Ecosyste.ms: Issues

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

GitHub / peleiden/daluke issues and pull requests

#136 - Investigate gradient locking/unlocking

Issue - State: open - Opened by asgerius over 3 years ago

#135 - Plot parameter changes by group

Pull Request - State: closed - Opened by asgerius over 3 years ago - 1 comment

#134 - Feature/roberta finetune

Pull Request - State: closed - Opened by sorenmulli over 3 years ago

#133 - Feature/no epochs

Pull Request - State: closed - Opened by asgerius over 3 years ago - 1 comment

#132 - Speed up data loading

Pull Request - State: closed - Opened by asgerius over 3 years ago - 2 comments

#131 - Make data loading faster

Issue - State: open - Opened by asgerius over 3 years ago

#130 - Make dataloading use multiple threads

Issue - State: closed - Opened by asgerius over 3 years ago - 1 comment

#129 - Validate model multiple times during epoch

Pull Request - State: closed - Opened by asgerius over 3 years ago

#128 - Feature/limit vocab size

Pull Request - State: closed - Opened by asgerius over 3 years ago

#127 - Feature/pca

Pull Request - State: closed - Opened by sorenmulli over 3 years ago - 1 comment

#126 - Remove rarely used tokens

Issue - State: closed - Opened by asgerius over 3 years ago

#125 - Auto-annotate DAGW and build data

Pull Request - State: closed - Opened by asgerius over 3 years ago - 1 comment

#124 - Unhardcode special tokens

Issue - State: closed - Opened by asgerius over 3 years ago

#123 - Feature/data

Pull Request - State: closed - Opened by asgerius over 3 years ago

#122 - Compare inference speed in reproduction

Issue - State: open - Opened by sorenmulli over 3 years ago - 1 comment

#121 - Calculate loss and top K acc. for validation every epoch

Pull Request - State: closed - Opened by sorenmulli over 3 years ago

#120 - Allow batching of NER API predictions

Pull Request - State: closed - Opened by sorenmulli over 3 years ago

#119 - Let entity represenations live in lower dim. space

Issue - State: closed - Opened by sorenmulli over 3 years ago

#118 - Let entity represenations live in lower dim. space

Issue - State: closed - Opened by sorenmulli over 3 years ago

#117 - Look into relational tokens using article categories

Issue - State: open - Opened by sorenmulli over 3 years ago

#116 - Train with MultiRoberta as base model

Issue - State: closed - Opened by sorenmulli over 3 years ago

#115 - Split dataset for validation

Issue - State: closed - Opened by sorenmulli over 3 years ago

#114 - Add code to prepare auto-annotated dataset

Issue - State: closed - Opened by sorenmulli over 3 years ago

#111 - Batched API (multiexamples)

Issue - State: closed - Opened by sorenmulli over 3 years ago

#110 - Allow giving entity ID's to predict_mlm in API

Issue - State: open - Opened by sorenmulli over 3 years ago
Labels: niceness

#109 - Stop API from logging with new version of pelutils

Issue - State: closed - Opened by sorenmulli over 3 years ago

#108 - Python API with state

Pull Request - State: closed - Opened by sorenmulli over 3 years ago

#107 - fetch_model's use of log levels makes it die with unconfigured logger

Issue - State: closed - Opened by sorenmulli over 3 years ago - 8 comments
Labels: bug

#106 - Dont create tokenizer multiple times in CLI

Issue - State: closed - Opened by sorenmulli over 3 years ago

#105 - Adding it to the DaCy robustness test

Issue - State: open - Opened by KennethEnevoldsen over 3 years ago - 2 comments

#104 - Add finetuning instructions

Issue - State: open - Opened by sorenmulli over 3 years ago
Labels: documentation, niceness

#103 - Make API modifications such that we can send it to DaNLP

Issue - State: closed - Opened by sorenmulli over 3 years ago

#101 - _

Issue - State: closed - Opened by sorenmulli over 3 years ago

#100 - Make daluke run on Python 3.7

Issue - State: closed - Opened by asgerius over 3 years ago

#99 - Calibration of DaLUKE predictions

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment
Labels: niceness

#98 - Correlation between DaLUKE and DaBERT and DaCy predictions

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment
Labels: niceness

#97 - Add probabilities to mask api output

Issue - State: closed - Opened by sorenmulli over 3 years ago

#96 - Make `daluke` command work on windows

Issue - State: closed - Opened by sorenmulli over 3 years ago

#94 - Download and store both pretrained and NER models

Issue - State: closed - Opened by sorenmulli over 3 years ago - 2 comments

#93 - DaCy medium

Issue - State: closed - Opened by sorenmulli over 3 years ago

#92 - Initialize query matrices for Bert attention in los fine-tuningos

Issue - State: closed - Opened by asgerius over 3 years ago
Labels: niceness

#91 - Confusion matrix of NER predictions

Issue - State: closed - Opened by asgerius over 3 years ago

#90 - Some kind of confusion matrix of NER predictions

Issue - State: closed - Opened by sorenmulli over 3 years ago
Labels: niceness

#89 - Evaluate pretraining performance

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment
Labels: important

#88 - mutate_for_ner should not be hardcoded for a padding ID of 0

Issue - State: closed - Opened by asgerius over 3 years ago - 4 comments
Labels: important

