wav2vec2 demo update: support different optimizer and lr_schedular, align mdoel, update input type, test=asr

demos/speech_ssl/run.sh

@@ -1,12 +1,10 @@
 #!/bin/bash
 
 # audio download
-wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav
+wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav
 
 # to recognize text 
 paddlespeech ssl --task asr --lang en --input ./en.wav
 
 # to get acoustic representation
 paddlespeech ssl --task vector --lang en --input ./en.wav
-
-README_cn

examples/librispeech/asr3/conf/preprocess.yaml

@@ -1,4 +1,3 @@
 process:
     # use raw audio
   - type: wav_process
-    dither: 0.0

examples/librispeech/asr3/conf/wav2vec2ASR.yaml

@@ -4,16 +4,21 @@
 freeze_wav2vec2: True
 normalize_wav: True
 output_norm: True
-dnn_blocks: 2
+init_type: 'kaiming_uniform' # !Warning: need to convergence
-dnn_neurons: 1024
+enc:
-blank_id: 0
+  input_shape: 1024
-ctc_dropout_rate: 0.0
+  dnn_blocks: 2
+  dnn_neurons: 1024
+  activation: True
+ctc:
+  enc_n_units: 1024
+  blank_id: 0
+  dropout_rate: 0.0
 wav2vec2_params_path: "exp/wav2vec2/wav2vec2-large-960h-lv60-self.pdparams"
 
 ############################################
 #               Wav2Vec2.0                 #
 ############################################
-vocab_size: 32
 hidden_size: 1024
 num_hidden_layers: 24
 num_attention_heads: 16
@@ -54,9 +59,6 @@ diversity_loss_weight: 0.1
 ctc_loss_reduction: "sum"
 ctc_zero_infinity: False
 use_weighted_layer_sum: False
-pad_token_id: 0
-bos_token_id: 1
-eos_token_id: 2
 add_adapter: False
 adapter_kernel_size: 3
 adapter_stride: 2
@@ -78,7 +80,7 @@ unit_type: 'char'
 mean_std_filepath: ""
 preprocess_config: conf/preprocess.yaml
 sortagrad: -1 # Feed samples from shortest to longest ; -1: enabled for all epochs 0: disabled other: enabled for 'other' epochs 
-batch_size: 10  # Different batch_size may cause large differences in results
+batch_size: 6  # Different batch_size may cause large differences in results
 maxlen_in: 51200000000  # if input length  > maxlen-in batchsize is automatically reduced
 maxlen_out: 1500000  # if output length > maxlen-out batchsize is automatically reduced
 minibatches: 0 # for debug
@@ -112,11 +114,20 @@ model_optim_conf:
   lr: 0.9
   epsilon: 1.0e-6
   rho: 0.95
-scheduler: constantlr    
+model_scheduler: constantlr    
-scheduler_conf:
+model_scheduler_conf:
+  warmup_steps: 25000
+  lr_decay: 1.0
+wav2vec2_optim: adadelta
+wav2vec2_optim_conf:
+  lr: 0.9
+  epsilon: 1.0e-6
+  rho: 0.95
+wav2vec2_scheduler: constantlr    
+wav2vec2_scheduler_conf:
   warmup_steps: 25000
   lr_decay: 1.0
 log_interval: 1
 checkpoint:
   kbest_n: 50
-  latest_n: 5
+  latest_n: 5

examples/librispeech/asr3/local/train.sh

@@ -10,7 +10,8 @@ echo "using $ngpu gpus..."
 
 config_path=$1
 ckpt_name=$2
-ips=$3
+resume=$3
+ips=$4
 
 if [ ! $ips ];then
   ips_config=
@@ -21,7 +22,7 @@ fi
 mkdir -p exp
 
 # seed may break model convergence
-seed=1998
+seed=1988
 if [ ${seed} != 0 ]; then
     export FLAGS_cudnn_deterministic=True
 fi
@@ -34,13 +35,15 @@ python3 -u ${BIN_DIR}/train.py \
 --ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
---seed ${seed} 
+--seed ${seed} \
+--resume ${resume}
 else
 python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${ips_config} ${BIN_DIR}/train.py \
 --ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
---seed ${seed}
+--seed ${seed} \
+--resume ${resume}
 fi
 
 if [ ${seed} != 0 ]; then

examples/librispeech/asr3/run.sh

@@ -11,7 +11,7 @@ conf_path=conf/wav2vec2ASR.yaml
 ips=            #xx.xx.xx.xx,xx.xx.xx.xx
 decode_conf_path=conf/tuning/decode.yaml
 avg_num=1
-dict_path=data/lang_char/vocab.txt
+resume=         # xx e.g. 30
 
 . ${MAIN_ROOT}/utils/parse_options.sh || exit 1;
 
