Merge pull request #141 from kan-bayashi/feature/template

egs/README.md

@@ -0,0 +1,110 @@
+# Kaldi-sytle all-in-one recipes
+
+This repository provides [Kaldi](https://github.com/kaldi-asr/kaldi)-style recipes, as the same as [ESPnet](https://github.com/espnet/espnet).  
+Currently, the following recipes are supported.
+
+- [LJSpeech](https://keithito.com/LJ-Speech-Dataset/): English female speaker
+- [JSUT](https://sites.google.com/site/shinnosuketakamichi/publication/jsut): Japanese female speaker
+- [CSMSC](https://www.data-baker.com/open_source.html): Mandarin female speaker
+- [CMU Arctic](http://www.festvox.org/cmu_arctic/): English speakers
+- [JNAS](http://research.nii.ac.jp/src/en/JNAS.html): Japanese multi-speaker
+- [VCTK](https://homepages.inf.ed.ac.uk/jyamagis/page3/page58/page58.html): English multi-speaker
+- [LibriTTS](https://arxiv.org/abs/1904.02882): English multi-speaker
+
+
+## How to run the recipe
+
+```bash
+# Let us move on the recipe directory
+$ cd egs/ljspeech/voc1
+
+# Run the recipe from scratch
+$ ./run.sh
+
+# You can change config via command line
+$ ./run.sh --conf <your_customized_yaml_config>
+
+# You can select the stage to start and stop
+$ ./run.sh --stage 2 --stop_stage 2
+
+# If you want to specify the gpu
+$ CUDA_VISIBLE_DEVICES=1 ./run.sh --stage 2
+
+# If you want to resume training from 10000 steps checkpoint
+$ ./run.sh --stage 2 --resume <path>/<to>/checkpoint-10000steps.pkl
+```
+
+You can check the command line options in `run.sh`.
+
+The integration with job schedulers such as [slurm](https://slurm.schedmd.com/documentation.html) can be done via `cmd.sh` and  `conf/slurm.conf`.  
+If you want to use it, please check [this page](https://kaldi-asr.org/doc/queue.html).
+
+All of the hyperparameters is written in a single yaml format configuration file.  
+Please check [this example](https://github.com/kan-bayashi/ParallelWaveGAN/blob/master/egs/ljspeech/voc1/conf/parallel_wavegan.v1.yaml) in ljspeech recipe.
+
+## How to make the recipe for your own dateset
+
+1. Setup your dataset to be the following structure.
+
+    ```bash
+    # For single-speaker case
+    $ tree /path/to/databse
+    /path/to/database
+    ├── utt_1.wav
+    ├── utt_2.wav
+    │   ...
+    └── utt_N.wav
+    # The directory can be nested, but each filename must be unique
+
+    # For multi-speaker case
+    $ tree /path/to/databse
+    /path/to/database
+    ├── spk_1
+    │   ├── utt1.wav
+    ├── spk_2
+    │   ├── utt1.wav
+    │   ...
+    └── spk_N
+        ├── utt1.wav
+        ...
+    # The directory under each speaker can be nested, but each filename in each speaker directory must be unique
+    ```
+
+2. Copy the template directory.
+
+    ```bash
+    cd egs
+
+    # For single speaker case
+    cp -r template_single_spk <your_dataset_name>
+
+    # For multi speaker case
+    cp -r template_multi_spk <your_dataset_name>
+
+    # Move on your recipe
+    cd egs/<your_dataset_name>/voc1
+    ```
+
+3. Modify the options in `run.sh`.
+
+    > What you need to change at least in `run.sh` is `db_root` option.
+
+4. Modify the hyperpameters in `conf/parallel_wavegan.v1.yaml`.
+
+    > What you need to change at least is `sampling_rate`
+
+5. (Optional) Change command backend in `cmd.sh`.
+
+    > If you are not familiar with kaldi and run in your local env, you do not need to change.
+
+6. Run your recipe.
+
+    ```bash
+    # Run all stages from the first stage
+    ./run.sh
+
+    # Specify CUDA device
+    CUDA_VISIBLE_DEVICES=0 ./run.sh
+    ```
+
+If you want to try the other advanced model, please check the config files in `egs/ljspeech/voc1/conf`.