#87 - Error analysis on daLUKE

Issue - State: closed - Opened by sorenmulli over 3 years ago - 2 comments

#86 - Print out some examples of masked language and entity modelling

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment

#85 - "Matrix" of dataset train and fine-tune

Issue - State: closed - Opened by sorenmulli over 3 years ago

#84 - Update plots

Issue - State: closed - Opened by asgerius over 3 years ago - 4 comments

#83 - NER return best (on dev)

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment

#82 - Add switch to bypass initialization using daBERT

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment

#81 - Entity vocabulary contains "Kategori"

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment
Labels: bug, important

#80 - Plot parameter dist.

Issue - State: closed - Opened by sorenmulli over 3 years ago

#79 - Add masking params to hyperparams

Issue - State: closed - Opened by sorenmulli over 3 years ago

#78 - Representation geometry over seq. length

Issue - State: closed - Opened by sorenmulli over 3 years ago

#77 - Investigate correctness of data, masking, and accuracy in pretraining

Issue - State: closed - Opened by asgerius over 3 years ago - 1 comment

#76 - Representation geometry wo. entities

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment
Labels: niceness

#75 - Ablation on daBERT initialization

Issue - State: closed - Opened by asgerius over 3 years ago
Labels: niceness

#74 - Ablation on finetuning concatenation

Issue - State: closed - Opened by sorenmulli over 3 years ago

#73 - Add better file not found to load_from_archive

Issue - State: closed - Opened by sorenmulli over 3 years ago

#72 - Plot PCA, t-SNE og UMAP /w positive classes

Issue - State: closed - Opened by sorenmulli over 3 years ago

#70 - Consider our own additions to daLUKE - what about relational tokens?

Issue - State: closed - Opened by sorenmulli over 3 years ago
Labels: niceness

#69 - Cross-valdate NER

Issue - State: closed - Opened by sorenmulli over 3 years ago
Labels: niceness

#67 - Make NER deterministic

Issue - State: closed - Opened by sorenmulli over 3 years ago

#65 - Clean up NER data-API

Issue - State: closed - Opened by sorenmulli over 3 years ago

#64 - Finalize PCA/t-SNE/UMAP

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment

#62 - NER: Show running F1 on train set

Issue - State: closed - Opened by sorenmulli over 3 years ago

#61 - Make entity augmentation experiment

Issue - State: closed - Opened by sorenmulli over 3 years ago

#59 - Create goals and get overview for our deployment plans

Issue - State: closed - Opened by sorenmulli over 3 years ago - 1 comment

#58 - Loss weighting in pretraining

Issue - State: closed - Opened by asgerius over 3 years ago

#57 - Consider requirements

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 1 comment
Labels: niceness

#56 - Allow some hyperparameters to be overwritten when resuming

Issue - State: closed - Opened by asgerius almost 4 years ago - 7 comments

#55 - Add more datasets to finetuning of daLUKE

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#54 - Add dacy to our NER reproduction

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#53 - Finetuning: Big boi hyperparameter seach

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 10 comments

#52 - Finetuning: Weighted loss?

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 1 comment

#51 - Finetuning should save jobs in subfolders

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#50 - Load daBERT in and analyse its' performance

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 1 comment

#49 - Check geometry of LUKE CWR's

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#48 - Make big K run every L sub-kai

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#47 - Continue "big boi"

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 3 comments

#46 - Make resume load hyperparams

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#45 - Fix off-by-one in resume training

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 2 comments

#44 - Debug NER

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#43 - Profiling produces incorrect measures

Issue - State: closed - Opened by asgerius almost 4 years ago

#39 - Choose learning rate during the training

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 3 comments

#37 - Choose weighting between word and entity loss

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 1 comment

#36 - Choose batch size

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 2 comments

#35 - Lower case training data as BotXO is trained on lower case

Issue - State: closed - Opened by sorenmulli almost 4 years ago

#13 - Analyse English LUKE

Issue - State: closed - Opened by sorenmulli almost 4 years ago
Labels: niceness

#12 - Perform more interesting analysis of Danish NER models predicitions

Issue - State: closed - Opened by sorenmulli almost 4 years ago
Labels: niceness

#11 - Save Danish NER models probability distributions

Issue - State: closed - Opened by sorenmulli almost 4 years ago
Labels: niceness

#10 - Make the Danish BERT better (allow MISC)

Issue - State: closed - Opened by sorenmulli almost 4 years ago - 1 comment
Labels: niceness