Lightning-AI/pytorch-lightning

Issues

• Stream outputs from Trainer.predict()

  #20334 opened 3 months ago by Turakar
  1

• _atomic_save with transaction cause "Invalid cross-device link" error

  #20270 opened 4 months ago by RichardChe
  3

• Mid-epoch resume causes a single unwanted validation step (which is not a sanity check)

  #20288 opened 3 months ago by Youyoun
  2

• Validation is incorrectly run on resume

  #20277 opened 3 months ago by PiotrDabkowski
  3

• Incorrect URI Prefix Stripping in MLflowLogger

  #20279 opened a month ago by awindmann
  0

• Custom Pytorch BatchSampler does not work well with pytorch lightning

  #20326 opened a month ago by dadwadw233
  0

• SLURM resubmission crashes because of multiprocessing error

  #20280 opened 3 months ago by antonzub99
  2

• `LightningCLI` doesn't fail when `config.yaml` contains invalid arguments

  #20337 opened 2 months ago by adosar
  4

• Save save_hyperparameters no longer respects linked arguments.

  #20311 opened 3 months ago by Erotemic
  4

• Model Checkpointing + FSDP causes Cuda OOM

  #20312 opened 3 months ago by profPlum
  1

• Import error on shutdown/KeyboardInterrupt if ran from Jupyter Lab notebook cell

  #20317 opened 3 months ago by asigalov61
  2

• Can't resume automatically a job, ckpt_path="hpc" throws ValueError from the start

  #20347 opened 2 months ago by F-Barto
  1

• Everything prints fine, but the loss doesn't descent

  #20344 opened 2 months ago by 2catycm
  6

• LearningRateFinder creates errors for schedulers in `val` stage

  #20355 opened 2 months ago by DeanLa
  2

• Gradient accumulation calcluation may be incorrect

  #20350 opened 2 months ago by tyler-rt
  0

• Add support S3 as a storage option for profiling results

  #20348 opened 2 months ago by kimminw00
  0

• tensorboard step and self.global_step do not correspond under accumulate_grad

  #20346 opened 2 months ago by wuzhiyue111
  0

• DDP and BackboneFinetuning: model weights get out of sync when unfreezing layers for training

  #20340 opened 2 months ago by ksikka
  2

• Impove how argument passing via CLI and config file is handled in regards to argument linking

  #20341 opened 2 months ago by MrWhatZitToYaa
  0

• Unreadable font color theme of YAML files

  #20335 opened 3 months ago by MrWhatZitToYaa
  3

• PyTorchProfiler: not showing CPU memory used even with `profile_memory=True`

  #20339 opened 2 months ago by Jack12xl
  0

• restore_training_state before on_fit_start?

  #20338 opened 2 months ago by lampuiho
  0

• `Trainer`'s `.init_module()` context does not initialize model on target device

  #20307 opened 3 months ago by jin-zhe
  1

• Add a Chinese version of README

  #20332 opened 3 months ago by nocoding03
  0

• Deepspeed Startegy doesn't set num_checkpoints while using activation partitions

  #20329 opened 3 months ago by Gforky
  1

• RuntimeError when running basic GAN model (from tutorial at lightning.ai) with DDP

  #20328 opened 3 months ago by pranavrao-qure
  3

• `strict = False` does not work when the checkpoint is distributed

  #20274 opened 3 months ago by NathanGodey
  1

• Bad practice in GAN example

  #20331 opened 3 months ago by MrWhatZitToYaa
  1

• Support A Variable Number of Batches

  #20330 opened 3 months ago by e-yi
  1

• Add list to torch.Tensor injection in yaml config

  #20324 opened 3 months ago by fguiotte
  0

• best-k-metrics in ModelCheckpoint

  #20321 opened 3 months ago by gonzachiar
  0

• Unable to serialize WandbLogger

  #20315 opened 3 months ago by cwallenwein
  2

• Incosistant memory usage comparing to huggingface trainer when using deepspeed

  #20299 opened 3 months ago by mickeysun0104
  5

• `hparams` not loaded when loading checkpoint via LightningCLI

  #20310 opened 3 months ago by YouRik
  0

• Split `reload_dataloaders_every_n_epochs` into separate parameters for train, val and test dataloaders

  #20309 opened 3 months ago by windring
  3

• `NeptuneCallback` produces lots of `X-coordinates (step) must be strictly increasing` errors

  #20281 opened 3 months ago by iirekm
  1

• The problem shows: version incompatibility from v1.3.x to v2.4

  #20308 opened 3 months ago by sunhan3787
  1

• NCCL backend fails during multi-node, multi-GPU training

  #20306 opened 3 months ago by raketenolli
  0

• the example that shows "The LightningModule also has access to the Hyperparameters" is not correct

  #20303 opened 3 months ago by XinleiRen
  0

• Problem in multi-gpu training

  #20264 opened 4 months ago by xizaoqu
  2

• RichProgressBar: refresh_rate doesn't affect metric_component

  #20300 opened 3 months ago by marios1861
  0

• Error encountered while using multiple optimizers inside a loop.

  #20296 opened 3 months ago by RAraghavarora
  0

• Fabric does not sync gradients?

  #20293 opened 3 months ago by RuABraun
  0

• Saving a checkpoint every n epochs does not work as expected

  #20282 opened 3 months ago by olly-writes-code
  2

• WandbLogger will cause error on TPU v3-8

  #20278 opened 3 months ago by buoyancy99
  0

• Lightning place model inputs and model to different devices

  #20276 opened 3 months ago by Kamichanw
  1

• MLFlow logger returns None when MLFlow server is used

  #20273 opened 4 months ago by lilruwu
  0

• Custom batch sampler fails to re-instantiate in `_dataloader_init_kwargs_resolve_sampler`

  #20272 opened 4 months ago by Kamichanw
  1

• rich progress bar shows v_num as 0.000

  #20268 opened 4 months ago by npuichigo
  0

• `_update_dataloader` improperly copies state of subclassed dataloader with attribute names that differ from `__init__` parameters.

  #20265 opened 4 months ago by spenceforce
  0