Ecosyste.ms: Issues

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

GitHub / i-machine-think/diagNNose issues and pull requests

#88 - Fix deprecated way of importing MutableMapping

Pull Request - State: closed - Opened by oskarvanderwal over 2 years ago

#87 - Add option for loading transformer model from file path

Issue - State: closed - Opened by oskarvanderwal over 2 years ago - 1 comment

#86 - Update requirements.txt to also install fairseq

Pull Request - State: closed - Opened by oskarvanderwal over 2 years ago

#85 - :bug: Add device param to pseudo ll method

Pull Request - State: closed - Opened by jumelet about 3 years ago

#84 - Development

Pull Request - State: closed - Opened by jumelet about 3 years ago

#83 - Use _convert_id_to_token for create tokenizer vocab dict

Issue - State: open - Opened by jumelet over 3 years ago
Labels: enhancement

#82 - Correctly handle multi-word expressions in BLiMP one_prefix_word_good

Issue - State: open - Opened by jumelet almost 4 years ago
Labels: enhancement, suggestion

#81 - AWD-LSTM fix

Pull Request - State: closed - Opened by Kalsir almost 4 years ago

#80 - Fix sen_idx for sliced corpus in Extractor

Issue - State: open - Opened by jumelet almost 4 years ago
Labels: bug

#79 - Change typing of Corpus.slice args

Issue - State: open - Opened by jumelet almost 4 years ago
Labels: refactor

#78 - Fix Explainer for Transformer LMs

Issue - State: open - Opened by jumelet almost 4 years ago
Labels: bug, priority

#77 - Various hotfixes and library improvements

Pull Request - State: closed - Opened by jumelet almost 4 years ago

#76 - Make library compatible with torchtext 0.9.0

Issue - State: open - Opened by jumelet almost 4 years ago
Labels: priority

#75 - Incorporate generator return type of ActivationReader throughout the library

Issue - State: open - Opened by jumelet almost 4 years ago
Labels: refactor

#74 - Use set_default_tensor_type for Cuda

Issue - State: open - Opened by jumelet almost 4 years ago

#73 - Update 'Research using DiagNNose'

Pull Request - State: closed - Opened by LucWeber almost 4 years ago

#72 - Test GPU support for both LSTMs and Transformers

Issue - State: open - Opened by jumelet about 4 years ago
Labels: enhancement

#71 - Fix model top_layer in syntax

Issue - State: closed - Opened by jumelet about 4 years ago
Labels: bug, priority

#70 - init_sentence.txt is not part of pip release

Issue - State: closed - Opened by jumelet about 4 years ago
Labels: bug, priority

#69 - Bring back concat_pickle_dumps after extraction has finished

Issue - State: open - Opened by jumelet about 4 years ago
Labels: bug

#68 - Look into subclassing from torch.Tensor directly

Issue - State: open - Opened by jumelet about 4 years ago

#67 - Allow class_weights to be set for unbalanced data

Issue - State: open - Opened by jumelet about 4 years ago

#66 - Incorporate tensor device for ShapleyTensors

Issue - State: closed - Opened by jumelet about 4 years ago

#65 - Look into parallelisation for Shapley computation

Issue - State: open - Opened by jumelet over 4 years ago

#64 - Check if old GCD scripts still comply with first release

Issue - State: open - Opened by jumelet over 4 years ago
Labels: suggestion

#63 - Add feature_mask to ShapleyTensor

Issue - State: open - Opened by jumelet over 4 years ago
Labels: enhancement, refactor

#62 - Skip extraction of sentences with no items from selection_func

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: enhancement

#61 - Investigate the "some weights were not initialized" warning on HF models

Issue - State: open - Opened by jumelet over 4 years ago
Labels: invalid, question

