Syncs repos after new Trainer successful experiment, Data Augmentation and Portfolio sample#11

Merged

nsorros merged 307 commits intowellcometrust:mainfrom MantisAI:main

Sep 6, 2023

+2,942-896

This pull request is big! We're only showing the most recent 250 commits

Commits on Jul 18, 2023

turn argument to option
Andrei Apostol
committed
revert separate preprocess command
Andrei Apostol
committed
Revert "revert separate preprocess command"
Andrei Apostol
committed
move cache disable inside preproc
Andrei Apostol
committed
add timings
Andrei Apostol
committed

Commits on Jul 19, 2023

Speed improvements on preprocessing
Jose J. Martinez
committed
Adds benchmarking utility
Jose J. Martinez
committed

Commits on Jul 20, 2023

Cleaning. Training using preprocessing output
Jose J. Martinez
committed

Commits on Jul 21, 2023

Cleaning. Training using preprocessing output
Jose J. Martinez
committed
Cleaning. Training using preprocessing output
Jose J. Martinez
committed
Improving some steps
Jose J. Martinez
committed
Improving some steps
Jose J. Martinez
committed
Improving some steps
Jose J. Martinez
committed
Improving some steps
Jose J. Martinez
committed
Improving save_to_disk
Jose J. Martinez
committed
Improving save_to_disk
Jose J. Martinez
committed
Removing save_to_disk
Jose J. Martinez
committed
Adapts training to call preprocess
Jose J. Martinez
committed
Adapts training to call preprocess
Jose J. Martinez
committed
Adapts training to call preprocess
Jose J. Martinez
committed
Adapts training to call preprocess
Jose J. Martinez
committed
Allows pretraining+saving or just pretraining before train
Jose J. Martinez
committed

Commits on Jul 24, 2023

Cleaning and rearranging
Jose J. Martinez
committed
Updates train.sh
Jose J. Martinez
committed
Adds allMesh_2021 jsonl version
Ubuntu
committed
Ignores .jsonl
Jose J. Martinez
committed
Fixes preprocessing unit tests
Jose J. Martinez
committed
Adapts preprocessing tests to the style of other tests
Jose J. Martinez
committed
Fixes tests in train. Adds more tests.
Jose J. Martinez
committed
Adds some comments
Jose J. Martinez
committed
Reformatting
josejuanmartinez
committed
Reformatting
Jose J. Martinez
committed
Reformatting
Jose J. Martinez
committed
Reformatting
Jose J. Martinez
committed
Reformatting
Jose J. Martinez
committed
Reformatting
Jose J. Martinez
committed
Reformatting
Jose J. Martinez
committed
Reformatting
Jose J. Martinez
committed
Reformatting
Jose J. Martinez
committed
Updates the Readme
Jose J. Martinez
committed
Merge branch '3-improve-trainer' into 3-improve-trainer-juan
josejuanmartinez
authored
Reformatting
Jose J. Martinez
committed

Commits on Jul 26, 2023

Merge pull request #12 from MantisAI/11-add-common-terms-column
Apostol Andrei
authored

Commits on Jul 28, 2023

Fixes training multigpu
Ubuntu
committed
Fixes train.sh
Ubuntu
committed
Merges duplicated code
Jose J. Martinez
committed
Merge remote-tracking branch 'origin/3-improve-trainer-juan' into 3-improve-trainer-juan
Jose J. Martinez
committed
Adds label2id to config
Jose J. Martinez
committed
Fixes train.sh
josejuanmartinez
committed
Removes exit
Jose J. Martinez
committed
Adds wandb param to train.sh
Jose J. Martinez
committed
Adds report_to none for non-wandb
Jose J. Martinez
committed
Updates train.sh
josejuanmartinez
committed
Updates README.md
Jose J. Martinez
committed
Black reformatting
Jose J. Martinez
committed

Commits on Jul 31, 2023

Modifies training
josejuanmartinez
committed
Merges train.sh
josejuanmartinez
committed
Saving then evaluating
Jose J. Martinez
committed
Updates train.sh
josejuanmartinez
committed
Updates train.sh
Jose J. Martinez
committed

Commits on Aug 1, 2023

Black reformatting
Jose J. Martinez
committed

Commits on Aug 2, 2023

Adding training resuming
Jose J. Martinez
committed
Adding training resuming
Jose J. Martinez
committed
Adds resume_train.sh
josejuanmartinez
committed
Parametrization
Jose J. Martinez
committed

