diff --git a/egs/librispeech/s5/local/chain/run_cnn_tdnn.sh b/egs/librispeech/s5/local/chain/run_cnn_tdnn.sh
new file mode 100755
index 00000000000..cd8f38d8309
--- /dev/null
+++ b/egs/librispeech/s5/local/chain/run_cnn_tdnn.sh
@@ -0,0 +1 @@
+tuning/run_cnn_tdnn_1a.sh
diff --git a/egs/librispeech/s5/local/chain/tuning/run_cnn_tdnn_1a.sh b/egs/librispeech/s5/local/chain/tuning/run_cnn_tdnn_1a.sh
new file mode 100755
index 00000000000..2a60587fc35
--- /dev/null
+++ b/egs/librispeech/s5/local/chain/tuning/run_cnn_tdnn_1a.sh
@@ -0,0 +1,278 @@
+#!/bin/bash
+
+# This is based on tdnn_1d_sp, but adding cnn as the front-end.
+# The cnn-tdnn-f (tdnn_cnn_1a_sp) outperforms the tdnn-f (tdnn_1d_sp).
+
+# bash local/chain/compare_wer.sh exp/chain_cleaned/tdnn_1d_sp exp/chain_cleaned/tdnn_cnn_1a_sp/
+# System                         tdnn_1d_sp  tdnn_cnn_1a_sp
+# WER on dev(fglarge)               3.29          3.34
+# WER on dev(tglarge)               3.44          3.39
+# WER on dev(tgmed)                 4.22          4.29
+# WER on dev(tgsmall)               4.72          4.77
+# WER on dev_other(fglarge)         8.71          8.62
+# WER on dev_other(tglarge)         9.05          9.00
+# WER on dev_other(tgmed)          11.09         10.93
+# WER on dev_other(tgsmall)        12.13         12.02
+# WER on test(fglarge)              3.80          3.69
+# WER on test(tglarge)              3.89          3.80
+# WER on test(tgmed)                4.72          4.64
+# WER on test(tgsmall)              5.19          5.16      
+# WER on test_other(fglarge)        8.76          8.71
+# WER on test_other(tglarge)        9.19          9.11
+# WER on test_other(tgmed)         11.22         11.00
+# WER on test_other(tgsmall)       12.24         12.16
+# Final train prob               -0.0378       -0.0420
+# Final valid prob               -0.0374       -0.0400
+# Final train prob (xent)        -0.6099       -0.6881
+# Final valid prob (xent)        -0.6353       -0.7180
+# Num-parameters                22623456      18100736
+
+
+set -e
+
+# configs for 'chain'
+stage=0
+decode_nj=50
+train_set=train_960_cleaned
+gmm=tri6b_cleaned
+nnet3_affix=_cleaned
+
+# The rest are configs specific to this script.  Most of the parameters
+# are just hardcoded at this level, in the commands below.
+affix=cnn_1a
+tree_affix=
+train_stage=-10
+get_egs_stage=-10
+decode_iter=
+
+# TDNN options
+frames_per_eg=150,110,100
+remove_egs=true
+common_egs_dir=
+xent_regularize=0.1
+dropout_schedule='0,0@0.20,0.5@0.50,0'
+
+test_online_decoding=true  # if true, it will run the last decoding stage.
+
+# End configuration section.
+echo "$0 $@"  # Print the command line for logging
+
+. ./cmd.sh
+. ./path.sh
+. ./utils/parse_options.sh
+
+if ! cuda-compiled; then
+  cat <<EOF && exit 1
+This script is intended to be used with GPUs but you have not compiled Kaldi with CUDA
+If you want to use GPUs (and have them), go to src/, and configure and make on a machine
+where "nvcc" is installed.
+EOF
+fi
+
+# The iVector-extraction and feature-dumping parts are the same as the standard
+# nnet3 setup, and you can skip them by setting "--stage 11" if you have already
+# run those things.
