From 559627c8de1399a021fb6728db2abf01db80eb4b Mon Sep 17 00:00:00 2001
From: "th.zhang" <15600919271@163.com>
Date: Fri, 21 Apr 2023 14:00:27 +0800
Subject: [PATCH] doc related

---
 demos/speech_ssl/README.md                    |   4 +-
 demos/speech_ssl/README_cn.md                 |   4 +-
 docs/source/released_model.md                 |   2 +
 examples/librispeech/asr3/local/train.sh      |   2 +-
 examples/librispeech/asr4/RESULTS.md          |   9 ++
 examples/librispeech/asr4/conf/hubertASR.yaml | 142 ++++++++++++++++++
 examples/librispeech/asr4/local/data.sh       | 110 ++++++++++++++
 examples/librispeech/asr4/local/test.sh       |  83 ++++++++++
 examples/librispeech/asr4/local/test_wav.sh   |  58 +++++++
 examples/librispeech/asr4/local/train.sh      |  58 +++++++
 examples/librispeech/asr4/run.sh              |   6 +-
 examples/librispeech/asr4/utils               |   1 +
 paddlespeech/s2t/models/hubert/__init__.py    |   4 +-
 13 files changed, 473 insertions(+), 10 deletions(-)
 create mode 100644 examples/librispeech/asr4/RESULTS.md
 create mode 100644 examples/librispeech/asr4/conf/hubertASR.yaml
 create mode 100755 examples/librispeech/asr4/local/data.sh
 create mode 100755 examples/librispeech/asr4/local/test.sh
 create mode 100755 examples/librispeech/asr4/local/test_wav.sh
 create mode 100755 examples/librispeech/asr4/local/train.sh
 mode change 100644 => 100755 examples/librispeech/asr4/run.sh
 create mode 120000 examples/librispeech/asr4/utils

diff --git a/demos/speech_ssl/README.md b/demos/speech_ssl/README.md
index b98a7cc61..937cd95a3 100644
--- a/demos/speech_ssl/README.md
+++ b/demos/speech_ssl/README.md
@@ -36,7 +36,7 @@ wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav
   ```
   Arguments:
   - `input`(required): Audio file to recognize.
-  - `model`: Model type of asr task. Default: `wav2vec2ASR_librispeech`.
+  - `model`: Model type of asr task. Default: `wav2vec2`, choices: [wav2vec2, hubert].
   - `task`: Output type. Default: `asr`.
   - `lang`: Model language. Default: `en`.
   - `sample_rate`: Sample rate of the model. Default: `16000`.
@@ -56,7 +56,7 @@ wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav
 
   # to recognize text 
   text = ssl_executor(
-      model='wav2vec2ASR_librispeech',
+      model='wav2vec2',
       task='asr',
       lang='en',
       sample_rate=16000,
diff --git a/demos/speech_ssl/README_cn.md b/demos/speech_ssl/README_cn.md
index 65961ce90..8455d2c77 100644
--- a/demos/speech_ssl/README_cn.md
+++ b/demos/speech_ssl/README_cn.md
@@ -36,7 +36,7 @@ wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav
   ```
   参数：
   - `input`(必须输入)：用于识别的音频文件。
-  - `model`：ASR 任务的模型，默认值：`wav2vec2ASR_librispeech`。
+  - `model`：ASR 任务的模型，默认值：`wav2vec2`, 可选项：[wav2vec2, hubert]。
   - `task`：输出类别，默认值：`asr`。
   - `lang`：模型语言，默认值：`en`。
   - `sample_rate`：音频采样率，默认值：`16000`。
@@ -56,7 +56,7 @@ wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav
 
   # 识别文本
   text = ssl_executor(
-      model='wav2vec2ASR_librispeech',
+      model='wav2vec2,
       task='asr',
       lang='en',
       sample_rate=16000,
diff --git a/docs/source/released_model.md b/docs/source/released_model.md
index 9e9221779..5b0cc70dc 100644
--- a/docs/source/released_model.md
+++ b/docs/source/released_model.md
@@ -26,6 +26,8 @@ Model | Pre-Train Method | Pre-Train Data | Finetune Data | Size | Descriptions
 [Wav2vec2ASR-large-960h-librispeech Model](https://paddlespeech.bj.bcebos.com/s2t/librispeech/asr3/wav2vec2ASR-large-960h-librispeech_ckpt_1.3.1.model.tar.gz) | wav2vec2 | Librispeech and LV-60k Dataset (5.3w h) | Librispeech (960 h) | 718 MB |Encoder: Wav2vec2.0, Decoder: CTC, Decoding method: Greedy search | - | 0.0189 | [Wav2vecASR Librispeech ASR3](../../examples/librispeech/asr3) |
 [Wav2vec2-large-wenetspeech-self Model](https://paddlespeech.bj.bcebos.com/s2t/aishell/asr3/wav2vec2-large-wenetspeech-self_ckpt_1.3.0.model.tar.gz) | wav2vec2 | Wenetspeech Dataset (1w h) | - | 714 MB |Pre-trained Wav2vec2.0 Model | - | - | - | 
