Lightning-AI/pytorch-lightning

AdvancedProfiler: ValueError: Attempting to stop recording an action (run_test_evaluation) which was never started. 19 days ago

bug help wanted good first issue priority: 1

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

support len(datamodule) 20 days ago

feature help wanted good first issue data handling

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Use :emphasize-lines: in sphinx docs to highlight code. 21 days ago

good first issue docs priority: 1

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Update the docformatter pre-commit action 22 days ago

good first issue ci

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Allow obtaining num_nodes from ClusterEnvironment 28 days ago

feature help wanted good first issue environment

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

[Fabric Lightning] Named barriers about 1 month ago

feature help wanted distributed

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Add checks for model spec and matching output values in `to_onnx()` method about 2 months ago

feature help wanted good first issue

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Improved control of device stats callbacks 3 months ago

help wanted refactor design callback: device stats

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Pruning callback causes GPU memory leak when used iteratively 3 months ago

bug good first issue waiting on author priority: 1

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Documentation: writing custom samplers compatible with multi GPU training 4 months ago

help wanted docs

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Adopt `typing_extensions.Override` 4 months ago

feature help wanted good first issue

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Make lazy initialization in plugins more robust 5 months ago

feature help wanted good first issue let's do it! distributed

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Incorrect batch progress saved in checkpoint at every_n_train_steps 5 months ago

bug help wanted loops ver: 1.9.x ver: 2.1.x repro needed

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

val_check_interval check not bypassed when limit_val_batches=0 5 months ago

bug good first issue trainer trainer: validate ver: 2.5.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Simple documentation of automatic optimization with learning rate scheduler 5 months ago

good first issue docs

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Support all iterator modes for fit/validate/test/predict 6 months ago

feature help wanted data handling pl

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Broken URLs on NeptuneLogger documentation page 7 months ago

good first issue docs logger: neptune

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Add `enable_device_summary` flag to disable device printout 7 months ago

feature good first issue callback trainer: argument

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Potential off by 1 error when resuming training of mid-epoch checkpoint 8 months ago

bug help wanted loops ver: 2.1.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Schedule in PyTorchProfiler doesn't work 8 months ago

bug help wanted profiler

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Confusing recommendation to use sync_dist=True even with TorchMetrics 8 months ago

bug help wanted logging ver: 2.2.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

trainer.fit render bug 9 months ago

good first issue docs trainer: fit

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

`RichProgressBar` in v1.6 is slower than v1.5 9 months ago

bug help wanted priority: 1 progress bar: rich performance pl

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

ModelPruning callback logs wrong sparsity 10 months ago

AI Summary: The `ModelPruning` callback incorrectly logs the overall sparsity when using verbosity levels 1 or 2, especially for layers with multiple parameters being pruned. The reported overall sparsity is lower than the actual applied amount, and the issue is traced to a potential double-counting of layers in the calculation at line 346 of the callback's code.

Complexity: 2/5

bug help wanted priority: 2 callback: pruning

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

How to use Webdataset in DDP setting? ValueError: you need to add an explicit nodesplitter to your input pipeline for multi-node training 11 months ago

help wanted docs ver: 2.2.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Using Stochastic Weight Averaging (SWA) and LearningRateFinder simultaneously can cause issues: 11 months ago

bug help wanted tuner callback: swa ver: 2.2.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Tuner's `scale_batch_size` edgecase resulting in a bug 11 months ago

bug help wanted tuner ver: 2.0.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

`save_hyperparameters` does not save the selected` learning_rate` when `auto_lr_find` is used 11 months ago

bug help wanted logger tuner

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Support AMD GPUs with the MPS backend 11 months ago

help wanted question accelerator: mps

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Resuming from checkpoint, mid epoch gives a very distorted time estimate 12 months ago

bug help wanted priority: 1 progress bar: tqdm ver: 2.0.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

ModelCheckpoint could not find key in returned metrics 12 months ago

bug help wanted callback: model checkpoint ver: 2.1.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Logging in `on_test_epoch_end` with multiple dataloaders 12 months ago

bug help wanted ver: 2.5.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

compatibility patch for BnB >=0.46 about 1 year ago

AI Summary: The task requires creating a compatibility patch for PyTorch Lightning (v2.5) to ensure it works correctly with the bitsandbytes library (versions >= 0.46). The issue stems from a change in the `bitsandbytes.functional` API, specifically the `int8_double_quant` function, which now returns a different number of values than previously expected. The patch needs to adjust the unpacking logic in PyTorch Lightning's code to accommodate this change.

Complexity: 4/5

bug help wanted good first issue 3rd party ver: 2.5.x

Lightning-AI/pytorch-lightning

31.2K

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

Python

#ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch

Open Issues Need Help