+
+local/nnet3/run_ivector_common.sh --stage $stage \
+                                  --train-set $train_set \
+                                  --gmm $gmm \
+                                  --num-threads-ubm 6 --num-processes 3 \
+                                  --nnet3-affix "$nnet3_affix" || exit 1;
+
+gmm_dir=exp/$gmm
+ali_dir=exp/${gmm}_ali_${train_set}_sp
+tree_dir=exp/chain${nnet3_affix}/tree_sp${tree_affix:+_$tree_affix}
+lang=data/lang_chain
+lat_dir=exp/chain${nnet3_affix}/${gmm}_${train_set}_sp_lats
+dir=exp/chain${nnet3_affix}/tdnn${affix:+_$affix}_sp
+train_data_dir=data/${train_set}_sp_hires
+lores_train_data_dir=data/${train_set}_sp
+train_ivector_dir=exp/nnet3${nnet3_affix}/ivectors_${train_set}_sp_hires
+
+# if we are using the speed-perturbed data we need to generate
+# alignments for it.
+
+for f in $gmm_dir/final.mdl $train_data_dir/feats.scp $train_ivector_dir/ivector_online.scp \
+    $lores_train_data_dir/feats.scp $ali_dir/ali.1.gz; do
+  [ ! -f $f ] && echo "$0: expected file $f to exist" && exit 1
+done
+
+# Please take this as a reference on how to specify all the options of
+# local/chain/run_chain_common.sh
+local/chain/run_chain_common.sh --stage $stage \
+                                --gmm-dir $gmm_dir \
+                                --ali-dir $ali_dir \
+                                --lores-train-data-dir ${lores_train_data_dir} \
+                                --lang $lang \
+                                --lat-dir $lat_dir \
+                                --num-leaves 7000 \
+                                --tree-dir $tree_dir || exit 1;
+
+if [ $stage -le 14 ]; then
+  echo "$0: creating neural net configs using the xconfig parser";
+
+  num_targets=$(tree-info $tree_dir/tree | grep num-pdfs | awk '{print $2}')
+  learning_rate_factor=$(echo "print 0.5/$xent_regularize" | python)
+  cnn_opts="l2-regularize=0.01"
+  ivector_affine_opts="l2-regularize=0.0"
+  affine_opts="l2-regularize=0.008 dropout-proportion=0.0 dropout-per-dim=true dropout-per-dim-continuous=true"
+  tdnnf_first_opts="l2-regularize=0.008 dropout-proportion=0.0 bypass-scale=0.0"
+  tdnnf_opts="l2-regularize=0.008 dropout-proportion=0.0 bypass-scale=0.75"
+  linear_opts="l2-regularize=0.008 orthonormal-constraint=-1.0"
+  prefinal_opts="l2-regularize=0.008"
+  output_opts="l2-regularize=0.005"
+
+  mkdir -p $dir/configs
+
+  cat <<EOF > $dir/configs/network.xconfig
+  input dim=100 name=ivector
+  input dim=40 name=input
+
+  # MFCC to filterbank
+  idct-layer name=idct input=input dim=40 cepstral-lifter=22 affine-transform-file=$dir/configs/idct.mat
+
+  linear-component name=ivector-linear $ivector_affine_opts dim=200 input=ReplaceIndex(ivector, t, 0)
+  batchnorm-component name=ivector-batchnorm target-rms=0.025
+  batchnorm-component name=idct-batchnorm input=idct
+
+  combine-feature-maps-layer name=combine_inputs input=Append(idct-batchnorm, ivector-batchnorm) num-filters1=1 num-filters2=5 height=40
+  conv-relu-batchnorm-layer name=cnn1 $cnn_opts height-in=40 height-out=40 time-offsets=-1,0,1 height-offsets=-1,0,1 num-filters-out=64
+  conv-relu-batchnorm-layer name=cnn2 $cnn_opts height-in=40 height-out=40 time-offsets=-1,0,1 height-offsets=-1,0,1 num-filters-out=64
+  conv-relu-batchnorm-layer name=cnn3 $cnn_opts height-in=40 height-out=20 height-subsample-out=2 time-offsets=-1,0,1 height-offsets=-1,0,1 num-filters-out=128
+  conv-relu-batchnorm-layer name=cnn4 $cnn_opts height-in=20 height-out=20 time-offsets=-1,0,1 height-offsets=-1,0,1 num-filters-out=128