egs/template_multi_spk/voc1/cmd.sh

@@ -0,0 +1,91 @@
+# ====== 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 documentaion 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 [ "${cmd_backend}" = local ]; then
+
+    # The other usage
+    export train_cmd="utils/run.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="utils/run.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="utils/run.pl"
+
+# Local machine, without any Job scheduling system
+elif [ "${cmd_backend}" = stdout ]; then
+
+    # The other usage
+    export train_cmd="utils/stdout.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="utils/stdout.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="utils/stdout.pl"
+
+# "qsub" (SGE, Torque, PBS, etc.)
+elif [ "${cmd_backend}" = sge ]; then
+    # The default setting is written in conf/queue.conf.
+    # You must change "-q g.q" for the "queue" for your environment.
+    # To know the "queue" names, type "qhost -q"
+    # Note that to use "--gpu *", you have to setup "complex_value" for the system scheduler.
+
+    export train_cmd="utils/queue.pl"
+    export cuda_cmd="utils/queue.pl"
+    export decode_cmd="utils/queue.pl"
+
+# "sbatch" (Slurm)
+elif [ "${cmd_backend}" = slurm ]; then
+    # The default setting is written in conf/slurm.conf.
+    # You must change "-p cpu" and "-p gpu" for the "partion" for your environment.
+    # To know the "partion" names, type "sinfo".
+    # You can use "--gpu * " by defualt for slurm and it is interpreted as "--gres gpu:*"
+    # The devices are allocated exclusively using "${CUDA_VISIBLE_DEVICES}".
+
+    export train_cmd="utils/slurm.pl"
+    export cuda_cmd="utils/slurm.pl"
+    export decode_cmd="utils/slurm.pl"
+
+elif [ "${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="utils/ssh.pl"
+    export cuda_cmd="utils/ssh.pl"
+    export decode_cmd="utils/ssh.pl"
+
+else
+    echo "$0: Error: Unknown cmd_backend=${cmd_backend}" 1>&2
+    return 1
+fi