@@ -28,7 +28,7 @@ fi
 
 if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
     # train model, all `ckpt` under `exp` dir
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/train.sh ${conf_path} ${ckpt} ${ips} 
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/train.sh ${conf_path} ${ckpt} ${resume} ${ips} 
 fi
 
 if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
@@ -38,10 +38,10 @@ fi
 
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
     # greedy search decoder
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/test.sh ${conf_path} ${decode_conf_path} exp/${ckpt}/checkpoints/${avg_ckpt} || exit -1
+    CUDA_VISIBLE_DEVICES=0 ./local/test.sh ${conf_path} ${decode_conf_path} exp/${ckpt}/checkpoints/${avg_ckpt} || exit -1
 fi
 
 if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
     # test a single .wav file
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/test_wav.sh ${conf_path} ${decode_conf_path} exp/${ckpt}/checkpoints/${avg_ckpt} ${audio_file} || exit -1
+    CUDA_VISIBLE_DEVICES=0 ./local/test_wav.sh ${conf_path} ${decode_conf_path} exp/${ckpt}/checkpoints/${avg_ckpt} ${audio_file} || exit -1
 fi

paddlespeech/audio/transform/spectrogram.py

@@ -383,7 +383,7 @@ class LogMelSpectrogramKaldi():
 
 
 class WavProcess():
-    def __init__(self, dither=0.0):
+    def __init__(self):
         """
         Args:
             dither (float): Dithering constant
@@ -391,9 +391,7 @@ class WavProcess():
         Returns:
         """
 
-        self.dither = dither
+    def __call__(self, x):
-
-    def __call__(self, x, train):
         """
         Args:
             x (np.ndarray): shape (Ti,)
@@ -405,10 +403,10 @@ class WavProcess():
         Returns:
             np.ndarray: (T, D)
         """
-        dither = self.dither if train else 0.0
         if x.ndim != 1:
             raise ValueError("Not support x: [Time, Channel]")
-        waveform = np.expand_dims(x, -1)
+        waveform = x.astype("float32") / 32768.0
+        waveform = np.expand_dims(waveform, -1)
         return waveform
 
 

paddlespeech/s2t/exps/wav2vec2/bin/train.py

@@ -34,9 +34,10 @@ def main(config, args):
 
 if __name__ == "__main__":
     parser = default_argument_parser()
+    parser.add_argument(
+        '--resume', type=str, default="", nargs="?", help='resume ckpt path.')
     args = parser.parse_args()
     print_arguments(args, globals())
-
     # https://yaml.org/type/float.html
     config = CfgNode(new_allowed=True)
     if args.config:

paddlespeech/s2t/exps/wav2vec2/model.py

@@ -15,6 +15,7 @@
 import json
 import math
 import os
+import re
 import time
 from collections import defaultdict
 from collections import OrderedDict
@@ -62,6 +63,19 @@ class Wav2Vec2ASRTrainer(Trainer):
             self.avg_train_loss -= self.avg_train_loss / (batch_index + 1)
             self.avg_train_loss += loss / (batch_index + 1)
 
+    def before_train(self):
+        from_scratch = self.resume_or_scratch()
+        if from_scratch:
+            # scratch: save init model, i.e. 0 epoch
+            self.save(tag='init', infos=None)
+        else:
+            # resume: train next_epoch and next_iteration
+            self.epoch += 1
+            logger.info(
+                f"Resume train: epoch {self.epoch }, step {self.iteration}!")
+
+        self.maybe_batch_sampler_step()
+
     def train_batch(self, batch_index, batch, msg):
         train_conf = self.config
         start = time.time()
@@ -69,14 +83,14 @@ class Wav2Vec2ASRTrainer(Trainer):
         # forward
         utt, wav, wavs_lens, target, target_lens = batch
         wavs_lens_rate = wavs_lens / wav.shape[1]
-        target_lens_rate = target_lens / target.shape[1]
+
         wav = wav[:, :, 0]
-        if hasattr(train_conf, 'speech_augment'):
+        if hasattr(train_conf, 'audio_augment'):
             wav = self.speech_augmentation(wav, wavs_lens_rate)
-        loss = self.model(wav, wavs_lens_rate, target, target_lens_rate)
+
+        loss = self.model(wav, wavs_lens_rate, target, target_lens)
         # loss div by `batch_size * accum_grad`
         loss /= train_conf.accum_grad
-
         # update self.avg_train_loss
         self.update_average(batch_index, float(loss))
 
@@ -98,11 +112,17 @@ class Wav2Vec2ASRTrainer(Trainer):
 
         # optimizer step old
         if (batch_index + 1) % train_conf.accum_grad == 0:
-            self.optimizer.step()
+            self.model_optimizer.step()
-            self.optimizer.clear_grad()
+            self.model_optimizer.clear_grad()
-            self.lr_scheduler.step()
+            if not train_conf.freeze_wav2vec2:
+                self.wav2vec2_optimizer.step()
+                self.wav2vec2_optimizer.clear_grad()
+            if self.config.model_scheduler != 'newbobscheduler':
+                self.model_lr_scheduler.step()
+            if self.config.wav2vec2_scheduler != 'newbobscheduler':
+                if not train_conf.freeze_wav2vec2:
+                    self.wav2vec2_lr_scheduler.step()
             self.iteration += 1
-
         losses_np = {'loss': self.avg_train_loss * train_conf.accum_grad}
         iteration_time = time.time() - start
         for k, v in losses_np.items():
@@ -114,7 +134,10 @@ class Wav2Vec2ASRTrainer(Trainer):
         if (batch_index + 1) % train_conf.accum_grad == 0:
             if dist.get_rank() == 0 and self.visualizer:
                 losses_np_v = losses_np.copy()
-                losses_np_v.update({"lr": self.lr_scheduler()})
+                losses_np_v.update({
+                    "model_lr": self.model_lr_scheduler(),
+                    "wav2vec2_lr": self.wav2vec2_lr_scheduler()
+                })
                 for key, val in losses_np_v.items():
                     self.visualizer.add_scalar(
                         tag='train/' + key, value=val, step=self.iteration - 1)
@@ -131,11 +154,10 @@ class Wav2Vec2ASRTrainer(Trainer):
         for i, batch in enumerate(self.valid_loader):
             utt, wav, wavs_lens, target, target_lens = batch
             wavs_lens_rate = wavs_lens / wav.shape[1]
-            target_lens_rate = target_lens / target.shape[1]
             wav = wav[:, :, 0]
-            loss = self.model(wav, wavs_lens_rate, target, target_lens_rate)
+            loss = self.model(wav, wavs_lens_rate, target, target_lens)
 