 [Wav2vec2ASR-large-aishell1 Model](https://paddlespeech.bj.bcebos.com/s2t/aishell/asr3/wav2vec2ASR-large-aishell1_ckpt_1.4.0.model.tar.gz) | wav2vec2 | Wenetspeech Dataset (1w h) | aishell1 (train set) | 1.18 GB |Encoder: Wav2vec2.0, Decoder: CTC, Decoding method: Greedy search | 0.0510 | - | - |
+[Hubert-large-lv60 Model](https://paddlespeech.bj.bcebos.com/hubert/hubert-large-lv60.pdparams) | hubert | LV-60k Dataset | - | 1.18 GB |Pre-trained hubert Model | - | - | - | 
+[Hubert-large-100h-librispeech Model](https://paddlespeech.bj.bcebos.com/s2t/librispeech/asr4/hubertASR-large-100h-librispeech_ckpt_1.4.0.model.tar.gz) | hubert | LV-60k Dataset | librispeech train-clean-100 | 1.27 GB |Encoder: Hubert, Decoder: Linear + CTC, Decoding method: Greedy search | - | 0.0587 | [HubertASR Librispeech ASR4](../../examples/librispeech/asr4) |
 
 ### Whisper Model
 Demo Link | Training Data | Size | Descriptions | CER | Model 
diff --git a/examples/librispeech/asr3/local/train.sh b/examples/librispeech/asr3/local/train.sh
index 10d254f0b..24776fd17 100755
--- a/examples/librispeech/asr3/local/train.sh
+++ b/examples/librispeech/asr3/local/train.sh
@@ -38,7 +38,7 @@ python3 -u ${BIN_DIR}/train.py \
 --seed ${seed} \
 --resume ${resume}
 else
-python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} --log_dir=exp/log/${ckpt_name} ${ips_config} ${BIN_DIR}/train.py \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${ips_config} ${BIN_DIR}/train.py \
 --ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
diff --git a/examples/librispeech/asr4/RESULTS.md b/examples/librispeech/asr4/RESULTS.md
new file mode 100644
index 000000000..81ce6ee9e
--- /dev/null
+++ b/examples/librispeech/asr4/RESULTS.md
@@ -0,0 +1,9 @@
+# LibriSpeech
+
+## hubertASR
+Fintuning on train-clean-100
+train: Epoch 3, 1*V100-32G, batchsize: 4, accum_grad: 8
+
+| Model | Params | Config | Augmentation| Test set | Decode method | WER |  
+| --- | --- | --- | --- | --- | --- | --- |
+| hubertASR | 326.16M | conf/hubertASR.yaml | spec_aug | test-clean | greedy search | 0.05868 |  
diff --git a/examples/librispeech/asr4/conf/hubertASR.yaml b/examples/librispeech/asr4/conf/hubertASR.yaml
new file mode 100644
index 000000000..e147815a8
--- /dev/null
+++ b/examples/librispeech/asr4/conf/hubertASR.yaml
@@ -0,0 +1,142 @@
+############################################
+#          Network Architecture           #
+############################################
+freeze_hubert: False
+normalize_wav: True
+output_norm: True
+init_type: kaiming_uniform # !Warning: need to convergence
+enc:
+  input_shape: 1024
+  dnn_blocks: 2
+  dnn_neurons: 1024
+  activation: True
+ctc:
+  enc_n_units: 1024
+  blank_id: 0
+  dropout_rate: 0.0
+hubert_params_path: "exp/hubert/pd_hubert_no_fintune.pdparams"
+
+
+task_cfg:
+  label_rate: 50.0
+  sample_rate: 16000
+  normalize: True
+  enable_padding: False
+  max_keep_size: None
+  max_sample_size: 250000
+  min_sample_size: 32000
+  single_target: False
+  random_crop: True
+  pad_audio: False
+
+model_cfg:
+  dropout_input: 0.0
+  final_dropout: 0.0
+  dropout: 0.0
+  attention_dropout: 0.0
+  activation_dropout: 0.1
+  apply_mask: True
+  mask_length: 10
+  mask_prob: 0.5
+  mask_selection: static
+  mask_other: 0.0
+  no_mask_overlap: False
+  mask_channel_length: 64
+  mask_channel_prob: 0.25
+  mask_channel_selection: static
+  mask_channel_other: 0.0
+  no_mask_channel_overlap: False
+  feature_grad_mult: 0.0
+  layerdrop: 0.1
+  normalize: True
+  fp16: True
+  label_rate: 50
+  extractor_mode: layer_norm
+  encoder_layers: 24
+  encoder_embed_dim: 1024
+  encoder_ffn_embed_dim: 4096
+  encoder_attention_heads: 16
+  activation_fn: gelu
+  encoder_layerdrop: 0.1
+  dropout_features: 0.0
+  final_dim: 768
+  untie_final_proj: True
+  layer_norm_first: True
+  conv_feature_layers: "[(512,10,5)] + [(512,3,2)] * 4 + [(512,2,2)] * 2"
+  conv_bias: False
+  logit_temp: 0.1
+  target_glu: False
+  mask_min_space: 1
+  mask_channel_min_space: 1
+  conv_pos: 128
+  conv_pos_groups: 16
+  latent_temp: [2.0, 0.5, 0.999995]
+  skip_masked: False
+  skip_nomask: True
+
+###########################################
+#                   Data                  #
+###########################################
