butchland/fastai_xla_extensions

Issues

• ImportError: cannot import name 'inf' from 'torch'

  #49 opened a year ago
  0

• NameError: name 'xmp' is not defined

  #48 opened a year ago
  0

• Please be compatible with the latest version of fastai, thank you!

  #44 opened 2 years ago
  0

• Is this repo still working? Can't run the colab notebooks provided in the doc

  #37 opened 3 years ago
  1

• xla_fit_one_cycle fails with an error when using fastai version 2.3.1

  #35 opened 3 years ago
  0

• xla_lr_find locks xm.xla_device in main process

  #33 opened 3 years ago
  0

• creating a model with concat_pool = True creates a model that can't run on multiple tpu cores

  #28 opened 3 years ago
  0

• using fastai data loaders is slower on multi core is slow if using large image sizes

  #27 opened 3 years ago
  0

• fastai_xla_extensions on kaggle errors out

  #25 opened 4 years ago
  7

• export and load_learner (for loading for inference) is not working

  #22 opened 3 years ago
  1

• default_device no longer sets to overridden default_device method

  #15 opened 4 years ago
  1

• Setting pretrained = true causes the cnn_learner model to not train

  #14 opened 4 years ago
  2

• run batch transforms on CPU if dataloader is using a TPU

  #11 opened 4 years ago
  1

• Add patch to LMLanguageLearner to fix problem on load_ignore_keys where state_dict error due to use of TPU

  #7 opened 4 years ago
  0

• Language model trainer is very slow or not finishing fit_one_cycle for 1 epoch

  #6 opened 4 years ago
  2

• batch transforms for vision are slow

  #5 opened 4 years ago
  4

• learner.lr_find returns wrong result

  #4 opened 4 years ago
  2

• TODO

  #2 opened 4 years ago
  0