-            if paddle.isfinite(loss):
+            if math.isfinite(float(loss)):
                 num_utts = batch[1].shape[0]
                 num_seen_utts += num_utts
                 total_loss += float(loss) * num_utts
@@ -160,6 +182,106 @@ class Wav2Vec2ASRTrainer(Trainer):
             dist.get_rank(), total_loss / num_seen_utts))
         return total_loss, num_seen_utts
 
+    @mp_tools.rank_zero_only
+    def save(self, tag=None, infos: dict=None):
+        """Save checkpoint (model parameters and optimizer states).
+
+        Args:
+            tag (int or str, optional): None for step, else using tag, e.g epoch. Defaults to None.
+            infos (dict, optional): meta data to save. Defaults to None.
+        """
+
+        infos = infos if infos else dict()
+        infos.update({
+            "epoch": self.epoch,
+            "model_lr": self.model_optimizer.get_lr(),
+            "wav2vec2_lr": self.wav2vec2_optimizer.get_lr()
+        })
+
+        checkpoint_path = os.path.join(
+            self.checkpoint_dir,
+            "{}".format(self.iteration if tag is None else tag))
+
+        model_dict = self.model.state_dict()
+        params_path = checkpoint_path + ".pdparams"
+        paddle.save(model_dict, params_path)
+        logger.info("Saved model to {}".format(params_path))
+
+        model_opt_dict = self.model_optimizer.state_dict()
+        wav2vec2_opt_dict = self.wav2vec2_optimizer.state_dict()
+
+        opt_dict = {'model': model_opt_dict, 'wav2vec2': wav2vec2_opt_dict}
+
+        optimizer_path = checkpoint_path + ".pdopt"
+        paddle.save(opt_dict, optimizer_path)
+        logger.info("Saved optimzier state to {}".format(optimizer_path))
+
+        scheduler_dict = {}
+
+        if self.config.model_scheduler == 'newbobscheduler':
+            scheduler_dict['model'] = self.model_lr_scheduler.save()
+        if self.config.wav2vec2_scheduler == 'newbobscheduler':
+            scheduler_dict['wav2vec2'] = self.wav2vec2_lr_scheduler.save()
+        if scheduler_dict:
+            scheduler_path = checkpoint_path + ".pdlrs"
+            paddle.save(scheduler_dict, scheduler_path)
+            logger.info("Saved scheduler state to {}".format(scheduler_path))
+        info_path = re.sub('.pdparams$', '.json', params_path)
+        infos = {} if infos is None else infos
+        with open(info_path, 'w') as fout:
+            data = json.dumps(infos)
+            fout.write(data)
+
+    def resume_or_scratch(self):
+        """Resume from latest checkpoint at checkpoints in the output
+        directory or load a specified checkpoint.
+
+        If ``args.checkpoint_path`` is not None, load the checkpoint, else
+        resume training.
+        """
+        scratch = None
+        if self.args.resume:
+            # just restore ckpt
+            # lr will resotre from optimizer ckpt
+            resume_json_path = os.path.join(self.checkpoint_dir,
+                                            self.args.resume + '.json')
+            with open(resume_json_path, 'r') as f:
+                resume_json = json.load(f)
+            self.iteration = 0
+            self.epoch = resume_json["epoch"]
+
+            # resotre model from *.pdparams
+            params_path = os.path.join(self.checkpoint_dir,
+                                       "{}".format(self.epoch)) + '.pdparams'
+            model_dict = paddle.load(params_path)
+            self.model.set_state_dict(model_dict)
+
+            # resotre optimizer from *.pdopt
+            optimizer_path = os.path.join(self.checkpoint_dir,
+                                          "{}".format(self.epoch)) + '.pdopt'
+            optimizer_dict = paddle.load(optimizer_path)
+            self.model_optimizer.set_state_dict(optimizer_dict['model'])
+            self.wav2vec2_optimizer.set_state_dict(optimizer_dict['wav2vec2'])
+
+            # resotre lr_scheduler from *.pdlrs
+            scheduler_path = os.path.join(self.checkpoint_dir,
+                                          "{}".format(self.epoch)) + '.pdlrs'
+            if os.path.isfile(os.path.join(scheduler_path)):
+                scheduler_dict = paddle.load(scheduler_path)
+                if self.config.model_scheduler == 'newbobscheduler':
+                    self.model_lr_scheduler.load(scheduler_dict['model'])
+                if self.config.wav2vec2_scheduler == 'newbobscheduler':
+                    self.wav2vec2_lr_scheduler.load(scheduler_dict['wav2vec2'])
+            logger.info(
+                f"Restore ckpt: epoch {self.epoch }, step {self.iteration}!")
+            scratch = False
+        else:
+            self.iteration = 0
+            self.epoch = 0
+            scratch = True
+            logger.info("Init from scratch!")
+        return scratch
+
     def do_train(self):
         """The training process control by step."""
         # !!!IMPORTANT!!!
@@ -170,7 +292,6 @@ class Wav2Vec2ASRTrainer(Trainer):
         # paddle.jit.save(script_model, script_model_path)
 
