Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / Lightning-AI/lightning issues and pull requests

#18624 - Update numpy requirement from <1.25.3,>=1.17.2 to >=1.17.2,<1.27 in /requirements

Pull Request - State: closed - Opened by dependabot[bot] about 1 year ago - 2 comments
Labels: ready, fabric, pl

#18624 - Update numpy requirement from <1.25.3,>=1.17.2 to >=1.17.2,<1.27 in /requirements

Pull Request - State: open - Opened by dependabot[bot] about 1 year ago - 2 comments
Labels: ready, fabric, pl

#18624 - Update numpy requirement from <1.25.3,>=1.17.2 to >=1.17.2,<1.27 in /requirements

Pull Request - State: open - Opened by dependabot[bot] about 1 year ago - 2 comments
Labels: ready, fabric, pl

#18623 - Update deepdiff requirement from <6.3.2,>=5.7.0 to >=5.7.0,<6.6 in /requirements

Pull Request - State: closed - Opened by dependabot[bot] about 1 year ago - 1 comment
Labels: ready, app

#18623 - Update deepdiff requirement from <6.3.2,>=5.7.0 to >=5.7.0,<6.6 in /requirements

Pull Request - State: open - Opened by dependabot[bot] about 1 year ago - 1 comment
Labels: app

#18623 - Update deepdiff requirement from <6.3.2,>=5.7.0 to >=5.7.0,<6.6 in /requirements

Pull Request - State: open - Opened by dependabot[bot] about 1 year ago - 1 comment
Labels: app

#18621 - allow fractional values for `every_n_train_steps` in `ModelCheckpoint`

Issue - State: closed - Opened by MF-FOOM about 1 year ago - 1 comment
Labels: feature, needs triage

#18620 - on_epoch=True reduction of low precision types (bf16, etc) results in very inaccurate metrics

Issue - State: closed - Opened by MF-FOOM about 1 year ago - 7 comments
Labels: bug, help wanted, logging, ver: 2.1.x

#18619 - Always pass the correct batch index to the automatic optimization loop

Pull Request - State: open - Opened by carmocca about 1 year ago - 2 comments
Labels: bug, loops, pl

#18619 - Always pass the correct batch index to the automatic optimization loop

Pull Request - State: closed - Opened by carmocca about 1 year ago - 2 comments
Labels: bug, ready, loops, pl

#18618 - Enable launching via torchrun in slurm environment

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 11 comments
Labels: feature, ready, environment: slurm, fabric, pl

#18617 - ci/groupcheck: fix TPU parameters [pytorch]

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ci, priority: 1

#18616 - LightningCLI: incorrect default value of kwarg used

Issue - State: open - Opened by adamjstewart about 1 year ago - 11 comments
Labels: bug, 3rd party, lightningcli, ver: 2.0.x

#18615 - rtfd: fix building with stable/latest

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready, docs, priority: 0

#18614 - [docs] explain how to use `torchrun` in a SLURM environment

Pull Request - State: open - Opened by stas00 about 1 year ago - 3 comments
Labels: docs, community, pl

#18613 - FSDP `ignored_modules` not moved to device automatically

Issue - State: closed - Opened by Sumith1896 about 1 year ago - 3 comments
Labels: question, strategy: fsdp, ver: 2.1.x

#18612 - Incorrect batch dtype at `on_train_batch_start/end` using 16-mixed / FSDP

Issue - State: closed - Opened by Sumith1896 about 1 year ago - 5 comments
Labels: bug, strategy: fsdp, ver: 2.1.x

#18611 - adding `make docs-{app,fabric,pytorch}`

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready

#18610 - Remove outdated num_workers warnings

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: pl

#18609 - ci/docs: fetch assets only for deployment, omit PR

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready, ci, pl

#18607 - Fabric.all_gather should support concatenation

Issue - State: open - Opened by cemde about 1 year ago - 1 comment
Labels: feature, discussion, fabric

#18606 - docs: better message for download extensions docs

Pull Request - State: closed - Opened by Borda about 1 year ago - 2 comments
Labels: ready, ci

#18605 - precommit: unify formatting with prettier +TPU

Pull Request - State: closed - Opened by Borda about 1 year ago - 4 comments
Labels: ready, ci, fabric, app, pl

#18604 - replace `tmpdir` by `tmp_path` in tests_data

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready, refactor, tests

#18603 - docs: 3/3 enable Sphinx nitpicky [app]

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready, docs, app

#18602 - docs: 2/3 enable Sphinx nitpicky [pytorch] part 2/n

Pull Request - State: closed - Opened by Borda about 1 year ago - 2 comments
Labels: ready, docs, app, pl

#18601 - W&B `dir` flag doesn't work.

