Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / dream-faster/modular-pipelines issues and pull requests
#219 - fix(DataSet): hate_speech_offensive dataset's labels were incorrect. now it's a lot smaller!
Pull Request -
State: open - Opened by almostintuitive almost 2 years ago
- 1 comment
#218 - Fix: Don't include data when pushing configs to ``wandb``
Issue -
State: closed - Opened by szemyd almost 2 years ago
#217 - fix(Plugins): Added ``key_exclude`` param to ``obj_to_dict`` to filter out data.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#216 - fix(DataLoader): added multiple sampling options to MergedDataLoader, because we don't want to over or undersample the training data, only the test data for now
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#215 - fix(DataLoader): you couldn't reuse Dataloaders before to load different splits
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#214 - feature(Library): added working multi-objective setup
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#213 - Fix: Plugins - outputanalyzer fails when training on tweeteval then evaluating on "offensive"
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#212 - Docs: More examples
Issue -
State: open - Opened by szemyd almost 2 years ago
#211 - feature(Library): added new set of Experiments, removed the ability to store multiple PreprocessingConfigs
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#210 - Feature(IO): Pickling shouldn't pickle models or loaded data.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#209 - Feature: Pickling should only pickle the pipeline definition.
Issue -
State: closed - Opened by szemyd almost 2 years ago
#208 - Feature(Package): Packaging python module for Demo usage.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#207 - Investigate: Can we get out more of the sklearn models
Issue -
State: open - Opened by almostintuitive almost 2 years ago
#206 - feature(Docs): Added diagram workflow.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#205 - feature(Plugin): Outputanalyser refractored to dframes and wandb logs outputs to tables now.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#204 - feature(Metrics): added f1/precision/recall for class 0
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#203 - chore(Plugin): Better printing of pipeline to wandb.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#202 - Feature: Log ``classification_report`` metric to ``wandb``
Issue -
State: open - Opened by szemyd almost 2 years ago
#201 - feature(Evaluation): Added multiple type of ``f1`` eval
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#200 - fix(Model): use a different bertweet repository (maybe it's more available?)
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#199 - feature(Pipeline): Masking of datasources works now.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#198 - Fix: Delete the run
Issue -
State: open - Opened by almostintuitive almost 2 years ago
#197 - feature(Plugin): Wandb now correctly logs the config of mutated pipeline.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#196 - feature(Tests): added a single experiment that we log to wandb on the CI
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#195 - feature(Experiment): finished multi-objective experiment
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#194 - fix(Blocks): speed up spacy tokenizer, only load Dataset when necessary (speed up initial phase)
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#193 - chore(Pipeline): Refractoring functions
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#192 - feature(DevOps): Added github workflow.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#191 - fix(Logging): output_analyzer plugin should use the predict datasource
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#190 - feature(Testing): Integration test added.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#189 - Feature: Add option to hand pick dataset categories for ``dev`` staging, as it fails if it only receives one category.
Issue -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#188 - fix(Pipeline): We also need to copy the list items.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#187 - fix(Runner): copy in obj_to_dict() to avoid mutating the original object, overriding global_dataloader's config, fixed HuggingfaceModel declarations, commented out try/catch
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#186 - fix(Pipeline): ``get_config`` gave key error
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#185 - feature(Wandb): Pipeline hierarchy is added to ``wandb``
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 2 comments
#184 - feature(Experiment): finished multi-objective experiment
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#183 - feature(Pipeline): Improved ``get_config`` and iteratively expanding objects to dicts.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#182 - feature(Library): reshuffled experiments to use a single project name, added DataLoader interface, renamed HuggingfaceDataLoader
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#181 - feature(Library): Refractored and added baseline experiments.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#180 - Investigate: In certain cases, 0s and 1s model is achieving 1 recall/precision on balanced datasets
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#179 - feature(Utils): added print_dataset_analytics()
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 2 comments
#178 - fix(Plugin): OutputAnalyser works on multi-objective
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#177 - fix(Experiments): added sourcetype parameter to get_labels, added Experiment.global_dataloader
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#176 - feature(Data): Multiple ``DataSources``
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#175 - feature(Library): Multi-task pipeline for transformers.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#174 - feature(Logging): We now centrally handle the logging pipe.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#173 - feature(Runner): ``try-except`` implemented on ``runner.run()``
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#172 - Feature: Log DataLoader information to ``wandb`` to recognize easier which run was done with which dataset.
Issue -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#171 - fix(Models): Huggingface tokenizer also truncates at inference time
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#170 - Feature: Perspective API hooked up.
