Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • add_default_vectorizer_and_pretrained_loading
  • change_system_act_in_env
  • clean_code
  • convlab_exp
  • crossowoz_ddpt
  • e2e-soloist
  • emoUS
  • generalized_action_dicts
  • issue127
  • issue_114
  • master default protected
  • nlg-scgpt
  • ppo_num_dialogues
  • pre-training
  • readme
  • remapping_actions
  • robust_masking_feature
  • scgpt_exp
  • soloist
19 results
Created with Raphaël 2.2.02Sep131Aug26252322181716151110985432323232129Jul27262522212019181715141312865430Jun2928272623222120141310832130May27242317161312111076530Apr29272625242221201918151413111098122Mar211918171510932128Feb262523191817161514102130Jan2624211913731Dec2827222122212220191514109872130Nov2928272413Sep19Aug30Jul718Jun1720May729Apr16215Mar1428Jan29Dec18Nov131Oct30292827262218171513762130Sep29282725242322sample normally from buffer when there is no regularizationneed to set NaN gradients to zero in vtraceweight regularization towards random policy to keep some amount of randomnesssaving policy config properly, not using domain prob in entropyoption to only use multiwoz like domains in sgdMerge branch 'setsumbt' into 'master'Setsumbtoption to not use slots when calculating log probsadded possibility to pre-load model during supervised trainingadded entropy loss to supervised trainingadded tqdm when loading data for supervised training and using timeout when trying to download databaseusing 40 epochs for supervised training, achieving 89% success and fixed a bug in action embedderprint loss during supervised trainingadded embeddings for multiwozto device..first version where SGD data set is running, now have to test whether it learns properlyfirst version of supervised training, need to test if it works for multiwoz and then for sgddeleted dialcrowd codeMerge branch 'master' into github_masterMerge branch 'GenTUS' of gitlab.cs.uni-duesseldorf.de:dsml/convlab/ConvLab3 into GenTUSfix typoMerge branch 'GenTUS' of gitlab.cs.uni-duesseldorf.de:dsml/convlab/ConvLab3 into GenTUStypoconflictzimrefactor gentus trainingsave two modelswipadd utiladd token mapadd kgadd gentus vectoradd agentconvert stepGenTUS to unify dataformetfix merge datasetfix typoremove unused tus functionMerge branch 'tus_training' into GenTUSclear unused argumentsupdate model name