All pull requests

Pull requests

• [BugFix] Pickle MemoryMappedTensor.from_storage results whose storage has a file

  CLA SignedmemmapTest
  #2037·peterdsharpe opened Oct 10, 2026Collaborator·

  ·
  1

• [BugFix] Fill only the masked entries in masked_fill_ on an indexed PersistentTensorDict

  bugCLA SignedpersistentTest
  #2036·peterdsharpe opened Oct 10, 2026Collaborator·

  ·
  1

• [BugFix] Make td % tensor work for a tensor with dims, as td + tensor does

  bugCLA SignedTest
  #2035·peterdsharpe opened Oct 10, 2026Collaborator·

  ·
  1

• [BugFix] Fill the masked batch elements in masked_fill_ of TensorDictStore and LazyStackedTensorDictStore[i]

  CLA SignedstoreTest
  #2031·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [BugFix] Use the sample key of a nested ProbabilisticTensorDictSequential in composite get_dist

  CLA SignednnTest
  #2028·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [BugFix] Run as_tensordict_module methods with the instance they are called on

  bugCLA SignednnTest
  #2026·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [BugFix] Move consolidated tensordicts to the meta device with non_blocking=False

  CLA SignedTest
  #2024·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [BugFix] Stop EnsembleModule.reset_parameters_recursive from warning about its own parameters and reordering batch dims

  bugCLA SignednnTest
  #2022·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [Doc] Add a guide to torch.compile, torch.func, torch.export and CUDA graphs with tensordict

  CLA Signeddocumentation
  #2019·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Feature] Write scalars into stores with store[index] = value

  CLA SignedFeaturestoreTest
  #2018·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Quality] Type the tensorclass constructor from its fields

  CLA SigneddocumentationQualitytensorclassTest
  #2017·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Quality] Fix the argument and return types of the TensorDict and tensorclass stubs

  CLA SigneddocumentationpersistentQualitytensorclassTest
  #2016·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Deprecation] Deprecate the fields and extend tensorclass methods and the safe argument of from_tensordict

  CLA SignedDeprecationnntensorclassTest
  #2015·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [BugFix] Reserve the names that only tensorclasses have as field names, and allow _cache

  bugCLA SigneddocumentationnntensorclassTest
  #2014·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [BugFix] Make set_get_defaults_to_none apply to tensorclass get

  bugCLA SignedCompiletensorclassTest
  #2013·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [BugFix] Rebuild nested tensorclass fields in from_tensordict and with autocast

  bugCLA SignedCompiledocumentationnntensorclassTest
  #2012·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1

• [BugFix] Support torch.reshape, torch.tensor_split, torch.chunk and torch.repeat_interleave on tensordicts, and torch.where on tensorclasses

  CLA SignedCompiletensorclassTest
  #2011·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [BugFix] Iterate over tensorclasses as over tensordicts, and stop NonTensorData iteration

  bugCLA SignedCompilepersistenttensorclassTest
  #2010·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Quality] Name the key in batch-size mismatch errors and reword unclear error messages

  CLA SignedpersistentQualitystoretensorclassTest
  #2009·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [BugFix] Count nested and leaf keys in len() of lazy-stack key views

  bugCLA SignedTest
  #2008·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Performance] Read the tensordict.nn modes without a lock on every module call

  BenchmarksCLA SignedCompilennTest
  #2007·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Performance] Speed up tensorclass field access and NonTensorData construction

  BenchmarksCLA SignedtensorclassTest
  #2006·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [Performance] Cut per-leaf Python overhead in key views, type checks, stack, cat and apply

  BenchmarksCLA SignedTest
  #2005·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [BugFix] Make modules exported with torch.export (non-strict) accept TensorDict inputs

  CLA SignedCompilefunctionalTest
  #2004·peterdsharpe opened Oct 9, 2026Collaborator·

  ·

• [BugFix] Make TensorDict arithmetic work inside torch.vmap

  bugCLA SignedCompiledocumentationfunctionalTest
  #2003·peterdsharpe opened Oct 9, 2026Collaborator·

  ·
  1