Issue -
State: closed - Opened by szemyd almost 2 years ago
#169 - feature(Model): Perspective-API created and hooked up
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#168 - feature(Experiments): added dynahate dataset, created all the cross-experiments for the blog post
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#167 - Feature: Create 2 more test splits that we can evaluate on (+ a merged test split)
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#166 - Feature: Create Huggingface meta model, where the predictions are concatenated with the original string
Issue -
State: open - Opened by almostintuitive almost 2 years ago
#165 - Feature: Add Perspective API
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#164 - Investigate: Reason for TextStats model to outperform on hatecheck
Issue -
State: open - Opened by almostintuitive almost 2 years ago
#163 - Fix: Calibrate SKLearn Model fails
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#162 - Investigate: Reason for Meta model underperformance
Issue -
State: open - Opened by almostintuitive almost 2 years ago
#161 - feature(Experiments): Added latest hate speech detection experiment variations
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#160 - feature(Models): huggingface variations added
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#159 - Feature: Clear delianation of which mode (train, infer) we are in.
Issue -
State: closed - Opened by szemyd almost 2 years ago
#158 - Feature(OutputAnalyser): Created a Correlation matrix metric
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#157 - Feature: Correlation matrix to ``OutputAnalyser``.
Issue -
State: closed - Opened by szemyd almost 2 years ago
#156 - feature(Models): added `calibrate` option to SKLearn
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#155 - Feature: Create development pipeline for quick development of features.
Issue -
State: closed - Opened by szemyd almost 2 years ago
#154 - refactor(Library): separated out models, datasets, pipelines, experiments
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#153 - Fix: Pipeline - get_configs() type mismatch
Issue -
State: open - Opened by almostintuitive almost 2 years ago
#152 - fix(Plugin): Added plugin order so we log entire to wandb.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#151 - fix(Models): explicitly add empty save() & load() to random & vader models
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#150 - fix(Plugin): Output Analyzer now type checked.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#149 - refactor(Types): flip `force_fit` to `frozen`
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#148 - Fix: OutputAnalyzer - fails on merged (hatespeech) dataset
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#147 - refactor(Models): pipeline is only constructed before inference for Huggingface, fixed types, saving
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#146 - Feature: Huggingface - We need to implement batching (at inference) ourselves with pipelines
Issue -
State: open - Opened by almostintuitive almost 2 years ago
#145 - fix(Huggingface): assertion correctly specified, inference after training now works, saving the full path on hfhub
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#144 - fix(Huggingface): Huggingface ``pipeline`` and ``AutoModelForSequenceClassification.from_pretrained`` seperated.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
- 2 comments
#143 - fix(DataLoader): only run dataset transformation once, fixed dataset transformations
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 2 comments
#142 - feature(Loading): If ``preferred_origin`` set, then no fallback. Naming starts with ``project_id``
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#141 - Fix: Huggingface - Should we append the project and pipeline name to the automatically created HF repo?
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#140 - feature(Plugin): Output Analyser
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#139 - fix(Huggingface): Preloading works now.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#138 - fix(Hierarchy): Hierarchy was badly defined for concat, fixed it now.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#137 - fix(DataLoader): override preprocessing config for merged datasets, only use a single, explicit transformation (no way to compose them right now)
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#136 - Fix(HF): Pretrained shouldn't train if we are doing only inference.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#135 - fix(Concat): Datasplit object ``.value`` and concat not a df
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#134 - feature(Experiments): added new Experiments, fixed random seed
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
#133 - chore(Dependencies): updated
Pull Request -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#132 - feature(Adaptor): ClassificationAdaptor now let's you select ``prediction`` of ``probabilities`` between models.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#131 - Fix: Resolve confusion in naming or usage ``DataType.PredictionsWithProbs`` and ``List[PredsWithProbs]``
Issue -
State: open - Opened by szemyd almost 2 years ago
#130 - feature(Dataloader): Dataloading wrapped in a class.
Pull Request -
State: closed - Opened by szemyd almost 2 years ago
#124 - Feature: ``random_seed`` across the entire project.
Issue -
State: closed - Opened by szemyd almost 2 years ago
- 2 comments
#117 - Fix: Formatting - Don't apply Black & isort at the same time (results in jumping around with long files)
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#116 - Fix: RunConfig's remote_save flag is not wired up?
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
- 1 comment
#112 - Feature: Create a multi-modal ``transformer`` pipeline, with a meta-model as a second step.
Issue -
State: closed - Opened by szemyd almost 2 years ago
#111 - Feature: Pluginize id injection.
Issue -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#88 - Feature: Sklearn - Add calibration step
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#77 - Feature: Train a multi-objective meta-model
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#71 - Feature: Add Imbalance transformation
Issue -
State: closed - Opened by almostintuitive almost 2 years ago
#28 - Feature: Visualize the pipeline graph.
Issue -
State: closed - Opened by szemyd almost 2 years ago
- 1 comment
#12 - Feature: Create 1 acceptance test
Issue -
State: closed - Opened by almostintuitive about 2 years ago