#60 - Apply Factory pattern to LM creation

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: refactor, priority

#59 - Recreate ContextualDecomposition from ShapleyTensor

Issue - State: open - Opened by jumelet over 4 years ago
Labels: suggestion

#58 - Incorporate Structural Probes

Issue - State: open - Opened by jumelet over 4 years ago
Labels: enhancement

#57 - Allow downstream corpora to be downloaded automatically

Issue - State: open - Opened by jumelet over 4 years ago
Labels: enhancement

#56 - Create subword averaging for transformer LMs

Issue - State: open - Opened by jumelet over 4 years ago
Labels: enhancement

#55 - Create visualiser for ShapleyTensors

Issue - State: open - Opened by jumelet over 4 years ago
Labels: enhancement

#54 - Incorporate Transformer LM tasks

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: enhancement

#53 - Create a model-agnostic attribution method

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: new feature

#52 - Unify vocab_path configuration

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: bug, priority

#51 - Reconsider GoogleLM support

Issue - State: open - Opened by jumelet over 4 years ago
Labels: suggestion

#50 - Remove indexing by key in ActivationReader

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: refactor

#49 - Add functionality for downloading Downstream corpora

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: new feature

#48 - Incorprate HF transformer compatability

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: enhancement, priority

#47 - Use ApacheArrow for activation storing

Issue - State: open - Opened by jumelet over 4 years ago
Labels: enhancement

#46 - Decrease coupling between config_dict and base classes

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: refactor, priority

#45 - Allow labels attribute name to be set for DCTrainer

Issue - State: closed - Opened by jumelet over 4 years ago
Labels: refactor

#44 - Add numpy compatibility next to torch tensors

Issue - State: closed - Opened by jumelet about 5 years ago

#43 - Add preproc_linzen code to downstream

Issue - State: closed - Opened by jumelet about 5 years ago
Labels: enhancement

#42 - Make suppress_print toggleable

Issue - State: closed - Opened by jumelet about 5 years ago

#41 - Make lib py3.6 able

Issue - State: closed - Opened by jumelet about 5 years ago

#40 - Create small suite of select functions

Issue - State: closed - Opened by jumelet over 5 years ago

#39 - Add Apache license

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: documentation

#38 - Consider structuring arg config as structured dict

Issue - State: closed - Opened by jumelet over 5 years ago

#37 - Check test.txt with uneven sen lens

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: bug

#36 - Input unk for GoogleLM

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: bug

#35 - Batchify CD

Issue - State: closed - Opened by jumelet over 5 years ago

#34 - Pass corpus to InitStates and use eos activations of that corpus

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: new feature

#33 - Port MaskedArray's to PackedSequences

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: refactor

#32 - Remove overrides

Issue - State: closed - Opened by jumelet over 5 years ago

#31 - Correct variable name change of corpus_path to corpus_vocab_path in Google LM

Pull Request - State: closed - Opened by LorianColtof over 5 years ago - 1 comment

#30 - Remove config overwrite option

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: suggestion

#29 - Make init states part of model

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: refactor, priority

#28 - Labels no longer part of extraction

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: suggestion

#27 - Fix three bugs in the Google LM

Pull Request - State: closed - Opened by LorianColtof over 5 years ago

#26 - Consider using h5py instead of pickle

Issue - State: closed - Opened by jumelet over 5 years ago - 1 comment
Labels: suggestion

#25 - Use os.path.join instead of custom trim

Issue - State: closed - Opened by jumelet over 5 years ago
Labels: enhancement

#24 - Decomposition

Pull Request - State: closed - Opened by jumelet almost 6 years ago

#23 - Allow decompositions to be created for a sentence directly

Issue - State: closed - Opened by jumelet almost 6 years ago
Labels: new feature

#22 - Investigate influence of using float32 vs. float64

Issue - State: closed - Opened by jumelet almost 6 years ago
Labels: enhancement

#21 - Consider using torchtext, torch.DataLoader, etc.

Issue - State: closed - Opened by jumelet almost 6 years ago
Labels: enhancement

#20 - Create array dim hints throughout code

Issue - State: open - Opened by jumelet almost 6 years ago
Labels: documentation

#19 - Move example_dir stuff outside scripts folder

Issue - State: closed - Opened by jumelet almost 6 years ago
Labels: enhancement

#18 - Activation indexing to tensor

Issue - State: closed - Opened by jumelet almost 6 years ago - 1 comment

#17 - Allow labels to be added to corpus afterwards

Issue - State: closed - Opened by jumelet almost 6 years ago - 2 comments
Labels: enhancement

#16 - Don't use defaultdict for w2i

Issue - State: closed - Opened by jumelet almost 6 years ago

#15 - Activation selection

Pull Request - State: closed - Opened by jumelet almost 6 years ago

#14 - Always save pickle dump sequence as 1 array

Issue - State: closed - Opened by jumelet almost 6 years ago - 2 comments

#13 - Lazy loading

Pull Request - State: closed - Opened by jumelet almost 6 years ago

#12 - Improve lazy loading functionality

Issue - State: closed - Opened by jumelet almost 6 years ago

#11 - Intervention nan fix

Pull Request - State: closed - Opened by Kaleidophon almost 6 years ago

#10 - Selection fix

Pull Request - State: closed - Opened by Kaleidophon almost 6 years ago

#9 - Commit messages

Pull Request - State: closed - Opened by Kaleidophon almost 6 years ago

#8 - Modifications to Activation Extraction and Training of DCs

Pull Request - State: closed - Opened by Kaleidophon about 6 years ago

#7 - Unify specification for activations

Issue - State: closed - Opened by Kaleidophon about 6 years ago
Labels: enhancement, priority

#6 - More control over activation extraction

Issue - State: closed - Opened by Kaleidophon about 6 years ago
Labels: enhancement

#5 - Batchify LSTM and Corpus

Issue - State: closed - Opened by Kaleidophon about 6 years ago
Labels: enhancement

#4 - Interventions

Pull Request - State: closed - Opened by Kaleidophon about 6 years ago

#3 - Interventions

Pull Request - State: closed - Opened by Kaleidophon about 6 years ago

#2 - feat: allow specification setup via commandline args

Pull Request - State: closed - Opened by dieuwkehupkes about 6 years ago - 1 comment

#1 - Transform usage of config

Pull Request - State: closed - Opened by dieuwkehupkes about 6 years ago