         self.before_train()
-
         if not self.use_streamdata:
             logger.info(
                 f"Train Total Examples: {len(self.train_loader.dataset)}")
@@ -187,7 +308,9 @@ class Wav2Vec2ASRTrainer(Trainer):
                             report("Rank", dist.get_rank())
                             report("epoch", self.epoch)
                             report('step', self.iteration)
-                            report("lr", self.lr_scheduler())
+                            report("model_lr", self.model_optimizer.get_lr())
+                            report("wav2vec2_lr",
+                                   self.wav2vec2_optimizer.get_lr())
                             self.train_batch(batch_index, batch, msg)
                             self.after_train_batch()
                             report('iter', batch_index + 1)
@@ -225,15 +348,25 @@ class Wav2Vec2ASRTrainer(Trainer):
                     cv_loss = float(cv_loss)
                 else:
                     cv_loss = total_loss / num_seen_utts
-
             logger.info(
                 'Epoch {} Val info val_loss {}'.format(self.epoch, cv_loss))
             if self.visualizer:
                 self.visualizer.add_scalar(
                     tag='eval/cv_loss', value=cv_loss, step=self.epoch)
                 self.visualizer.add_scalar(
-                    tag='eval/lr', value=self.lr_scheduler(), step=self.epoch)
+                    tag='eval/model_lr',
-
+                    value=self.model_lr_scheduler(),
+                    step=self.epoch)
+                self.visualizer.add_scalar(
+                    tag='eval/wav2vec2_lr',
+                    value=self.wav2vec2_lr_scheduler(),
+                    step=self.epoch)
+
+            if self.config.model_scheduler == 'newbobscheduler':
+                self.model_lr_scheduler.step(cv_loss)
+            if self.config.wav2vec2_scheduler == 'newbobscheduler':
+                if not self.config.freeze_wav2vec2:
+                    self.wav2vec2_lr_scheduler.step(cv_loss)
             self.save(tag=self.epoch, infos={'val_loss': cv_loss})
             self.new_epoch()
 
@@ -268,14 +401,11 @@ class Wav2Vec2ASRTrainer(Trainer):
                 model_conf.output_dim = self.test_loader.vocab_size
 
         model = Wav2vec2ASR.from_config(model_conf)
-
+        model_dict = paddle.load(config.wav2vec2_params_path)
-        # load pretrained wav2vec2 model params
+        model.wav2vec2.set_state_dict(model_dict)
-        wav2vec2_dict = paddle.load(config.wav2vec2_params_path)
-        model.wav2vec2.set_state_dict(wav2vec2_dict)
 
         if self.parallel:
             model = paddle.DataParallel(model, find_unused_parameters=True)
-
         logger.info(f"{model}")
         layer_tools.print_params(model, logger.info)
         self.model = model
@@ -290,46 +420,74 @@ class Wav2Vec2ASRTrainer(Trainer):
             return
 
         train_config = config
-        optim_type = train_config.model_optim
+        model_optim_type = train_config.model_optim
-        optim_conf = train_config.model_optim_conf
+        model_optim_conf = train_config.model_optim_conf
-        scheduler_type = train_config.scheduler
+        wav2vec2_optim_type = train_config.model_optim
-        scheduler_conf = train_config.scheduler_conf
+        wav2vec2_optim_conf = train_config.wav2vec2_optim_conf
-
+
-        scheduler_args = {
+        model_scheduler_type = train_config.model_scheduler
-            "learning_rate": optim_conf.lr,
+        model_scheduler_conf = train_config.model_scheduler_conf
-            "verbose": False,
+        wav2vec2_scheduler_type = train_config.wav2vec2_scheduler
-            "warmup_steps": scheduler_conf.warmup_steps,
+        wav2vec2_scheduler_conf = train_config.wav2vec2_scheduler_conf
-            "gamma": scheduler_conf.lr_decay,
+
-            "d_model": model_conf.dnn_neurons,
+        model_scheduler_args = dict(
-        }
+            **{"learning_rate": model_optim_conf.lr,
-        lr_scheduler = LRSchedulerFactory.from_args(scheduler_type,
+               "verbose": False}, **(dict(model_scheduler_conf)))
-                                                    scheduler_args)
+
+        wav2vec2_scheduler_args = dict(
+            **{"learning_rate": wav2vec2_optim_conf.lr,
+               "verbose": False}, **(dict(wav2vec2_scheduler_conf)))
+
+        model_lr_scheduler = LRSchedulerFactory.from_args(model_scheduler_type,
+                                                          model_scheduler_args)
+        wav2vec2_lr_scheduler = LRSchedulerFactory.from_args(
+            wav2vec2_scheduler_type, wav2vec2_scheduler_args)
 
