See also http://www.danielpovey.com/files/2017_interspeech_dropout.pdf
this improves on the best recipes.

This is a fix so the --silence-weighting option will work correctly when decoding online with chain models.  Navneeth tested it on a couple of setups so we believe it works.

[srcipts] steps/nnet3/report/generate_plots.py: plot 5,50,95th percentile of value and derivative instead of mean+-stddev (#1472)

* [scripts,egs] Adding options for using PCA instead of LDA+MLLT for ivectors used in ASR. Results are reported in the default TDNN recipe in AMI. Updating steps/online/nnet2/{train_diag_ubm.sh,train_ivector_extractor.sh} so that they now backup the contents of their destination directory if it already exists.

* [egs,scripts] Updating AMI TDNN results to reflect the current recipe (tdnn1d). Fixing minor bug in egs/ami/s5b/local/chain/tuning/run_tdnn_*.sh scripts.

* [egs] Updating chain scripts in AMI so that they do not default to keeping egs

[egs,scripts,src] SID and LID tools and scripts: cosmetic improvements, better error-handling, and various minor fixes; results unchanged. (#1543)

[src] Change ConvertStringToReal to be locale-independent (i.e. always-US).  Fixes android issue. (#1513)

The issue was that in Android you can't easily override the phone's locale, so export LC_ALL=C does not really work to enable reading of models if the phone was (e.g.) Brazilian or German.

xconfig : Added delay option for FixedAffineLayer. This will be used for ensuring the model specified in ref.config has at least the context required by the model specified in init.config

[src] Fix copy-feats for using the --write-num-frames and --compress true flags at the same time (#1541)

- we auto-detect the 'compute capability' problems (these appear as the 'invalid device function'),
- we also provide guidelines what to try before posting to forum, and which info to send to us,

CuVector::AddColSumMat<float>[no-trans],    16  0.0057  0.0172 3.01x
    CuVector::AddColSumMat<float>[no-trans],    32  0.0242  0.0668 2.76x
    CuVector::AddColSumMat<float>[no-trans],    64  0.0992  0.2577 2.60x
    CuVector::AddColSumMat<float>[no-trans],   128  0.3747  0.9280 2.48x
    CuVector::AddColSumMat<float>[no-trans],   256  1.4711  3.0541 2.08x
    CuVector::AddColSumMat<float>[no-trans],   512  5.1709  9.4713 1.83x
    CuVector::AddColSumMat<float>[no-trans],  1024 12.4352 20.4517 1.64x
    CuVector::AddColSumMat<double>[no-trans],   16  0.0060  0.0175 2.91x
    CuVector::AddColSumMat<double>[no-trans],   32  0.0240  0.0672 2.80x
    CuVector::AddColSumMat<double>[no-trans],   64  0.1006  0.2712 2.70x
    CuVector::AddColSumMat<double>[no-trans],  128  0.3691  0.9097 2.46x
    CuVector::AddColSumMat<double>[no-trans],  256  1.4530  3.1044 2.14x
    CuVector::AddColSumMat<double>[no-trans],  512  4.4524  7.5872 1.70x
    CuVector::AddColSumMat<double>[no-trans], 1024 11.1212 16.1423 1.45x

It appears there may be no good reason to disallow system-wide OpenFst.

[build,src,doc] Modify get_version.sh to deal better with whitespace (avoid space in version); minor fixes (#1526)

[scripts,egs] Adding options for using PCA instead of LDA+MLLT for ivectors used in ASR. Results are reported in the default TDNN recipe in AMI. Updating steps/online/nnet2/{train_diag_ubm.sh,train_ivector_extractor.sh} so that they now backup the contents of their destination directory if it already exists. (#1514)

[src] (minor) Added missing SetZero() to NaturalGradientAffineComponent::Scale() if scale==0.0 (#1522)

this improves speed when using hashes.

[egs] ami : Added tdnn_lstm recipe with fast-lstmp layer. Added tdnn_lstm recipe with -1 delay at lowest lstm layer (#1505)

swbd : Added tdnn_lstm recipe with delay -1 at the lowest lstm layer