+  conv-relu-batchnorm-layer name=cnn5 $cnn_opts height-in=20 height-out=10 height-subsample-out=2 time-offsets=-1,0,1 height-offsets=-1,0,1 num-filters-out=256
+  conv-relu-batchnorm-layer name=cnn6 $cnn_opts height-in=10 height-out=10 time-offsets=-1,0,1 height-offsets=-1,0,1 num-filters-out=256
+
+  # the first TDNN-F layer has no bypass
+  tdnnf-layer name=tdnnf7 $tdnnf_first_opts dim=1536 bottleneck-dim=256 time-stride=0
+  tdnnf-layer name=tdnnf8 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf9 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf10 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf11 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf12 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf13 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf14 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf15 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf16 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf17 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  tdnnf-layer name=tdnnf18 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
+  linear-component name=prefinal-l dim=256 $linear_opts
+
+  prefinal-layer name=prefinal-chain input=prefinal-l $prefinal_opts big-dim=1536 small-dim=256
+  output-layer name=output include-log-softmax=false dim=$num_targets $output_opts
+
+  prefinal-layer name=prefinal-xent input=prefinal-l $prefinal_opts big-dim=1536 small-dim=256
+  output-layer name=output-xent dim=$num_targets learning-rate-factor=$learning_rate_factor $output_opts
+EOF
+
+  steps/nnet3/xconfig_to_configs.py --xconfig-file $dir/configs/network.xconfig --config-dir $dir/configs/
+fi
+
+if [ $stage -le 15 ]; then
+  if [[ $(hostname -f) == *.clsp.jhu.edu ]] && [ ! -d $dir/egs/storage ]; then
+    utils/create_split_dir.pl \
+     /export/b{09,10,11,12}/$USER/kaldi-data/egs/swbd-$(date +'%m_%d_%H_%M')/s5c/$dir/egs/storage $dir/egs/storage
+  fi
+
+  steps/nnet3/chain/train.py --stage $train_stage \
+    --use-gpu "wait" \
+    --cmd "$decode_cmd" \
+    --feat.online-ivector-dir $train_ivector_dir \
+    --feat.cmvn-opts "--norm-means=false --norm-vars=false" \
+    --chain.xent-regularize $xent_regularize \
+    --chain.leaky-hmm-coefficient 0.1 \
+    --chain.l2-regularize 0.0 \
+    --chain.apply-deriv-weights false \
+    --chain.lm-opts="--num-extra-lm-states=2000" \
+    --egs.dir "$common_egs_dir" \
+    --egs.stage $get_egs_stage \
+    --egs.opts "--frames-overlap-per-eg 0 --constrained false" \
+    --egs.chunk-width $frames_per_eg \
+    --trainer.dropout-schedule $dropout_schedule \
+    --trainer.add-option="--optimization.memory-compression-level=2" \
+    --trainer.num-chunk-per-minibatch 64 \
+    --trainer.frames-per-iter 2500000 \
+    --trainer.num-epochs 4 \
+    --trainer.optimization.num-jobs-initial 3 \
+    --trainer.optimization.num-jobs-final 16 \
+    --trainer.optimization.initial-effective-lrate 0.00015 \
+    --trainer.optimization.final-effective-lrate 0.000015 \
+    --trainer.max-param-change 2.0 \
+    --cleanup.remove-egs $remove_egs \
+    --feat-dir $train_data_dir \
+    --tree-dir $tree_dir \
+    --lat-dir $lat_dir \
+    --dir $dir  || exit 1;
+
+fi
+
+graph_dir=$dir/graph_tgsmall
+if [ $stage -le 16 ]; then
+  # Note: it might appear that this $lang directory is mismatched, and it is as
+  # far as the 'topo' is concerned, but this script doesn't read the 'topo' from
+  # the lang directory.
