Skip to content
Snippets Groups Projects
Select Git revision
  • master default protected
1 result
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.022Dec2120191816151413121110985432130Nov29272624232221201918151413109876432129Oct2725242322212019171613111094328Sep27262522171513121185131Aug302928252423212018171514121110987543231Jul2624232119171614119876330Jun292827262423222120191615141310872131May292726252423222120191817151311109765432130Apr2928262524232221201917161514131211109542131Mar2928272322201917161413108632128Feb2725242322212019171513111098762131Jan302927262524232223222120191817[scripts] Fixing bug in fast-lstm-layer and lstmb-layer whereby c was used instead of m for affine transform.[doc] Fix typo in docs (#2101)[egs] Fix to decode context option in swbd (#2100)[src] Fix bug in compilation with Scale() expressions. (#2098)[src] Fix bug in compilation with Scale() expressions.[src] Simplify/refactor natural gradient code in nnet3[src] Prevent RNNLM training from exhausting GPU memory in diagnostic jobs (#2094)[src,egs,scripts] Merging RNNLM-related changes which were in wrong branch (#2092)[scripts,egs] Some fixes to OPGRU scripts from Lu Huang (#2091)Merge remote-tracking branch 'upstream/master' into reorthonormalize-with-memonorm[src] Fix optimization bug in nnet3, regarding Scale() expressions>[scripts] Cosmetic change: add message[src] Fix bug in nnet3 optimization, affecting Scale() operation; cosmetic fixes. (#2088)[src] Some refactoring of lstmb layer: using memnorm.[src] Fix bug in LinearComponent whereby use-natural-gradient defaulted to false.[scripts] Some changes to lstm.py which I don't want to keep all of, but backing up via git.Merge remote-tracking branch 'origin/memory-norm-with-combine' into reorthonormalize-with-memonormMerge remote-tracking branch 'upstream/master'[src,scripts] Add extra nnet3 diagnostics; add reothonormalize option for LinearComponent[egs] Mac compatibility fix to SGMM+MMI: remove -T option to cp (#2087)[scripts] Script fix; update batchnorm(/memnorm) stats faster in first part of training.Merge remote-tracking branch 'upstream/master' into memory-norm-with-combine[src] Simplify MemoryNormComponent and always store stats during backprop on nnet_to_store_stats_[egs] Copy dictionary-preparation-script fix from fisher-english(8e7793f) to fisher-swbd and ami (#2084)[src] Simplify NormalizeLayer.[egs] Small fix to backstitch in AMI scripts (#2083)[scripts] Fix augment_data_dir.py (relates to non-pipe case of wav.scp) (#2081)Merge branch 'model_combine' of https://github.com/freewym/kaldi into memory-norm-with-combine[src] Changes to how MemoryNormComponent behaves (use more up-to-date stats)fixdocs fixesfix[src,scripts] Changing which model stats are accumulated on (will affect memory-norm). Other small changes to MemoryNormComponent, will rework most of this. Adding script changes for memory-norm.[egs,scripts] Add OPGRU scripts and recipes (#1950)fix Makefilepython-level changes, added more documentations.changes according to the review[egs] Add an l2-regularize-based recipe for image recognition setups (#2066)[src] Bug-fix to assertion in cu-sparse-matrix.cc (RE large matrices) (#2077)[src] Some reorganizations of MemoryNormComponent code