Commits on Aug 4, 2023

Updates documentation, script name and adds filter by years and tags to preprocessing_mesh.py
Jose J. Martinez
committed
Adding train/test split using years
Jose J. Martinez
committed
Adds multibatching to filtering train/test by years
Jose J. Martinez
committed
Updates message
Jose J. Martinez
committed
Changes shards to cpu_count() by default
Jose J. Martinez
committed
Adds preprocess example
Jose J. Martinez
committed
Adds evaluation by epoch
Jose J. Martinez
committed
Adds evaluation by epoch
Jose J. Martinez
committed
Train/test split by years
Jose J. Martinez
committed

Commits on Aug 5, 2023

Fixes drop out and standardizes calls
Jose J. Martinez
committed
Fixes years
Jose J. Martinez
committed

Commits on Aug 6, 2023

Test size frac vs row
Jose J. Martinez
committed
Test size frac vs row
Jose J. Martinez
committed
Test size frac vs row
Jose J. Martinez
committed
Test size frac vs row
Jose J. Martinez
committed
Test size frac vs row
Jose J. Martinez
committed
Test size frac vs row
Jose J. Martinez
committed
Test size frac vs row
Jose J. Martinez
committed

Commits on Aug 7, 2023

Logging model params
Jose J. Martinez
committed
Adds hidden size to train params
Jose J. Martinez
committed
Adds hidden size to train params
Jose J. Martinez
committed
Adds hidden size to train params
Jose J. Martinez
committed
Fixes bug with number of rows as test size
Jose J. Martinez
committed

Commits on Aug 8, 2023

Adds multilabel_attention and freeze_backbone
Jose J. Martinez
committed
Adds multilabel_attention and freeze_backbone
Jose J. Martinez
committed

Commits on Aug 10, 2023

Adds some debug info
Jose J. Martinez
committed
Removes forward logs
Jose J. Martinez
committed
Checking last implementation of BertMesh
Jose J. Martinez
committed
Roll back early implementation of BertMesh
Jose J. Martinez
committed

Commits on Aug 11, 2023

Fixing bug with number of rows
Jose J. Martinez
committed

Commits on Aug 14, 2023

Adds OpenAI augmentation
Jose J. Martinez
committed

Commits on Aug 16, 2023

Parallel augmentation prototype
Jose J. Martinez
committed
Adds batch size to augment
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Adds numpy
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Fixes bug
Jose J. Martinez
committed
Args to kwargs
Jose J. Martinez
committed
Args to kwargs
Jose J. Martinez
committed
Args to kwargs
Jose J. Martinez
committed
Removes frequence penalty
Jose J. Martinez
committed
Removes frequence penalty
Jose J. Martinez
committed
Payload response fix
Jose J. Martinez
committed
Payload response fix
Jose J. Martinez
committed
Payload response fix
Jose J. Martinez
committed
Payload response fix
Jose J. Martinez
committed
Payload response fix
Jose J. Martinez
committed
Fixing uuid
Jose J. Martinez
committed
Adds inspiration
Jose J. Martinez
committed
Debugs
Jose J. Martinez
committed
Debugs
Jose J. Martinez
committed
Asks to reformat the quotes
Jose J. Martinez
committed
Asks to reformat the quotes
Jose J. Martinez
committed
Moves to a csv format
Jose J. Martinez
committed
Moves to a csv format
Jose J. Martinez
committed
Moves to a csv format
Jose J. Martinez
committed
Moves to a csv format
Jose J. Martinez
committed
Moves to a csv format
Jose J. Martinez
committed
Json asking to escape quotes
Jose J. Martinez
committed
Json asking to escape quotes
Jose J. Martinez
committed
Json asking to escape quotes
Jose J. Martinez
committed
Json asking to escape quotes
Jose J. Martinez
committed
Json asking to escape quotes
Jose J. Martinez
committed
Json asking to escape quotes
Jose J. Martinez
committed
Json asking to escape quotes
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed
Printing what is generating
Jose J. Martinez
committed

Commits on Aug 17, 2023

Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed
Adds gradient clipping and cosine optimizer
Jose J. Martinez
committed

Commits on Aug 18, 2023

Freezing bias
Jose J. Martinez
committed
Freezing bias
Jose J. Martinez
committed
Freezing bias
Jose J. Martinez
committed