egs/template_multi_spk/voc1/conf/parallel_wavegan.v1.yaml

@@ -0,0 +1,121 @@
+# This is the hyperparameter configuration file for Parallel WaveGAN.
+# This configuration requires 12 GB GPU memory and takes ~3 days on TITAN V.
+# You need to change the setting depneding on your dataset.
+
+###########################################################
+#                FEATURE EXTRACTION SETTING               #
+###########################################################
+sampling_rate: 22050     # Sampling rate.
+fft_size: 1024           # FFT size.
+hop_size: 256            # Hop size.
+win_length: null         # Window length.
+                         # If set to null, it will be the same as fft_size.
+window: "hann"           # Window function.
+num_mels: 80             # Number of mel basis.
+fmin: 80                 # Minimum freq in mel basis calculation.
+fmax: 7600               # Maximum frequency in mel basis calculation.
+global_gain_scale: 1.0   # Will be multiplied to all of waveform.
+trim_silence: false      # Whether to trim the start and end of silence.
+trim_threshold_in_db: 60 # Need to tune carefully if the recording is not good.
+trim_frame_size: 2048    # Frame size in trimming.
+trim_hop_size: 512       # Hop size in trimming.
+format: "hdf5"           # Feature file format. "npy" or "hdf5" is supported.
+
+###########################################################
+#         GENERATOR NETWORK ARCHITECTURE SETTING          #
+###########################################################
+generator_params:
+    in_channels: 1        # Number of input channels.
+    out_channels: 1       # Number of output channels.
+    kernel_size: 3        # Kernel size of dilated convolution.
+    layers: 30            # Number of residual block layers.
+    stacks: 3             # Number of stacks i.e., dilation cycles.
+    residual_channels: 64 # Number of channels in residual conv.
+    gate_channels: 128    # Number of channels in gated conv.
+    skip_channels: 64     # Number of channels in skip conv.
+    aux_channels: 80      # Number of channels for auxiliary feature conv.
+                          # Must be the same as num_mels.
+    aux_context_window: 2 # Context window size for auxiliary feature.
+                          # If set to 2, previous 2 and future 2 frames will be considered.
+    dropout: 0.0          # Dropout rate. 0.0 means no dropout applied.
+    use_weight_norm: true # Whether to use weight norm.
+                          # If set to true, it will be applied to all of the conv layers.
+    upsample_net: "ConvInUpsampleNetwork" # Upsampling network architecture.
+    upsample_params:                      # Upsampling network parameters.
+        upsample_scales: [4, 4, 4, 4]     # Upsampling scales. Prodcut of these must be the same as hop size.
+
+###########################################################
+#       DISCRIMINATOR NETWORK ARCHITECTURE SETTING        #
+###########################################################
+discriminator_params:
+    in_channels: 1        # Number of input channels.
+    out_channels: 1       # Number of output channels.
+    kernel_size: 3        # Number of output channels.
+    layers: 10            # Number of conv layers.
+    conv_channels: 64     # Number of chnn layers.
+    bias: true            # Whether to use bias parameter in conv.
+    use_weight_norm: true # Whether to use weight norm.
+                          # If set to true, it will be applied to all of the conv layers.
+    nonlinear_activation: "LeakyReLU" # Nonlinear function after each conv.
+    nonlinear_activation_params:      # Nonlinear function parameters
+        negative_slope: 0.2           # Alpha in LeakyReLU.
+
+###########################################################
+#                   STFT LOSS SETTING                     #
+###########################################################
+stft_loss_params:
+    fft_sizes: [1024, 2048, 512]  # List of FFT size for STFT-based loss.
+    hop_sizes: [120, 240, 50]     # List of hop size for STFT-based loss
+    win_lengths: [600, 1200, 240] # List of window length for STFT-based loss.
+    window: "hann_window"         # Window function for STFT-based loss
+
+###########################################################
+#               ADVERSARIAL LOSS SETTING                  #
+###########################################################
+lambda_adv: 4.0  # Loss balancing coefficient.
+
+###########################################################
+#                  DATA LOADER SETTING                    #
+###########################################################
+batch_size: 6              # Batch size.
+batch_max_steps: 25600     # Length of each audio in batch. Make sure dividable by hop_size.
+pin_memory: true           # Whether to pin memory in Pytorch DataLoader.
+num_workers: 2             # Number of workers in Pytorch DataLoader.
+remove_short_samples: true # Whether to remove samples the length of which are less than batch_max_steps.
+allow_cache: true          # Whether to allow cache in dataset. If true, it requires cpu memory.
+
+###########################################################
+#             OPTIMIZER & SCHEDULER SETTING               #
+###########################################################
+generator_optimizer_params:
+    lr: 0.0001             # Generator's learning rate.
+    eps: 1.0e-6            # Generator's epsilon.
+    weight_decay: 0.0      # Generator's weight decay coefficient.
+generator_scheduler_params:
+    step_size: 200000      # Generator's scheduler step size.
+    gamma: 0.5             # Generator's scheduler gamma.
+                           # At each step size, lr will be multiplied by this parameter.
+generator_grad_norm: 10    # Generator's gradient norm.
+discriminator_optimizer_params:
+    lr: 0.00005            # Discriminator's learning rate.
+    eps: 1.0e-6            # Discriminator's epsilon.
+    weight_decay: 0.0      # Discriminator's weight decay coefficient.
+discriminator_scheduler_params:
+    step_size: 200000      # Discriminator's scheduler step size.
+    gamma: 0.5             # Discriminator's scheduler gamma.
+                           # At each step size, lr will be multiplied by this parameter.
+discriminator_grad_norm: 1 # Discriminator's gradient norm.
+
+###########################################################
+#                    INTERVAL SETTING                     #
+###########################################################
+discriminator_train_start_steps: 100000 # Number of steps to start to train discriminator.
+train_max_steps: 400000                 # Number of training steps.
+save_interval_steps: 5000               # Interval steps to save checkpoint.
+eval_interval_steps: 1000               # Interval steps to evaluate the network.
+log_interval_steps: 100                 # Interval steps to record the training log.
+
+###########################################################
+#                     OTHER SETTING                       #
+###########################################################
+num_save_intermediate_results: 4  # Number of results to be saved as intermediate results.

egs/template_multi_spk/voc1/conf/slurm.conf

@@ -0,0 +1,12 @@
+# Default configuration
+command sbatch --export=PATH  --ntasks-per-node=1
+option time=* --time $0
+option mem=* --mem-per-cpu $0
+option mem=0          # Do not add anything to qsub_opts
+option num_threads=* --cpus-per-task $0 --ntasks-per-node=1
+option num_threads=1 --cpus-per-task 1  --ntasks-per-node=1 # Do not add anything to qsub_opts
+default gpu=0
+option gpu=0 -p cpu
+option gpu=* -p gpu --gres=gpu:$0
+# 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.