 29, 2026
functional/text/eed.py still carries the RWTH non-commercial licence (same class as #2665) Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

Lightning-AI/torchmetrics#3435 · 1 comment · Opened Jul 24, 2026
sync_dist mean-reduced epoch metrics are inflated when ranks log unequal batch counts (integer batch-size accumulator floors during mean-sync) Open

bug logger ver: 2.6.x ver: 2.7.x

Difficulty 2/5 1-3 hours Newbie friendliness 25/100

Lightning-AI/pytorch-lightning#21859 · 2 comments · Opened Jul 23, 2026
nan_stategy=disable dropping weights Open

bug / fix help wanted

Difficulty 3/5 1-2 days Newbie friendliness 35/100

Lightning-AI/torchmetrics#3434 · 1 comment · Opened Jul 23, 2026
ModelCheckpoint: manual-optimization pre-step snapshot is not applied when the save is deferred to on_validation_end Open

callback: model checkpoint ver: 2.6.x ver: 2.7.x

Difficulty 3/5 1-2 days Newbie friendliness 74/100

Lightning-AI/pytorch-lightning#21842 · 3 comments · Opened Jul 17, 2026
Mixed precision autocast is disabled when running FSDP on CPU Open

bug ver: 2.5.x ver: 2.6.x

Difficulty 3/5 1-2 days Newbie friendliness 68/100

Lightning-AI/pytorch-lightning#21815 · Opened Jul 9, 2026
Device mesh initialization fails when running FSDP on CPU Open

bug ver: 2.5.x ver: 2.6.x

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

Lightning-AI/pytorch-lightning#21814 · Opened Jul 9, 2026

Showing the newest 100

This page lists what was indexed most recently. Advanced filter has the whole inventory, narrowed by language, difficulty and how long a task takes.

Open advanced filter

Lightning-AI

Showing the newest 100

Get new issues in your inbox