         def optimizer_args(
                 config,
+                optim_type,
+                optim_conf,
                 parameters,
                 lr_scheduler=None, ):
             train_config = config
-            optim_type = train_config.model_optim
+            optim_arg = dict(optim_conf)
-            optim_conf = train_config.model_optim_conf
+            optim_arg.update({
-            scheduler_type = train_config.scheduler
+                "grad_clip":
-            scheduler_conf = train_config.scheduler_conf
+                train_config.global_grad_clip,
-            return {
+                "learning_rate":
-                "grad_clip": train_config.global_grad_clip,
+                lr_scheduler if lr_scheduler else optim_conf.lr,
-                "learning_rate": lr_scheduler
+                "parameters":
-                if lr_scheduler else optim_conf.lr,
+                parameters
-                "epsilon": optim_conf.epsilon,
+            })
-                "rho": optim_conf.rho,
+            return optim_arg
-                "parameters": parameters,
+
-                "beta1": 0.9 if optim_type == 'noam' else None,
+        model_optimizer_args = optimizer_args(config, model_optim_type,
-                "beat2": 0.98 if optim_type == 'noam' else None,
+                                              model_optim_conf, [{
-            }
+                                                  'params':
-
+                                                  model._layers.enc.parameters()
-        optimzer_args = optimizer_args(config, model.parameters(), lr_scheduler)
+                                              }, {
-        optimizer = OptimizerFactory.from_args(optim_type, optimzer_args)
+                                                  'params':
-
+                                                  model._layers.ctc.parameters()
-        self.optimizer = optimizer
+                                              }] if self.parallel else [{
-        self.lr_scheduler = lr_scheduler
+                                                  'params':
+                                                  model.enc.parameters()
+                                              }, {
+                                                  'params':
+                                                  model.ctc.parameters()
+                                              }], model_lr_scheduler)
+        wav2vec2_optimizer_args = optimizer_args(
+            config, wav2vec2_optim_type, wav2vec2_optim_conf,
+            model._layers.wav2vec2.parameters() if self.parallel else
+            model.wav2vec2.parameters(), wav2vec2_lr_scheduler)
+        model_optimizer = OptimizerFactory.from_args(model_optim_type,
+                                                     model_optimizer_args)
+        wav2vec2_optimizer = OptimizerFactory.from_args(wav2vec2_optim_type,
+                                                        wav2vec2_optimizer_args)
+
+        self.model_optimizer = model_optimizer
+        self.wav2vec2_optimizer = wav2vec2_optimizer
+        self.model_lr_scheduler = model_lr_scheduler
+        self.wav2vec2_lr_scheduler = wav2vec2_lr_scheduler
         logger.info("Setup optimizer/lr_scheduler!")
 
 

paddlespeech/s2t/models/wav2vec2/modules/VanillaNN.py

@@ -1,23 +1,12 @@
-# Authors
+"""Vanilla Neural Network for simple tests.
-# * Elena Rastorgueva 2020
+Authors
-# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+* Elena Rastorgueva 2020
-#
+"""
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-# Modified from speechbrain(https://github.com/speechbrain/speechbrain/blob/develop/speechbrain/lobes/models/VanillaNN.py).
 import paddle
 
 from paddlespeech.s2t.models.wav2vec2.modules import containers
 from paddlespeech.s2t.models.wav2vec2.modules import linear
+from paddlespeech.s2t.models.wav2vec2.modules.normalization import BatchNorm1d
 
 
 class VanillaNN(containers.Sequential):
@@ -39,18 +28,34 @@ class VanillaNN(containers.Sequential):
     paddle.shape([10, 120, 512])
     """
 
-    def __init__(
+    def __init__(self,
-            self,
+                 input_shape,
-            input_shape,
+                 dnn_blocks=2,
-            activation=paddle.nn.LeakyReLU,
+                 dnn_neurons=512,
-            dnn_blocks=2,
+                 activation=True,
-            dnn_neurons=512, ):
+                 normalization=False,
-        super().__init__(input_shape=input_shape)
+                 dropout_rate=0.0):
+        super().__init__(input_shape=[None, None, input_shape])
+
+        if not isinstance(dropout_rate, list):
+            dropout_rate = [dropout_rate] * dnn_blocks
+        else:
+            assert len(
+                dropout_rate
+            ) == dnn_blocks, "len(dropout_rate) must equal to dnn_blocks"
 
         for block_index in range(dnn_blocks):
             self.append(
                 linear.Linear,
                 n_neurons=dnn_neurons,
-                bias=True,
+                bias_attr=None,
                 layer_name="linear", )
-            self.append(activation(), layer_name="act")
+            if normalization:
+                self.append(
+                    BatchNorm1d, input_size=dnn_neurons, layer_name='bn')
+            if activation:
+                self.append(paddle.nn.LeakyReLU(), layer_name="act")
+            self.append(
+                paddle.nn.Dropout(),
+                p=dropout_rate[block_index],
+                layer_name='dropout')

paddlespeech/s2t/models/wav2vec2/modules/containers.py

@@ -141,5 +141,4 @@ class Sequential(paddle.nn.LayerDict):
             x = layer(x)
             if isinstance(x, tuple):
                 x = x[0]
-
         return x

paddlespeech/s2t/models/wav2vec2/modules/linear.py

@@ -1,20 +1,8 @@
-# Authors
+"""Library implementing linear transformation.
-#  * Mirco Ravanelli 2020
+Authors
-#  * Davide Borra 2021
+ * Mirco Ravanelli 2020
-# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+ * Davide Borra 2021
-#
+"""
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-# Modified from speechbrain(https://github.com/speechbrain/speechbrain/blob/develop/speechbrain/nnet/linear.py).
 import logging
 
 import paddle
@@ -53,7 +41,7 @@ class Linear(paddle.nn.Layer):
             n_neurons,
             input_shape=None,
             input_size=None,
-            bias=True,
+            bias_attr=None,
             combine_dims=False, ):
         super().__init__()
         self.combine_dims = combine_dims
@@ -67,7 +55,7 @@ class Linear(paddle.nn.Layer):
                 input_size = input_shape[2] * input_shape[3]
 
