i-machine-think/diagNNose

Issues

• How to use pretrained models from AutoModelForSeq2SeqLM (google/flan-t5-xxl)

  #89 opened a year ago by DavidAdamczyk
  3

• Add option for loading transformer model from file path

  #87 opened 2 years ago by oskarvanderwal
  1

• init_sentence.txt is not part of pip release

  #70 opened 3 years ago by jumelet
  0

• Use _convert_id_to_token for create tokenizer vocab dict

  #83 opened 3 years ago by jumelet
  0

• Correctly handle multi-word expressions in BLiMP one_prefix_word_good

  #82 opened 3 years ago by jumelet
  0

• Fix sen_idx for sliced corpus in Extractor

  #80 opened 4 years ago by jumelet
  0

• Change typing of Corpus.slice args

  #79 opened 4 years ago by jumelet
  0

• Fix Explainer for Transformer LMs

  #78 opened 4 years ago by jumelet
  0

• Make library compatible with torchtext 0.9.0

  #76 opened 4 years ago by jumelet
  0

• Incorporate generator return type of ActivationReader throughout the library

  #75 opened 4 years ago by jumelet
  0

• Use set_default_tensor_type for Cuda

  #74 opened 4 years ago by jumelet
  0

• Test GPU support for both LSTMs and Transformers

  #72 opened 4 years ago by jumelet
  0

• Fix model top_layer in syntax

  #71 opened 4 years ago by jumelet
  0

• Make lib py3.6 able

  #41 opened 4 years ago by jumelet
  0

• Bring back concat_pickle_dumps after extraction has finished

  #69 opened 4 years ago by jumelet
  0

• Incorprate HF transformer compatability

  #48 opened 4 years ago by jumelet
  0

• Look into subclassing from torch.Tensor directly

  #68 opened 4 years ago by jumelet
  0

• Allow labels attribute name to be set for DCTrainer

  #45 opened 4 years ago by jumelet
  0

• Allow class_weights to be set for unbalanced data

  #67 opened 4 years ago by jumelet
  0

• Incorporate tensor device for ShapleyTensors

  #66 opened 4 years ago by jumelet
  0

• Add numpy compatibility next to torch tensors

  #44 opened 4 years ago by jumelet
  0

• Decrease coupling between config_dict and base classes

  #46 opened 4 years ago by jumelet
  0

• Add functionality for downloading Downstream corpora

  #49 opened 4 years ago by jumelet
  0

• Create a model-agnostic attribution method

  #53 opened 4 years ago by jumelet
  0

• Skip extraction of sentences with no items from selection_func

  #62 opened 4 years ago by jumelet
  0

• Incorporate Transformer LM tasks

  #54 opened 4 years ago by jumelet
  0

• Look into parallelisation for Shapley computation

  #65 opened 4 years ago by jumelet
  0

• Check if old GCD scripts still comply with first release

  #64 opened 4 years ago by jumelet
  0

• Add feature_mask to ShapleyTensor

  #63 opened 4 years ago by jumelet
  0

• Investigate the "some weights were not initialized" warning on HF models

  #61 opened 4 years ago by jumelet
  0

• Make suppress_print toggleable

  #42 opened 4 years ago by jumelet
  0

• Unify vocab_path configuration

  #52 opened 4 years ago by jumelet
  0

• Remove indexing by key in ActivationReader

  #50 opened 4 years ago by jumelet
  0

• Add preproc_linzen code to downstream

  #43 opened 4 years ago by jumelet
  0

• Apply Factory pattern to LM creation

  #60 opened 4 years ago by jumelet
  0

• Recreate ContextualDecomposition from ShapleyTensor

  #59 opened 4 years ago by jumelet
  0

• Incorporate Structural Probes

  #58 opened 4 years ago by jumelet
  0

• Allow downstream corpora to be downloaded automatically

  #57 opened 4 years ago by jumelet
  0

• Create subword averaging for transformer LMs

  #56 opened 4 years ago by jumelet
  0

• Create visualiser for ShapleyTensors

  #55 opened 4 years ago by jumelet
  0

• Reconsider GoogleLM support

  #51 opened 4 years ago by jumelet
  0

• Use ApacheArrow for activation storing

  #47 opened 4 years ago by jumelet
  0

• Create small suite of select functions

  #40 opened 5 years ago by jumelet
  0

• Consider structuring arg config as structured dict

  #38 opened 5 years ago by jumelet
  0

• Check test.txt with uneven sen lens

  #37 opened 5 years ago by jumelet
  0

• Add Apache license

  #39 opened 5 years ago by jumelet
  0

• Batchify CD

  #35 opened 5 years ago by jumelet
  0

• Pass corpus to InitStates and use eos activations of that corpus

  #34 opened 5 years ago by jumelet
  0

• Input unk for GoogleLM

  #36 opened 5 years ago by jumelet
  0

• Port MaskedArray's to PackedSequences

  #33 opened 5 years ago by jumelet
  0