Issue - State: open - Opened by cemde about 1 year ago - 4 comments
Labels: bug, help wanted, logger: wandb, ver: 2.0.x

#18600 - Drop the "global" prefix in the seeding info message

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: feature, ready, fabric

#18599 - "Global" wording in seeding message can be confusing

Issue - State: closed - Opened by awaelchli about 1 year ago
Labels: feature

#18598 - Input validation for `num_nodes` argument

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: feature, ready, fabric, trainer: connector, pl

#18597 - Trainer does not work when accelerator="mps" (but works fine if using accelerator="cpu" or "gpu")

Issue - State: open - Opened by plannaAlain about 1 year ago - 1 comment
Labels: bug, 3rd party, accelerator: mps, ver: 2.0.x

#18596 - Fixed confusing exception message in Tuner

Pull Request - State: closed - Opened by sameertantry about 1 year ago - 1 comment
Labels: bug, ready, tuner, community, pl

#18595 - Validation metrics not available when resuming training from checkpoint

Issue - State: open - Opened by TreeMage about 1 year ago - 1 comment
Labels: bug, needs triage, ver: 2.0.x

#18594 - Update `LightningModule.optimizers()` docs

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, docs, optimization, pl

#18593 - Force consistent it/s display in TQDM progress bar

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: feature, ready, progress bar: tqdm, pl

#18592 - Remove `process_group` property

Pull Request - State: closed - Opened by carmocca about 1 year ago - 2 comments
Labels: ready, refactor, strategy: fsdp, pl

#18592 - Remove `process_group` property

Pull Request - State: closed - Opened by carmocca about 1 year ago - 2 comments
Labels: ready, refactor, strategy: fsdp, pl

#18591 - Improve the suggested `num_workers` warning

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: feature, ready, fabric, performance, pl

#18591 - Improve the suggested `num_workers` warning

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: feature, ready, fabric, performance, pl

#18590 - ModelCheckpoint saves multiple checkpoints when trainer is using DDP

Issue - State: open - Opened by sebquetin about 1 year ago
Labels: bug, callback: model checkpoint, strategy: ddp, ver: 2.0.x

#18590 - ModelCheckpoint saves multiple checkpoints when trainer is using DDP

Issue - State: open - Opened by sebquetin about 1 year ago
Labels: bug, needs triage, ver: 2.0.x

#18589 - Can't seem to change distributed backend to gloo on Windows

Issue - State: closed - Opened by amansingh427 about 1 year ago - 2 comments
Labels: question, strategy: ddp, ver: 2.1.x

#18589 - Can't seem to change distributed backend to gloo on Windows

Issue - State: closed - Opened by amansingh427 about 1 year ago - 2 comments
Labels: question, strategy: ddp, ver: 2.1.x

#18588 - Resuming training with custom scheduler loads wrong learning rate

Issue - State: closed - Opened by rob-hen about 1 year ago - 4 comments
Labels: bug, optimization, lr scheduler, ver: 2.0.x

#18588 - Resuming training with custom scheduler loads wrong learning rate

Issue - State: open - Opened by rob-hen about 1 year ago
Labels: bug, needs triage, ver: 2.0.x

#18587 - load_from_checkpoint uses default parameters instead of supplied argument (size mismatch)

Issue - State: closed - Opened by B-lanc about 1 year ago - 3 comments
Labels: question, ver: 2.0.x

#18587 - load_from_checkpoint uses default parameters instead of supplied argument (size mismatch)

Issue - State: open - Opened by B-lanc about 1 year ago - 2 comments
Labels: question, ver: 2.0.x

