local/chain/run_ tdnn.sh：

bash -x local/nnet3/run_ivector_common.sh --stage $stage || exit 1 
# It shows that randomly changing the speed and volume of training data to expand diversity 


if [ $stage -le 7 ]; then
  # Get the alignments as lattices (gives the LF-MMI training more freedom).
  # use the same num-jobs as the alignments Use alignment as a word map file 
  nj=$(cat $ali_dir/num_jobs) || exit 1;
  steps/align_fmllr_lats.sh --nj $nj --cmd "$train_cmd" data/$train_set \
    data/lang exp/tri5a exp/tri5a_sp_lats
  rm exp/tri5a_sp_lats/fsts.*.gz # save space
fi
# Down to 9 Reading a book P237
if [ $stage -le 8 ]; then
  # Create a version of the lang/ directory that has one state per phone in the
  # topo file. [note, it really has two states.. the first one is only repeated
  # once, the second one has zero or more repeats.]
  rm -rf $lang
  cp -r data/lang $lang
  silphonelist=$(cat $lang/phones/silence.csl) || exit 1;
  nonsilphonelist=$(cat $lang/phones/nonsilence.csl) || exit 1;
  # Use our special topology... note that later on may have to tune this
  # topology.
  steps/nnet3/chain/gen_topo.py $nonsilphonelist $silphonelist >$lang/topo
fi

if [ $stage -le 9 ]; then
  # Build a tree using our new topology. This is the critically different
  # step compared with other recipes.
  silphonelist=$(cat $lang/phones/silence.csl) || exit 1;
  nonsilphonelist=$(cat $lang/phones/nonsilence.csl) || exit 1;
  # Use our special topology... note that later on may have to tune this
  # topology.
  steps/nnet3/chain/gen_topo.py $nonsilphonelist $silphonelist >$lang/topo
fi

if [ $stage -le 9 ]; then
  # Build a tree using our new topology. This is the critically different
  # step compared with other recipes.
  steps/nnet3/chain/build_tree.sh --frame-subsampling-factor 3 \
      --context-opts "--context-width=2 --central-position=1" \
      --cmd "$train_cmd" 5000 data/$train_set $lang $ali_dir $treedir
fi


 Definition of model structure , In two parts , One part is the description of the model structure that is easy for users to understand , stay kaldi Called xconfig; The second part is kaldi What is really used in the training process config, Generally, the following script 
steps/nnet3/xconfig_to_configs.py --xconfig-file $dir/configs/network.xconfig --config-dir $dir/configs/ 
 According to the user-defined xconfig produce .
if [ $stage -le 10 ]; then
  echo "$0: creating neural net configs using the xconfig parser";

  num_targets=$(tree-info $treedir/tree |grep num-pdfs|awk '{print $2}')# Number of neural network outputs 
  learning_rate_factor=$(echo "print (0.5/$xent_regularize)" | python)# Learning rate adjustment ratio 



# Decoding part 
graph_dir=$dir/graph
if [ $stage -le 13 ]; then
  for test_set in dev test; do
    steps/nnet3/decode.sh --acwt 1.0 --post-decode-acwt 10.0 \
      --nj 10 --cmd "$decode_cmd" \
      --online-ivector-dir exp/nnet3/ivectors_$test_set \
      $graph_dir data/${test_set}_hires $dir/decode_${test_set} || exit 1;
  done
fi