remove useless code.

examples/aishell/asr3/conf/wav2vec2ASR.yaml

@@ -147,14 +147,12 @@ wav2vec2_optim_conf:
   lr: 0.0001
   weight_decay: 0.0
 
-
-
-model_scheduler: newbobscheduler    
+model_scheduler: newbobscheduler
 model_scheduler_conf:
   improvement_threshold: 0.0025
   annealing_factor: 0.8
   patient: 0
-wav2vec2_scheduler: newbobscheduler    
+wav2vec2_scheduler: newbobscheduler
 wav2vec2_scheduler_conf:
   improvement_threshold: 0.0025
   annealing_factor: 0.9

examples/aishell/asr3/local/test.sh

@@ -54,30 +54,30 @@ for type in ctc_greedy_search; do
     echo "decoding ${type} done."
 done
 
-# for type in ctc_prefix_beam_search; do
-#     echo "decoding ${type}"
-#     batch_size=1
-#     python3 -u ${BIN_DIR}/test.py \
-#         --ngpu ${ngpu} \
-#         --config ${config_path} \
-#         --decode_cfg ${decode_config_path} \
-#         --result_file ${ckpt_prefix}.${type}.rsl \
-#         --checkpoint_path ${ckpt_prefix} \
-#         --opts decode.decoding_method ${type} \
-#         --opts decode.decode_batch_size ${batch_size}
-
-#     if [ $? -ne 0 ]; then
-#         echo "Failed in evaluation!"
-#         exit 1
-#     fi
-#     python3 utils/format_rsl.py \
-#         --origin_hyp ${ckpt_prefix}.${type}.rsl \
-#         --trans_hyp ${ckpt_prefix}.${type}.rsl.text
-
-#     python3 utils/compute-wer.py --char=1 --v=1 \
-#         data/manifest.test-clean.text ${ckpt_prefix}.${type}.rsl.text > ${ckpt_prefix}.${type}.error
-#     echo "decoding ${type} done."
-# done
+for type in ctc_prefix_beam_search; do
+    echo "decoding ${type}"
+    batch_size=1
+    python3 -u ${BIN_DIR}/test.py \
+        --ngpu ${ngpu} \
+        --config ${config_path} \
+        --decode_cfg ${decode_config_path} \
+        --result_file ${ckpt_prefix}.${type}.rsl \
+        --checkpoint_path ${ckpt_prefix} \
+        --opts decode.decoding_method ${type} \
+        --opts decode.decode_batch_size ${batch_size}
+
+    if [ $? -ne 0 ]; then
+        echo "Failed in evaluation!"
+        exit 1
+    fi
+    python3 utils/format_rsl.py \
+        --origin_hyp ${ckpt_prefix}.${type}.rsl \
+        --trans_hyp ${ckpt_prefix}.${type}.rsl.text
+
+    python3 utils/compute-wer.py --char=1 --v=1 \
+        data/manifest.test-clean.text ${ckpt_prefix}.${type}.rsl.text > ${ckpt_prefix}.${type}.error
+    echo "decoding ${type} done."
+done
 
 echo "Finished"
 

examples/aishell/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
-resume=20         # xx e.g. 30
+resume=         # xx e.g. 30
 export FLAGS_cudnn_deterministic=1
 . ${MAIN_ROOT}/utils/parse_options.sh || exit 1;
 
@@ -38,11 +38,11 @@ fi
 
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
     # greedy search decoder
-    CUDA_VISIBLE_DEVICES=4 ./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/s2t/exps/wav2vec2/model.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -17,14 +17,12 @@ import math
 import os
 import re
 import time
-from collections import defaultdict
 from collections import OrderedDict
 from contextlib import nullcontext
 
 import jsonlines
 import numpy as np
 import paddle
-import tqdm
 import transformers
 from hyperpyyaml import load_hyperpyyaml
 from paddle import distributed as dist
@@ -142,7 +140,6 @@ class Wav2Vec2ASRTrainer(Trainer):
         self.avg_train_loss = 0.0
         self.flag = False
         self.use_sb = True
-        self.epoch = 1
 
     def update_average(self, batch_index, loss):
         """Update running average of the loss.

paddlespeech/s2t/models/wav2vec2/io/data_pipeline.py

@@ -110,29 +110,6 @@ class GeneratorDynamicItem(DynamicItem):
     The main benefit is to be able to define the pipeline in a clear function,
     even if parts of the pipeline depend on others for their initialization.
 
-    Example
-    -------
-    >>> lab2ind = {}
-    >>> def text_pipeline(text):
-    ...     text = text.lower().strip()
-    ...     text = "".join(c for c in text if c.isalpha() or c == " ")
-    ...     words = text.split()
-    ...     yield words
-    ...     encoded = [lab2ind[word] for word in words]
-    ...     yield encoded
-    >>> item = GeneratorDynamicItem(
-    ...         func=text_pipeline,
-    ...         takes=["text"],
-    ...         provides=["words", "words_encoded"])
-    >>> # First create the integer-encoding:
-    >>> ind = 1
-    >>> for token in item("Is this it? - This is it."):
-    ...     if token not in lab2ind:
-    ...         lab2ind[token] = ind
-    ...         ind += 1
-    >>> # Now the integers can be encoded!
-    >>> item()
-    [1, 2, 3, 2, 1, 3]
     """
 
     def __init__(self, *args, **kwargs):

paddlespeech/s2t/models/wav2vec2/io/sampler.py

@@ -112,8 +112,6 @@ class ReproducibleWeightedRandomSampler(WeightedRandomSampler):
         to use a value which has a good mix of 0 and 1 bits.
     epoch : int
         The epoch to start at.
-
-
     """
 
     def __init__(

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

@@ -83,7 +83,6 @@ class SpeedPerturb(nn.Layer):
                 "new_freq": self.orig_freq * speed // 100,
             }
             self.resamplers.append(Resample(**config))
-        paddle.seed(2)
 
     def forward(self, waveform):
         """
@@ -465,8 +464,6 @@ class DropFreq(nn.Layer):
             low=self.drop_count_low,
             high=self.drop_count_high + 1,
             shape=(1, ), )
-        ##对齐固定drop_count
-        # drop_count = paddle.to_tensor([2])
 
         # Filter parameters
         filter_length = 101
@@ -481,8 +478,6 @@ class DropFreq(nn.Layer):
             drop_range = self.drop_freq_high - self.drop_freq_low
             drop_frequency = (
                 paddle.rand(drop_count) * drop_range + self.drop_freq_low)
-            #对齐固定drop_frequency
-            # drop_frequency = torch.tensor([0.8102, 0.7742])
             # Subtract each frequency
             for frequency in drop_frequency:
                 notch_kernel = notch_filter(
@@ -752,8 +747,7 @@ class SpecAugment(paddle.nn.Layer):
         # 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]),

paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py

@@ -88,7 +88,7 @@ class Wav2vec2ASR(nn.Layer):
 
         if decoding_method == 'ctc_prefix_beam_search' and batch_size > 1:
             logger.error(
-                f'decoding mode {decoding_method} must be running with batch_size == 1'
+                f"decoding mode {decoding_method} must be running with batch_size == 1"
             )
             logger.error(f"current batch_size is {batch_size}")
 
@@ -311,4 +311,4 @@ class Wav2vec2Base(nn.Layer):
 
     def forward(self, wav):
         out = self.wav2vec2(wav)
-        return
+        return out