#18586 - Utility function to check shared filesystem

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: feature, ready, distributed, fabric, pl

#18586 - Utility function to check shared filesystem

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: feature, ready, distributed, fabric, pl

#18585 - Correct reload_dataloaders_every_n_epochs docstring

Pull Request - State: closed - Opened by f0k about 1 year ago
Labels: ready, docs, trainer: argument, pl

#18584 - `CombinedLoader` takes a long time when `num_workers > 0`

Issue - State: open - Opened by johnathanchiu about 1 year ago
Labels: bug, help wanted, performance, ver: 2.0.x, repro needed

#18583 - Avoid passing process group to enable FSDP's hybrid-shard

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: feature, ready, fabric, strategy: fsdp, pl

#18583 - Avoid passing process group to enable FSDP's hybrid-shard

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: feature, ready, fabric, strategy: fsdp, pl

#18582 - Set up dataloaders in k-fold cross validation example

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: bug, example, ready, fabric

#18581 - DDP communication hooks with Fabric

Issue - State: closed - Opened by patchmeifyoucan about 1 year ago - 2 comments
Labels: help wanted, docs

#18581 - DDP communication hooks with Fabric

Issue - State: closed - Opened by patchmeifyoucan about 1 year ago - 2 comments
Labels: help wanted, docs

#18580 - Lightning fabric k-fold example has device issue

Issue - State: closed - Opened by fealty94 about 1 year ago - 1 comment
Labels: bug, example, fabric, ver: 2.1.x

#18579 - Bump docker/build-push-action from 4 to 5

Pull Request - State: closed - Opened by dependabot[bot] about 1 year ago
Labels: ready, ci

#18578 - Bump docker/setup-buildx-action from 2 to 3

Pull Request - State: closed - Opened by dependabot[bot] about 1 year ago - 1 comment
Labels: ready, ci

#18577 - Bump docker/login-action from 2 to 3

Pull Request - State: closed - Opened by dependabot[bot] about 1 year ago
Labels: ready, ci

#18576 - Add multiple ModelCheckpoint callbacks support to WandbLogger and adjust model file namings

Issue - State: open - Opened by royvelich about 1 year ago
Labels: feature, logger: wandb

#18575 - Move `_KINETO_AVAILABLE` check to profiler

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, refactor, pl, fun

#18574 - Remove confusing TensorBoardLogger doctest

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, docs, logger: tensorboard, pl

#18573 - Lazily import dependencies for NeptuneLogger

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, refactor, performance, pl, fun

#18570 - Schedulers don't work

Issue - State: closed - Opened by ririya about 1 year ago - 10 comments
Labels: bug, waiting on author, optimization, ver: 2.0.x

#18569 - docs: update chlog after `2.0.8` and `2.0.9`

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready, docs, fabric, app, pl

#18568 - Export to ONNX

Issue - State: closed - Opened by MarcoPrassel about 1 year ago - 4 comments
Labels: question, waiting on author, ver: 2.1.x

#18567 - Avoid rewriting the metrics file in CSVLogger unless necessary

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: bug, ready, priority: 0, fabric, performance, pl

#18566 - Wokarounds to avoid dynamo graph breaks with common precision settings

Pull Request - State: closed - Opened by carmocca about 1 year ago - 1 comment
Labels: feature, fabric, torch.compile

#18565 - Conda package lightning (v2.0.9) appears to be corrupt when installing on Mac Darwin

Issue - State: closed - Opened by FlorisCalkoen about 1 year ago - 3 comments
Labels: bug, priority: 1, release, ver: 2.0.x

#18564 - xla_fsdp support in lightning

Issue - State: closed - Opened by yihui-he about 1 year ago - 1 comment
Labels: duplicate, feature

#18563 - Pip warns about non-standard dependency specifier

Issue - State: open - Opened by mowangmodi about 1 year ago - 4 comments
Labels: bug, release, dependencies, ver: 1.7.x

#18562 - Support hybrid-shard in FSDP

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: feature, fabric, strategy: fsdp, pl

#18561 - Optimize import paths for optional dependencies

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, docs, ci, refactor, fabric, pl, fun

