Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 17, 2020

• added further details to the readme

  KamWithK

  authored
  1c92597

  Copy full SHA for 1c92597

Commits on Feb 15, 2020

• added final MobileNet V2 model trained with Super-Convergence and LDAM loss

  

  DeepLearning VM

  committed
  9bea38a

  Copy full SHA for 9bea38a

Commits on Feb 14, 2020

• corrected code to allow loading of custom data and model simultaniouslly (instead of just what was bundled with the model's saves)

  

  DeepLearning VM

  committed
  062a5aa

  Copy full SHA for 062a5aa

• added further comments to aid documentation

  

  DeepLearning VM

  committed
  d78ae4a

  Copy full SHA for d78ae4a

• added link to Android app

  KamWithK

  authored
  b27b0ea

  Copy full SHA for b27b0ea

Commits on Feb 13, 2020

• added the best found max learning rate for LDAM loss (with MobileNet V2)

  

  DeepLearning VM

  committed
  6b867ce

  Copy full SHA for 6b867ce

• added LDAM loss

  

  DeepLearning VM

  committed
  fd6fc24

  Copy full SHA for fd6fc24

• replaced squeezenet model with a resnext50 model

  

  DeepLearning VM

  committed
  0af5001

  Copy full SHA for 0af5001

• added test dataset transforms

  

  DeepLearning VM

  committed
  7a2ec47

  Copy full SHA for 7a2ec47

• use label names instead of numbers for scikit learn evaluation report

  

  DeepLearning VM

  committed
  6d16829

  Copy full SHA for 6d16829

• always save all evaluation data which is calulated (instead of just confusion matrix)

  

  DeepLearning VM

  committed
  05b240f

  Copy full SHA for 05b240f

• changed code to always calculate weights and label counts in case used elsewhere

  

  DeepLearning VM

  committed
  8c189ef

  Copy full SHA for 8c189ef

• converted certain variables into properties for access by classes which use organiser's

  

  DeepLearning VM

  committed
  26c9dad

  Copy full SHA for 26c9dad

• made balancing data optional

  

  DeepLearning VM

  committed
  83cb8c6

  Copy full SHA for 83cb8c6

Commits on Feb 4, 2020

• updated evaluation function for greater flexibility

  

  DeepLearning VM

  committed
  471f6f0

  Copy full SHA for 471f6f0

• corrected typo in variable names

  

  DeepLearning VM

  committed
  003ef51

  Copy full SHA for 003ef51

• added ability to balance data classes

  

  DeepLearning VM

  committed
  0e5cded

  Copy full SHA for 0e5cded

Commits on Jan 22, 2020

• Further specified about unique PyTorch techniques

  KamWithK

  authored
  be47e65

  Copy full SHA for be47e65

• created readme file

  KamWithK

  authored
  7458d4a

  Copy full SHA for 7458d4a

Commits on Jan 21, 2020

• ensured smooth model loading when moving between CPU's and GPU's

  

  DeepLearning VM

  committed
  76dd2e4

  Copy full SHA for 76dd2e4

• made loading image information more modular and pythonic

  

  DeepLearning VM

  committed
  3ed33c2

  Copy full SHA for 3ed33c2

• added function to create executable TorchScript module

  

  DeepLearning VM

  committed
  66f14d3

  Copy full SHA for 66f14d3

Commits on Jan 16, 2020

• modified network last layers

  

  DeepLearning VM

  committed
  5818e21

  Copy full SHA for 5818e21

• major code refactor

  

  DeepLearning VM

  committed
  1139177

  Copy full SHA for 1139177

Commits on Jan 11, 2020

• removed redundent variables for path

  

  DeepLearning VM

  committed
  d14dccb

  Copy full SHA for d14dccb

• finetuned hyperparameters

  

  DeepLearning VM

  committed
  006924d

  Copy full SHA for 006924d

Commits on Jan 10, 2020

• added a learning rate finder to get the best learning rate

  

  DeepLearning VM

  committed
  9ac8524

  Copy full SHA for 9ac8524

• corrected imports and variable naming

  

  DeepLearning VM

  committed
  1da2405

  Copy full SHA for 1da2405

• add more image augmentation

  KamWithK

  committed
  47095a7

  Copy full SHA for 47095a7

• call the scheduler after every training batch instead of validation loop (as required for the one-cycle policy)

  KamWithK

  committed
  38eb4d2

  Copy full SHA for 38eb4d2

• correct minor bugs in code

  KamWithK

  committed
  b28e5ac

  Copy full SHA for b28e5ac

Commits on Jan 9, 2020

• moved data splitting away from the Trainer class

  KamWithK

  committed
  90ddef7

  Copy full SHA for 90ddef7

• added experimental evaluation code

  

  DeepLearning VM

  committed
  889ba80

  Copy full SHA for 889ba80

• change hyperparameters for super convergence and removed extra TransferLearning class

  

  DeepLearning VM

  committed
  676d6c2

  Copy full SHA for 676d6c2

• modularised data loading

  

  DeepLearning VM

  committed
  f67d599

  Copy full SHA for f67d599

Pagination

Previous

Next