8000 Commits · mayanksuman/kaldi · GitHub

More Web Proxy on the site http://driver.im/

Commits

Commits on Jan 11, 2017

[src] nnet3: fixed bug (LstmNonlinearityComponent::ZeroStats() was not implemented) (#1302)

freewym
authored and
danpovey
committed

Commits on Dec 6, 2016

minor bug fix in XconfigAffineLayer when initializing param-stddev
freewym
committed

Commits on Dec 2, 2016

changed default value of zeroing-threshold in BackpropTruncationComponent to 15; updated the results on AMI (#1240)

freewym
authored and
danpovey
committed

Commits on Nov 29, 2016

Commits on Nov 28, 2016

some scripts fixes in blstm_6j and tdnn_lstm recipes (#1220)

freewym
authored and
vijayaditya
committed

Commits on Nov 17, 2016

Commits on Nov 16, 2016

fixed max_deriv_time unset issue for BLSTM
freewym
committed

Commits on Nov 13, 2016

WIP: BackpropTruncationComponent (#1126)

freewym
authored and
danpovey
committed

Commits on Oct 27, 2016

printed out stats for self-repair in nonlinearity (self-repaired-proportion) (#1141)

freewym
authored and
danpovey
committed

Commits on Oct 26, 2016

Max change per component (#918)

freewym
authored and
danpovey
committed

Commits on Oct 4, 2016

added the latest cleanup code and left-biphone models to the librispeech recipe (#991)

freewym
authored and
vijayaditya
committed

Commits on Sep 27, 2016

fixed an arg-parsing bug introduced from a previous commit which will cause tdnn/make_configs.py to crash (#1067)

freewym
authored and
danpovey
committed

Commits on Sep 26, 2016

introduced a new splice configuration for tdnn+xent and tdnn+chain on swbd as default and updateed their improved results (#1035)

freewym
authored and
vijayaditya
committed

Commits on Aug 24, 2016

change option name self-repair-scale to self-repair-scale-nonlinearity for all tdnn recipes (#1002)

freewym
authored and
vijayaditya
committed

Commits on Aug 17, 2016

added assert for the matrix dimension agreement in CuMatrixBase::AddMatMatElements()
freewym
committed

Commits on Aug 11, 2016

Fixed some bugs in librispeech tdnn+{xent,chain}+sMBR recipes and added (#960)

freewym
authored and
vijayaditya
committed

Commits on Jul 30, 2016

fix a bug in ClipGradientComponent::RepairGradients(), which may cause "divided by 0" problem
freewym
committed

Commits on Jul 25, 2016

changes in ClipGradientComponent::Read() to ensure back-compatibility when reading old models
freewym
committed

Commits on Jul 17, 2016

introduce a new splice configuration for tdnn+xent on swbd as default and its improved results
freewym
committed

Commits on Jul 15, 2016

Shrink in-value of ClipGradientComponent toward some smaller value when clipping proportion exceeds some threshold (#803)

freewym
authored and
vijayaditya
committed

Commits on Jul 13, 2016

add an option to set random seed for model initialization and egs shu… (#892)

freewym
authored and
vijayaditya
committed

Commits on Jun 22, 2016

add AddBLstmLayer function to components.py (#856)

freewym
authored and
vijayaditya
committed

Commits on May 17, 2016

Commits on Apr 17, 2016

force to generate utt2dur for every srcdir while calling combine_data.sh
freewym
committed

Commits on Apr 12, 2016

get duration from feats.scp when wav.scp or segments are absent
freewym
committed

Commits on Apr 11, 2016

fix
freewym
committed

Commits on Apr 10, 2016

read entire file according to if there are sox commands with speed perturb
freewym
committed

Commits on Apr 9, 2016

fix for utt2dur when applied on speed-pertubed data
freewym
committed

Commits on Mar 11, 2016

add chain tdnn AMI recipe
freewym
committed

Commits on Jan 2, 2016

a minor bug in cu-sparse-matrix.cc
freewym
committed

Commits on Dec 31, 2015

Add more results of BLSTM on swbd
freewym
committed

Commits on Dec 1, 2015

some fixes for BLSTM
freewym
committed

0