pytorch-lightning · Issues· 1085 open
Open on GitHubLocally synced open issues (discussions stay on GitHub)
- #18985
BatchSizeFinder throws KeyError: 'limit_eval_batches'
bugduplicatehelp wantedtunerver: 2.1.xrepro neededUpdated Sep 9, 2026 - #21895
Many time passes between epochs
bugneeds triagever: 2.2.xver: 2.4.xUpdated Sep 8, 2026 - #21938
First epoch after resuming from checkpoint uses sampler epoch 0 (regression from #20775, num_workers > 0)
bugneeds triagever: 2.6.xUpdated Sep 8, 2026 - #21932
test_tensorboard_with_symlink fails on Windows without symlink privilege
buglogger: tensorboardtestsver: 2.7.xUpdated Sep 8, 2026 - #16431
`self.log` raised error when number of dataloader is not consistent
bughelp wantedloggingUpdated Sep 5, 2026 - #21928
[Fabric] FSDPStrategy ignores error_if_nonfinite in gradient clipping
Updated Sep 3, 2026 - #12119
Use :emphasize-lines: in sphinx docs to highlight code.
good first issuedocspriority: 1Updated Sep 2, 2026 - #5965
support len(datamodule)
featurehelp wantedgood first issuedata handlingUpdated Sep 2, 2026 - #17544
Stepwise LR-Scheduler not working across epochs
bughelp wantedlr schedulerver: 2.0.xUpdated Aug 30, 2026 - #17281
Strict warning about using optimizers manually would be needed in the docs
help wanteddocsUpdated Aug 26, 2026 - #21911
[mobile iOS Safari] Hyperlinks from the main docs page all give "Version not found"
docsneeds triageUpdated Aug 20, 2026 - #21905
Unified storage_options support for FSDPStrategy, ModelParallelStrategy, and TorchCheckpointIO
featureneeds triageUpdated Aug 17, 2026 - #21903
scale_batch_size runs zero batches per trial when started after training has begun
ver: 2.7.xUpdated Aug 16, 2026 - #21901
BaseFinetuning re-freezes already unfrozen modules when resuming from a checkpoint
ver: 2.7.xUpdated Aug 14, 2026 - #21865
compatibility with rich 15.0.0
bugprogress bar: richver: 2.6.xUpdated Aug 12, 2026