Commits on Aug 19, 2023

Removes cosine scheduler
Jose J. Martinez
committed
Removes cosine scheduler
Jose J. Martinez
committed
Removes cosine scheduler
Jose J. Martinez
committed
Sending metadata
Jose J. Martinez
committed
Refactors
Jose J. Martinez
committed
Refactors
Jose J. Martinez
committed
Adds sleep
Jose J. Martinez
committed
Adds sleep
Jose J. Martinez
committed
Adds sleep
Jose J. Martinez
committed
Adds sleep
Jose J. Martinez
committed
Adds sleep
Jose J. Martinez
committed
Sends one by one
Jose J. Martinez
committed
Sends one by one
Jose J. Martinez
committed
Sends one by one
Jose J. Martinez
committed
Sends one by one
Jose J. Martinez
committed
Changes from static to global
Jose J. Martinez
committed
Removes param
Jose J. Martinez
committed
Removes param
Jose J. Martinez
committed
Removes param
Jose J. Martinez
committed
Checks error
Jose J. Martinez
committed
Fixes bug with metadata field name
Jose J. Martinez
committed
Adds JsonParser
Jose J. Martinez
committed
Adds JsonParser
Jose J. Martinez
committed
Removes sleep
Jose J. Martinez
committed
Prevents locks
Jose J. Martinez
committed
Write > Append
Jose J. Martinez
committed
Write > Append
Jose J. Martinez
committed
Adds different schedulers
Jose J. Martinez
committed
Parametrizes temperature
Jose J. Martinez
committed
Fixes schedule name bug
Jose J. Martinez
committed
Fixes schedule name bug
Jose J. Martinez
committed
Fixes schedule name bug
Jose J. Martinez
committed
Refactors and adds augment script
Jose J. Martinez
committed
Adds 25 concurrent calls by default
Jose J. Martinez
committed
Adds 25 concurrent calls by default
Jose J. Martinez
committed
Adds 25 concurrent calls by default
Jose J. Martinez
committed
Adds 25 concurrent calls by default
Jose J. Martinez
committed

Commits on Aug 21, 2023

Adds more examples
Jose J. Martinez
committed

Commits on Aug 22, 2023

Adds scheduler type
Jose J. Martinez
committed

Commits on Aug 23, 2023

Freezes everything except weights
Jose J. Martinez
committed

Commits on Aug 24, 2023

Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Changes threshold, evaluation on tags, freezing backbone
Jose J. Martinez
committed
Adds filtering by tags
Jose J. Martinez
committed
Adds edge case for the remaining X end texts to augment
Jose J. Martinez
committed
Removes metrics from tags absent in training
Jose J. Martinez
committed
Removes metrics from tags absent in training
Jose J. Martinez
committed

Commits on Aug 25, 2023

Rolls back filtering tags in metrics
Jose J. Martinez
committed
Adds back tag filtering
Jose J. Martinez
committed
Adds weight_decay, correct_bias, dropout probs, attention dropout...
Jose J. Martinez
committed
Adds weight_decay, correct_bias, dropout probs, attention dropout...
Jose J. Martinez
committed
Adds best params
josejuanmartinez
committed

Commits on Aug 26, 2023

Updates resume_train_by_steps.sh
Jose J. Martinez
committed
Updates resume_train_by_steps.sh
Jose J. Martinez
committed

Commits on Aug 28, 2023

Adds tags-based augmentation
Jose J. Martinez
committed
Adds id2label for augmentation
Jose J. Martinez
committed
Adds dataset folder
Jose J. Martinez
committed
Decodes id back into labels
Jose J. Martinez
committed
Decodes id back into labels
Jose J. Martinez
committed
Generates examples also for not-underrepresented
Jose J. Martinez
committed
Generates examples also for not-underrepresented
Jose J. Martinez
committed
Generates examples also for not-underrepresented
Jose J. Martinez
committed
Adds more columns to preprocessing
Jose J. Martinez
committed
Adds more columns to preprocessing
Jose J. Martinez
committed
Adds more columns to preprocessing
Jose J. Martinez
committed
Adds more columns to preprocessing
Jose J. Martinez
committed
Adds more columns to preprocessing
Jose J. Martinez
committed
Prevents crashes
Jose J. Martinez
committed
Better hyperparams
Jose J. Martinez
committed

Commits on Aug 31, 2023

Check if fixes tests
Jose J. Martinez
committed
Tries to fix torch-cpu recent issue
Jose J. Martinez
committed

Commits on Sep 1, 2023

Refactors augmentation
Jose J. Martinez
committed
Refactors augmentation
Jose J. Martinez
committed
Refactors augmentation
Jose J. Martinez
committed
Fixes tests
Jose J. Martinez
committed
Fixes black
Jose J. Martinez
committed
Fixes black
Jose J. Martinez
committed
Fixes ruff
Jose J. Martinez
committed
Fixes ruff
Jose J. Martinez
committed
Fixes black
Jose J. Martinez
committed
Fixes black
Jose J. Martinez
committed
Merge pull request #10 from MantisAI/3-improve-trainer-juan
josejuanmartinez
authored

Commits on Sep 4, 2023

Merge branch 'wellcometrust:main' into main
josejuanmartinez
authored

Commits on Sep 6, 2023

Modify script to include active portfolio sample
Ubuntu
committed
Update data
Ubuntu
committed
Merge pull request #14 from MantisAI/active-portfolio-sample
nsorros
authored