+  utils/mkgraph.sh --self-loop-scale 1.0 --remove-oov data/lang_test_tgsmall $dir $graph_dir
+  # remove <UNK> from the graph, and convert back to const-FST.
+  fstrmsymbols --apply-to-output=true --remove-arcs=true "echo 3|" $graph_dir/HCLG.fst - | \
+    fstconvert --fst_type=const > $graph_dir/temp.fst
+  mv $graph_dir/temp.fst $graph_dir/HCLG.fst
+fi
+
+iter_opts=
+if [ ! -z $decode_iter ]; then
+  iter_opts=" --iter $decode_iter "
+fi
+if [ $stage -le 17 ]; then
+  rm $dir/.error 2>/dev/null || true
+  for decode_set in test_clean test_other dev_clean dev_other; do
+      (
+      steps/nnet3/decode.sh --acwt 1.0 --post-decode-acwt 10.0 \
+          --nj $decode_nj --cmd "$decode_cmd" $iter_opts \
+          --online-ivector-dir exp/nnet3${nnet3_affix}/ivectors_${decode_set}_hires \
+          $graph_dir data/${decode_set}_hires $dir/decode_${decode_set}${decode_iter:+_$decode_iter}_tgsmall || exit 1
+      steps/lmrescore.sh --cmd "$decode_cmd" --self-loop-scale 1.0 data/lang_test_{tgsmall,tgmed} \
+          data/${decode_set}_hires $dir/decode_${decode_set}${decode_iter:+_$decode_iter}_{tgsmall,tgmed} || exit 1
+      steps/lmrescore_const_arpa.sh \
+          --cmd "$decode_cmd" data/lang_test_{tgsmall,tglarge} \
+          data/${decode_set}_hires $dir/decode_${decode_set}${decode_iter:+_$decode_iter}_{tgsmall,tglarge} || exit 1
+      steps/lmrescore_const_arpa.sh \
+          --cmd "$decode_cmd" data/lang_test_{tgsmall,fglarge} \
+          data/${decode_set}_hires $dir/decode_${decode_set}${decode_iter:+_$decode_iter}_{tgsmall,fglarge} || exit 1
+      ) || touch $dir/.error &
+  done
+  wait
+  if [ -f $dir/.error ]; then
+    echo "$0: something went wrong in decoding"
+    exit 1
+  fi
+fi
+
+if $test_online_decoding && [ $stage -le 18 ]; then
+  # note: if the features change (e.g. you add pitch features), you will have to
+  # change the options of the following command line.
+  steps/online/nnet3/prepare_online_decoding.sh \
+       --mfcc-config conf/mfcc_hires.conf \
+       $lang exp/nnet3${nnet3_affix}/extractor $dir ${dir}_online
+
+  rm $dir/.error 2>/dev/null || true
+  for data in test_clean test_other dev_clean dev_other; do
+    (
+      nspk=$(wc -l <data/${data}_hires/spk2utt)
+      # note: we just give it "data/${data}" as it only uses the wav.scp, the
+      # feature type does not matter.
+      steps/online/nnet3/decode.sh \
+          --acwt 1.0 --post-decode-acwt 10.0 \
+          --nj $nspk --cmd "$decode_cmd" \
+          $graph_dir data/${data} ${dir}_online/decode_${data}_tgsmall || exit 1
+
+    ) || touch $dir/.error &
+  done
+  wait
+  if [ -f $dir/.error ]; then
+    echo "$0: something went wrong in decoding"
+    exit 1
+  fi
+fi
+
+exit 0;
+