Merge pull request espnet#3900 from YushiUeda/librimix_diar

Fix issues in EEND-EDA & add Librimix_diar recipe

egs2/TEMPLATE/asr1/db.sh

@@ -42,6 +42,7 @@ SLURP=
 VOXCELEB=
 MINI_LIBRISPEECH=downloads
 MISP2021=
+LIBRIMIX=downloads
 LIBRITTS=
 LJSPEECH=downloads
 NSC=

egs2/TEMPLATE/diar1/scripts/utils/show_diar_result.sh

@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
-mindepth=0
-maxdepth=1
+mindepth=1
+maxdepth=5
 
 . utils/parse_options.sh
 
@@ -44,26 +44,21 @@ cat << EOF
 EOF
 
 while IFS= read -r expdir; do
-    if ls "${expdir}"/*/*/*/scoring/result_* &> /dev/null; then
+#    if ls "${expdir}"/*/*/*/scoring/result_* &> /dev/null; then
+    if ls "${expdir}"/scoring/result_* &> /dev/null; then
+    diardir=${expdir#*/}
         cat << EOF
-## $(basename ${expdir})
+## ${diardir%%/*}
 ### DER
-EOF
-        while IFS= read -r datasetdir; do
-            if ls "${datasetdir}"/scoring/result_* &> /dev/null; then
-                cat << EOF
-$(basename ${datasetdir})
-
+${expdir##*/}
 |threshold_median_collar|DER|
 |---|---|
 EOF
-                for file in "${datasetdir}"/scoring/result_*; do
-                    grep OVER ${file} \
-                        | grep -v nooverlap \
-                        | sed "s/^.*[^0-9]\([0-9]\{1,3\}\.[0-9]\{2\}\).*$/\|$(basename ${file})\|\1\|/"
-                    echo -n
-                done
-            fi
-        done < <(find ${expdir} -mindepth 1 -maxdepth 5 -type d)
+        for file in "${expdir}"/scoring/result_*; do
+            grep OVER ${file} \
+                | grep -v nooverlap \
+                | sed "s/^.*[^0-9]\([0-9]\{1,3\}\.[0-9]\{2\}\).*$/\|$(basename ${file})\|\1\|/"
+            echo -n
+        done
     fi
 done < <(find ${exp} -mindepth ${mindepth} -maxdepth ${maxdepth} -type d)