#18560 - Adding test for legacy checkpoint created with 2.0.9

Pull Request - State: closed - Opened by pl-ghost about 1 year ago - 1 comment
Labels: ready, checkpointing, tests, pl

#18559 - Enable Quantization with `BitsandbytesQuantization` Plugin

Pull Request - State: closed - Opened by JustinGoheen about 1 year ago
Labels: fabric

#18558 - Redirect users from neptune-client to the neptune package

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 4 comments
Labels: 3rd party, logger: neptune, pl

#18557 - [Fabric] Replace `@contextlib.contextmanager`

Pull Request - State: closed - Opened by carmocca about 1 year ago - 2 comments
Labels: ready, refactor, breaking change, fabric

#18556 - automatic_optimization=False disables checkpoint saving

Issue - State: closed - Opened by petargyurov about 1 year ago - 4 comments
Labels: question, checkpointing, optimization, ver: 2.0.x

#18555 - cannot find neptune.new.utils with `Neptune-client==0.16.3`

Issue - State: open - Opened by filipporemonato about 1 year ago - 9 comments
Labels: bug, help wanted, logger: neptune, ver: 2.0.x

#18555 - cannot find neptune.new.utils with `Neptune-client==0.16.3`

Issue - State: open - Opened by filipporemonato about 1 year ago - 9 comments
Labels: bug, help wanted, logger: neptune, ver: 2.0.x

#18554 - Remove reference to training in checkpoint loading error message

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 1 comment
Labels: ready, fabric, pl

#18554 - Remove reference to training in checkpoint loading error message

Pull Request - State: open - Opened by awaelchli about 1 year ago - 1 comment
Labels: ready, fabric, pl

#18553 - ci: add description how to clean machines

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready, ci

#18553 - ci: add description how to clean machines

Pull Request - State: closed - Opened by Borda about 1 year ago - 1 comment
Labels: ready, ci

#18552 - Support FSDP's hybrid-shard

Issue - State: closed - Opened by awaelchli about 1 year ago - 3 comments
Labels: feature, strategy: fsdp

#18552 - Support FSDP's hybrid-shard

Issue - State: closed - Opened by awaelchli about 1 year ago - 3 comments
Labels: feature, strategy: fsdp

#18551 - Prefer local imports for optional dependencies

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, refactor, performance, pl, fun

#18551 - Prefer local imports for optional dependencies

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, refactor, performance, pl, fun

#18550 - Avoid warning about logging interval for fast dev run

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: bug, ready, logging, trainer: fit, pl

#18550 - Avoid warning about logging interval for fast dev run

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: bug, ready, logging, trainer: fit, pl

#18549 - Refactor NeptuneLogger tests from unittest to pytest

Pull Request - State: open - Opened by awaelchli about 1 year ago - 2 comments
Labels: refactor, tests, logger: neptune, fun

#18549 - Refactor NeptuneLogger tests from unittest to pytest

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, refactor, tests, logger: neptune, fun

#18549 - Refactor NeptuneLogger tests from unittest to pytest

Pull Request - State: closed - Opened by awaelchli about 1 year ago - 2 comments
Labels: ready, refactor, tests, logger: neptune, fun

#18548 - Fix `Trainer`'s `log_dir` method for `CSVLogger`

Pull Request - State: closed - Opened by ioangatop about 1 year ago - 1 comment
Labels: bug, ready, logger: csv, community, pl

#18547 - `log_dir` in `Trainer` is wrong for `CSVLogger`

Issue - State: closed - Opened by ioangatop about 1 year ago - 3 comments
Labels: bug, duplicate, logger: csv, ver: 2.1.x

#18546 - Bug with add_subclass_arguments() or ParamData() (or smth else?)

Issue - State: closed - Opened by usernameisntavailableble about 1 year ago - 7 comments
Labels: bug, lightningcli

#18545 - How to unit test LightningCLI without UserWarning?

Issue - State: open - Opened by adamjstewart about 1 year ago - 8 comments
Labels: bug, lightningcli, ver: 2.0.x

#18544 - Replace LightningClient with import from lightning_cloud

Pull Request - State: closed - Opened by justusschock about 1 year ago - 1 comment
Labels: ready, app