         # Weights are initialized following paddle approach
-        self.w = align.Linear(input_size, n_neurons, bias_attr=bias)
+        self.w = align.Linear(input_size, n_neurons, bias_attr=bias_attr)
 
     def forward(self, x):
         """Returns the linear transformation of input tensor.

paddlespeech/s2t/models/wav2vec2/modules/modeling_wav2vec2.py

@@ -1120,9 +1120,6 @@ class Wav2Vec2ConfigPure():
         self.output_hidden_states = False
         self.use_return_dict = True
 
-        self.pad_token_id = config.pad_token_id
-        self.bos_token_id = config.bos_token_id
-        self.eos_token_id = config.eos_token_id
         self.hidden_size = config.hidden_size
         self.feat_extract_norm = config.feat_extract_norm
         self.feat_extract_activation = config.feat_extract_activation
@@ -1145,7 +1142,6 @@ class Wav2Vec2ConfigPure():
         self.layerdrop = config.layerdrop
         self.layer_norm_eps = config.layer_norm_eps
         self.initializer_range = config.initializer_range
-        self.vocab_size = config.vocab_size
         self.do_stable_layer_norm = config.do_stable_layer_norm
         self.use_weighted_layer_sum = config.use_weighted_layer_sum
 

paddlespeech/s2t/models/wav2vec2/processing/speech_augmentation.py

@@ -639,6 +639,170 @@ class DropChunk(nn.Layer):
         return dropped_waveform
 
 
+class SpecAugment(paddle.nn.Layer):
+    """An implementation of the SpecAugment algorithm.
+    Reference:
+        https://arxiv.org/abs/1904.08779
+    Arguments
+    ---------
+    time_warp : bool
+        Whether applying time warping.
+    time_warp_window : int
+        Time warp window.
+    time_warp_mode : str
+        Interpolation mode for time warping (default "bicubic").
+    freq_mask : bool
+        Whether applying freq mask.
+    freq_mask_width : int or tuple
+        Freq mask width range.
+    n_freq_mask : int
+        Number of freq mask.
+    time_mask : bool
+        Whether applying time mask.
+    time_mask_width : int or tuple
+        Time mask width range.
+    n_time_mask : int
+        Number of time mask.
+    replace_with_zero : bool
+        If True, replace masked value with 0, else replace masked value with mean of the input tensor.
+    Example
+    -------
+    >>> aug = SpecAugment()
+    >>> a = paddle.rand([8, 120, 80])
+    >>> a = aug(a)
+    >>> print(a.shape)
+    paddle.Size([8, 120, 80])
+    """
+
+    def __init__(
+            self,
+            time_warp=True,
+            time_warp_window=5,
+            time_warp_mode="bicubic",
+            freq_mask=True,
+            freq_mask_width=(0, 20),
+            n_freq_mask=2,
+            time_mask=True,
+            time_mask_width=(0, 100),
+            n_time_mask=2,
+            replace_with_zero=True, ):
+        super().__init__()
+        assert (
+            time_warp or freq_mask or time_mask
+        ), "at least one of time_warp, time_mask, or freq_mask should be applied"
+
+        self.apply_time_warp = time_warp
+        self.time_warp_window = time_warp_window
+        self.time_warp_mode = time_warp_mode
+
+        self.freq_mask = freq_mask
+        if isinstance(freq_mask_width, int):
+            freq_mask_width = (0, freq_mask_width)
+        self.freq_mask_width = freq_mask_width
+        self.n_freq_mask = n_freq_mask
+
+        self.time_mask = time_mask
+        if isinstance(time_mask_width, int):
+            time_mask_width = (0, time_mask_width)
+        self.time_mask_width = time_mask_width
+        self.n_time_mask = n_time_mask
+
+        self.replace_with_zero = replace_with_zero
+
+    def forward(self, x):
+        """Takes in input a tensors and returns an augmented one."""
+        if self.apply_time_warp:
+            x = self.time_warp(x)
+        if self.freq_mask:
+            x = self.mask_along_axis(x, dim=2)
+        if self.time_mask:
+            x = self.mask_along_axis(x, dim=1)
+        return x
+
+    def time_warp(self, x):
+        """Time warping with paddle.nn.functional.interpolate"""
+        original_size = x.shape
+        window = self.time_warp_window
+
+        # 2d interpolation requires 4D or higher dimension tensors
+        # x: (Batch, Time, Freq) -> (Batch, 1, Time, Freq)
+        if x.dim() == 3:
+            x = x.unsqueeze(1)
+
+        time = x.shape[2]
+        if time - window <= window:
+            return x.view(*original_size)
+
+        # compute center and corresponding window
+        c = paddle.randint(window, time - window, (1, ))[0]
+        w = paddle.randint(c - window, c + window, (1, ))[0] + 1
+        # c = 5
+        # w = 10
+        left = paddle.nn.functional.interpolate(
+            x[:, :, :c],
+            (w, x.shape[3]),
+            mode=self.time_warp_mode,
+            align_corners=True, )
+        right = paddle.nn.functional.interpolate(
+            x[:, :, c:],
+            (time - w, x.shape[3]),
+            mode=self.time_warp_mode,
+            align_corners=True, )
+
+        x[:, :, :w] = left
+        x[:, :, w:] = right
+        return x.view(*original_size)
+
+    def mask_along_axis(self, x, dim):
+        """Mask along time or frequency axis.
+        Arguments
+        ---------
+        x : tensor
+            Input tensor.
+        dim : int
+            Corresponding dimension to mask.
+        """
+        original_size = x.shape
+        if x.dim() == 4:
+            x = x.view(-1, x.shape[2], x.shape[3])
+
+        batch, time, fea = x.shape
+
+        if dim == 1:
+            D = time
+            n_mask = self.n_time_mask
+            width_range = self.time_mask_width
+        else:
+            D = fea
+            n_mask = self.n_freq_mask
+            width_range = self.freq_mask_width
+
+        mask_len = paddle.randint(width_range[0], width_range[1],
+                                  (batch, n_mask)).unsqueeze(2)
+
+        mask_pos = paddle.randint(0, max(1, D - mask_len.max()),
+                                  (batch, n_mask)).unsqueeze(2)
+
+        # compute masks
+        arange = paddle.arange(end=D).view(1, 1, -1)
+        mask = (mask_pos <= arange) * (arange < (mask_pos + mask_len))
+        mask = mask.any(axis=1)
+
+        if dim == 1:
+            mask = mask.unsqueeze(2)
+        else:
+            mask = mask.unsqueeze(1)
+
+        if self.replace_with_zero:
+            val = 0.0
+        else:
+            val = x.mean()
+        # same to x.masked_fill_(mask, val)
+        y = paddle.full(x.shape, val, x.dtype)
+        x = paddle.where(mask, y, x)
+        return x.view(*original_size)
+
+
 class TimeDomainSpecAugment(nn.Layer):
     """A time-domain approximation of the SpecAugment algorithm.
     This augmentation module implements three augmentations in

paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py

@@ -23,7 +23,9 @@ import paddle.nn.functional as F
 from paddlespeech.s2t.models.wav2vec2.modules.modeling_wav2vec2 import Wav2Vec2ConfigPure
 from paddlespeech.s2t.models.wav2vec2.modules.modeling_wav2vec2 import Wav2Vec2Model
 from paddlespeech.s2t.models.wav2vec2.modules.VanillaNN import VanillaNN
+from paddlespeech.s2t.models.wav2vec2.processing.speech_augmentation import SpecAugment
 from paddlespeech.s2t.modules.ctc import CTCDecoderBase as CTC
+from paddlespeech.s2t.modules.initializer import DefaultInitializerContext
 from paddlespeech.s2t.utils.ctc_utils import remove_duplicates_and_blank
 from paddlespeech.s2t.utils.utility import log_add
 
@@ -31,42 +33,41 @@ from paddlespeech.s2t.utils.utility import log_add
 class Wav2vec2ASR(nn.Layer):
     def __init__(self, config: dict):
         super().__init__()
+        init_type = config.get("init_type", None)
+        with DefaultInitializerContext(init_type):
+            self.config = config
+            wav2vec2_config = Wav2Vec2ConfigPure(config)
+            wav2vec2 = Wav2Vec2Model(wav2vec2_config)
+            self.normalize_wav = config.normalize_wav
+            self.output_norm = config.output_norm
+            if hasattr(config, 'spec_augment'):
+                self.spec_augment = SpecAugment(**config.spec_augment)
 
-        wav2vec2_config = Wav2Vec2ConfigPure(config)
+            if config.freeze_wav2vec2:
-        wav2vec2 = Wav2Vec2Model(wav2vec2_config)
+                wav2vec2.eval()
-        self.normalize_wav = config.normalize_wav
+                for parm in wav2vec2.parameters():
-        self.output_norm = config.output_norm
+                    parm.trainable = False
-        if config.freeze_wav2vec2:
+            self.wav2vec2 = wav2vec2
-            wav2vec2.eval()
+            self.enc = VanillaNN(**config.enc)
-            for parm in wav2vec2.parameters():
+            self.ctc = CTC(**config.ctc,
-                parm.trainable = False
+                           odim=config.output_dim,
-        self.wav2vec2 = wav2vec2
+                           batch_average=False,
-        self.enc = VanillaNN(
+                           reduction='mean')
-            input_shape=[None, None, wav2vec2_config.hidden_size],
-            activation=nn.LeakyReLU,
-            dnn_blocks=config.dnn_blocks,
-            dnn_neurons=config.dnn_neurons)
-        self.ctc = CTC(odim=config.output_dim,
-                       enc_n_units=config.dnn_neurons,
-                       blank_id=config.blank_id,
-                       dropout_rate=config.ctc_dropout_rate,
-                       reduction='mean')
 