egs2/librimix/diar1/cmd.sh

@@ -0,0 +1,110 @@
+# ====== About run.pl, queue.pl, slurm.pl, and ssh.pl ======
+# Usage: <cmd>.pl [options] JOB=1:<nj> <log> <command...>
+# e.g.
+#   run.pl --mem 4G JOB=1:10 echo.JOB.log echo JOB
+#
+# Options:
+#   --time <time>: Limit the maximum time to execute.
+#   --mem <mem>: Limit the maximum memory usage.
+#   -–max-jobs-run <njob>: Limit the number parallel jobs. This is ignored for non-array jobs.
+#   --num-threads <ngpu>: Specify the number of CPU core.
+#   --gpu <ngpu>: Specify the number of GPU devices.
+#   --config: Change the configuration file from default.
+#
+# "JOB=1:10" is used for "array jobs" and it can control the number of parallel jobs.
+# The left string of "=", i.e. "JOB", is replaced by <N>(Nth job) in the command and the log file name,
+# e.g. "echo JOB" is changed to "echo 3" for the 3rd job and "echo 8" for 8th job respectively.
+# Note that the number must start with a positive number, so you can't use "JOB=0:10" for example.
+#
+# run.pl, queue.pl, slurm.pl, and ssh.pl have unified interface, not depending on its backend.
+# These options are mapping to specific options for each backend and
+# it is configured by "conf/queue.conf" and "conf/slurm.conf" by default.
+# If jobs failed, your configuration might be wrong for your environment.
+#
+#
+# The official documentation for run.pl, queue.pl, slurm.pl, and ssh.pl:
+#   "Parallelization in Kaldi": http://kaldi-asr.org/doc/queue.html
+# =========================================================~
+
+
+# Select the backend used by run.sh from "local", "stdout", "sge", "slurm", or "ssh"
+cmd_backend='local'
+
+# Local machine, without any Job scheduling system
+if [ "${cmd_backend}" = local ]; then
+
+    # The other usage
+    export train_cmd="run.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="run.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="run.pl"
+
+# Local machine logging to stdout and log file, without any Job scheduling system
+elif [ "${cmd_backend}" = stdout ]; then
+
+    # The other usage
+    export train_cmd="stdout.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="stdout.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="stdout.pl"
+
+
+# "qsub" (Sun Grid Engine, or derivation of it)
+elif [ "${cmd_backend}" = sge ]; then
+    # The default setting is written in conf/queue.conf.
+    # You must change "-q g.q" for the "queue" for your environment.
+    # To know the "queue" names, type "qhost -q"
+    # Note that to use "--gpu *", you have to setup "complex_value" for the system scheduler.
+
+    export train_cmd="queue.pl"
+    export cuda_cmd="queue.pl"
+    export decode_cmd="queue.pl"
+
+
+# "qsub" (Torque/PBS.)
+elif [ "${cmd_backend}" = pbs ]; then
+    # The default setting is written in conf/pbs.conf.
+
+    export train_cmd="pbs.pl"
+    export cuda_cmd="pbs.pl"
+    export decode_cmd="pbs.pl"
+
+
+# "sbatch" (Slurm)
+elif [ "${cmd_backend}" = slurm ]; then
+    # The default setting is written in conf/slurm.conf.
+    # You must change "-p cpu" and "-p gpu" for the "partition" for your environment.
+    # To know the "partion" names, type "sinfo".
+    # You can use "--gpu * " by default for slurm and it is interpreted as "--gres gpu:*"
+    # The devices are allocated exclusively using "${CUDA_VISIBLE_DEVICES}".
+
+    export train_cmd="slurm.pl"
+    export cuda_cmd="slurm.pl"
+    export decode_cmd="slurm.pl"
+
+elif [ "${cmd_backend}" = ssh ]; then
+    # You have to create ".queue/machines" to specify the host to execute jobs.
+    # e.g. .queue/machines
+    #   host1
+    #   host2
+    #   host3
+    # Assuming you can login them without any password, i.e. You have to set ssh keys.
+
+    export train_cmd="ssh.pl"
+    export cuda_cmd="ssh.pl"
+    export decode_cmd="ssh.pl"
+
+# This is an example of specifying several unique options in the JHU CLSP cluster setup.
+# Users can modify/add their own command options according to their cluster environments.
+elif [ "${cmd_backend}" = jhu ]; then
+
+    export train_cmd="queue.pl --mem 2G"
+    export cuda_cmd="queue-freegpu.pl --mem 2G --gpu 1 --config conf/queue.conf"
+    export decode_cmd="queue.pl --mem 4G"
+
+else
+    echo "$0: Error: Unknown cmd_backend=${cmd_backend}" 1>&2
+    return 1
+fi

egs2/librimix/diar1/conf/decode_diar.yaml

@@ -0,0 +1 @@
+tuning/decode_diar.yaml

egs2/librimix/diar1/conf/decode_diar_eda.yaml

@@ -0,0 +1 @@
+tuning/decode_diar_eda.yaml

egs2/librimix/diar1/conf/pbs.conf

@@ -0,0 +1,11 @@
+# Default configuration
+command qsub -V -v PATH -S /bin/bash
+option name=* -N $0
+option mem=* -l mem=$0
+option mem=0          # Do not add anything to qsub_opts
+option num_threads=* -l ncpus=$0
+option num_threads=1  # Do not add anything to qsub_opts
+option num_nodes=* -l nodes=$0:ppn=1
+default gpu=0
+option gpu=0
+option gpu=* -l ngpus=$0

egs2/librimix/diar1/conf/queue.conf

@@ -0,0 +1,12 @@
+# Default configuration
+command qsub -v PATH -cwd -S /bin/bash -j y -l arch=*64*
+option name=* -N $0
+option mem=* -l mem_free=$0,ram_free=$0
+option mem=0          # Do not add anything to qsub_opts
+option num_threads=* -pe smp $0
+option num_threads=1  # Do not add anything to qsub_opts
+option max_jobs_run=* -tc $0
+option num_nodes=* -pe mpi $0  # You must set this PE as allocation_rule=1
+default gpu=0
+option gpu=0
+option gpu=* -l gpu=$0 -q g.q

egs2/librimix/diar1/conf/slurm.conf

@@ -0,0 +1,14 @@
+# Default configuration
+command sbatch --export=PATH
+option name=* --job-name $0
+option time=* --time $0
+option mem=* --mem-per-cpu $0
+option mem=0
+option num_threads=* --cpus-per-task $0
+option num_threads=1 --cpus-per-task 1
+option num_nodes=* --nodes $0
+default gpu=0
+option gpu=0 -p cpu
+option gpu=* -p gpu --gres=gpu:$0 -c $0  # Recommend allocating more CPU than, or equal to the number of GPU
+# note: the --max-jobs-run option is supported as a special case
+# by slurm.pl and you don't have to handle it in the config file.

egs2/librimix/diar1/conf/train_diar.yaml

@@ -0,0 +1 @@
+tuning/train_diar_2.yaml

egs2/librimix/diar1/conf/train_diar_eda.yaml

@@ -0,0 +1 @@
+tuning/train_diar_eda_5.yaml

egs2/librimix/diar1/conf/train_diar_eda_adapt.yaml

@@ -0,0 +1 @@
+tuning/train_diar_eda_adapt.yaml

egs2/librimix/diar1/conf/tuning/decode_diar.yaml

@@ -0,0 +1,2 @@
+fs: 8000
+num_spk: 2 # The number of speakers will be estimated if "num_spk" is "None"

egs2/librimix/diar1/conf/tuning/decode_diar_eda.yaml

@@ -0,0 +1,2 @@
+fs: 8000
+#num_spk: 2 # The number of speakers will be estimated if "num_spk" is "None"

egs2/librimix/diar1/conf/tuning/train_diar_2.yaml

@@ -0,0 +1,58 @@
+#    This config file is for SA-EEND.
+#    For the details about SA-EEND, refer to the following paper:
+#    SA-EEND: https://arxiv.org/pdf/1909.06247.pdf
+
+# network architecture
+# encoder related
+encoder: transformer
+encoder_conf:
+    input_layer: "linear"
+    num_blocks: 4
+    linear_units: 512
+    dropout_rate: 0.1
+    output_size: 256  # dimension of attention
+    attention_heads: 4
+    attention_dropout_rate: 0.1
+
+# decoder related
+decoder: linear
+
+# minibatch related
+batch_type: folded
+batch_size: 64
+
+# optimization related
+optim: adam
+accum_grad: 2
+grad_clip: 5
+max_epoch: 100
+optim_conf:
+    lr: 0.0002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 100000
+
+# others:
+best_model_criterion:
+-  - valid
+   - acc
+   - max
+keep_nbest_models: 10
+
+patience: none
+# The initialization method for model parameters
+init: xavier_uniform
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: false
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2

egs2/librimix/diar1/conf/tuning/train_diar_eda_5.yaml

@@ -0,0 +1,61 @@
+#    This config file is for EEND-EDA pre-training stage (training on fixed number of speakers).
+#    For the details about EEND-EDA, refer to the following papers:
+#    EEND-EDA: https://arxiv.org/pdf/2005.09921.pdf, https://arxiv.org/pdf/2106.10654.pdf
+
+# network architecture
+# encoder related
+encoder: transformer
+encoder_conf:
+    input_layer: "linear"
+    num_blocks: 4
+    linear_units: 512
+    dropout_rate: 0.1
+    output_size: 256  # dimension of attention
+    attention_heads: 4
+    attention_dropout_rate: 0.1
+
+# attractor related
+attractor: rnn
+attractor_conf:
+    unit: 256 # same as encoder output size
+    layer: 1
+    dropout: 0.1
+    attractor_grad: True
+
+# optimization related
+optim: adam
+grad_clip: 5
+max_epoch: 250
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 30000
+
+# others:
+best_model_criterion:
+-  - valid
+   - acc
+   - max
+keep_nbest_models: 10
+
+batch_type: numel
+batch_bins: 15000000
+accum_grad: 6
+patience: none
+# The initialization method for model parameters
+init: xavier_uniform
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: false
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2