-    def forward(self, wav, wavs_lens_rate, target, target_lens_rate):
+    def forward(self, wav, wavs_lens_rate, target, target_lens):
         if self.normalize_wav:
-            wav = F.layer_norm(wav, wav.shape[1:])
+            wav = F.layer_norm(wav, wav.shape)
         # Extract wav2vec output
         out = self.wav2vec2(wav)[0]
         # We normalize the output if required
         if self.output_norm:
-            out = F.layer_norm(out, out.shape[1:])
+            out = F.layer_norm(out, out.shape)
-        feats = out
+        if self.train and hasattr(self.config, 'spec_augment'):
-
+            feats = self.spec_augment(out)
+        else:
+            feats = out
         x = self.enc(feats)
         x_lens = (wavs_lens_rate * x.shape[1]).round().astype(paddle.int64)
-        target_lens = (target_lens_rate *
-                       target.shape[1]).round().astype(paddle.int64)
-
         ctc_loss = self.ctc(x, x_lens, target, target_lens)
         return ctc_loss
 

paddlespeech/s2t/training/scheduler.py

@@ -17,6 +17,7 @@ from typing import Dict
 from typing import Text
 from typing import Union
 
+import paddle
 from paddle.optimizer.lr import LRScheduler
 from typeguard import check_argument_types
 
@@ -107,6 +108,125 @@ class ConstantLR(LRScheduler):
         return self.base_lr
 
 
+@register_scheduler
+class NewBobScheduler(LRScheduler):
+    """Scheduler with new-bob technique, used for LR annealing.
+
+    The learning rate is annealed based on the validation performance.
+    In particular: if (past_loss-current_loss)/past_loss< impr_threshold:
+    lr=lr * annealing_factor.
+
+    Arguments
+    ---------
+    initial_value : float
+        The initial hyperparameter value.
+    annealing_factor : float
+        It is annealing factor used in new_bob strategy.
+    improvement_threshold : float
+        It is the improvement rate between losses used to perform learning
+        annealing in new_bob strategy.
+    patient : int
+        When the annealing condition is violated patient times,
+        the learning rate is finally reduced.
+
+    Example
+    -------
+    >>> scheduler = NewBobScheduler(initial_value=1.0)
+    >>> scheduler(metric_value=10.0)
+    (1.0, 1.0)
+    >>> scheduler(metric_value=2.0)
+    (1.0, 1.0)
+    >>> scheduler(metric_value=2.5)
+    (1.0, 0.5)
+    """
+
+    def __init__(
+            self,
+            learning_rate,
+            last_epoch=-1,
+            verbose=False,
+            annealing_factor=0.5,
+            improvement_threshold=0.0025,
+            patient=0, ):
+        self.hyperparam_value = learning_rate
+        self.annealing_factor = annealing_factor
+        self.improvement_threshold = improvement_threshold
+        self.patient = patient
+        self.metric_values = []
+        self.current_patient = self.patient
+        super().__init__(learning_rate, last_epoch, verbose)
+
+    def step(self, metric_value=None):
+        """
+
+        ``step`` should be called after ``optimizer.step`` . It will update the learning rate in optimizer according to current ``epoch`` .
+        The new learning rate will take effect on next ``optimizer.step`` .
+
+        Args:
+            epoch (int, None): specify current epoch. Default: None. Auto-increment from last_epoch=-1.
+
+        Returns:
+            None
+        """
+        if metric_value is None:
+            self.last_epoch += 1
+            self.last_lr = self.hyperparam_value
+        else:
+            self.last_epoch += 1
+            self.last_lr = self.get_lr(metric_value)
+
+        if self.verbose:
+            print('Epoch {}: {} set learning rate to {}.'.format(
+                self.last_epoch, self.__class__.__name__, self.last_lr))
+
+    def get_lr(self, metric_value):
+        """Returns the current and new value for the hyperparameter.
+
+        Arguments
+        ---------
+        metric_value : int
+            A number for determining whether to change the hyperparameter value.
+        """
+        new_value = self.hyperparam_value
+        if len(self.metric_values) > 0:
+            prev_metric = self.metric_values[-1]
+            # Update value if improvement too small and patience is 0
+            if prev_metric == 0:  # Prevent division by zero
+                improvement = 0
+            else:
+                improvement = (prev_metric - metric_value) / prev_metric
+            if improvement < self.improvement_threshold:
+                if self.current_patient == 0:
+                    new_value *= self.annealing_factor
+                    self.current_patient = self.patient
+                else:
+                    self.current_patient -= 1
+
+        # Store relevant info
+        self.metric_values.append(metric_value)
+        self.hyperparam_value = new_value
+
+        return new_value
+
+    def save(self):
+        """Saves the current metrics on the specified path."""
+        data = {
+            "current_epoch_index": self.last_epoch,
+            "hyperparam_value": self.hyperparam_value,
+            "metric_values": self.metric_values,
+            "current_patient": self.current_patient
+        }
+        return data
+
+    def load(self, data):
+        """Loads the needed information."""
+        data = paddle.load(data)
+        self.last_epoch = data["current_epoch_index"]
+        self.hyperparam_value = data["hyperparam_value"]
+        self.metric_values = data["metric_values"]
+        self.current_patient = data["current_patient"]
+
+
 def dynamic_import_scheduler(module):
     """Import Scheduler class dynamically.