+train_manifest: data/manifest.train-clean-100
+dev_manifest: data/manifest.dev
+test_manifest: data/manifest.test-clean
+
+###########################################
+#              Dataloader                 #
+###########################################
+vocab_filepath: data/lang_char/vocab.txt 
+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: 2  # 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
+batch_count: auto
+batch_bins: 0 
+batch_frames_in: 0
+batch_frames_out: 0
+batch_frames_inout: 0
+num_workers: 0
+subsampling_factor: 1
+num_encs: 1
+dist_sampler: True
+shortest_first: True
+return_lens_rate: True
+  
+############################################
+#             Data Augmentation            #
+############################################
+audio_augment:  # for raw audio 
+  sample_rate: 16000
+  speeds: [95, 100, 105]
+
+###########################################
+#                 Training                #
+###########################################
+n_epoch: 3
+accum_grad: 8
+global_grad_clip: 5.0
+model_optim: adadelta
+model_optim_conf:
+  lr: 1.0
+  epsilon: 1.0e-6
+  rho: 0.95
+model_scheduler: constantlr    
+model_scheduler_conf:
+  warmup_steps: 25000
+  lr_decay: 1.0
+hubert_optim: adadelta
+hubert_optim_conf:
+  lr: 1.0
+  epsilon: 1.0e-6
+  rho: 0.95
+hubert_scheduler: constantlr    
+hubert_scheduler_conf:
+  warmup_steps: 25000
+  lr_decay: 1.0
+log_interval: 1
+checkpoint:
+  kbest_n: 50
+  latest_n: 5
diff --git a/examples/librispeech/asr4/local/data.sh b/examples/librispeech/asr4/local/data.sh
new file mode 100755
index 000000000..7d0613d5a
--- /dev/null
+++ b/examples/librispeech/asr4/local/data.sh
@@ -0,0 +1,110 @@
+#!/bin/bash
+
+stage=-1
+stop_stage=100
+
+unit_type=char
+dict_dir=data/lang_char
+
+source ${MAIN_ROOT}/utils/parse_options.sh
+
+mkdir -p data
+mkdir -p ${dict_dir}
+TARGET_DIR=${MAIN_ROOT}/dataset
+mkdir -p ${TARGET_DIR}
+
+if [ ${stage} -le -1 ] && [ ${stop_stage} -ge -1 ]; then
+    # download data, generate manifests
+    python3 ${TARGET_DIR}/librispeech/librispeech.py \
+    --manifest_prefix="data/manifest" \
+    --target_dir="${TARGET_DIR}/librispeech" \
+    --full_download="True"
+
+    if [ $? -ne 0 ]; then
+        echo "Prepare LibriSpeech failed. Terminated."
+        exit 1
+    fi
+
+    for set in train-clean-100 train-clean-360 train-other-500 dev-clean dev-other test-clean test-other; do
+        mv data/manifest.${set} data/manifest.${set}.raw
+    done
+
+    rm -rf data/manifest.train.raw data/manifest.dev.raw  data/manifest.test.raw
+    for set in train-clean-100 train-clean-360 train-other-500; do
+        cat data/manifest.${set}.raw >> data/manifest.train.raw
+    done
+
+    for set in dev-clean dev-other; do
+        cat data/manifest.${set}.raw >> data/manifest.dev.raw
+    done
+
+    for set in test-clean test-other; do
+        cat data/manifest.${set}.raw >> data/manifest.test.raw
+    done
+fi
+
+if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
+    # compute mean and stddev for normalizer
+    num_workers=$(nproc)
+    python3 ${MAIN_ROOT}/utils/compute_mean_std.py \
+    --manifest_path="data/manifest.train.raw" \
+    --num_samples=2000 \
+    --spectrum_type="fbank" \
+    --feat_dim=161 \
+    --delta_delta=false \
+    --sample_rate=16000 \
+    --stride_ms=10 \
+    --window_ms=25 \
+    --use_dB_normalization=False \
+    --num_workers=${num_workers} \
+    --output_path="data/mean_std.json"
+
+    if [ $? -ne 0 ]; then
+        echo "Compute mean and stddev failed. Terminated."
+        exit 1
+    fi
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    # build vocabulary
+    python3 ${MAIN_ROOT}/utils/build_vocab.py \
+    --unit_type ${unit_type} \
+    --count_threshold=0 \
+    --vocab_path="${dict_dir}/vocab.txt" \
+    --manifest_paths="data/manifest.train.raw"
+
+    if [ $? -ne 0 ]; then
+        echo "Build vocabulary failed. Terminated."
+        exit 1
+    fi
+fi
+
+if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
+    # format manifest with tokenids, vocab size
+    for set in train dev test dev-clean dev-other test-clean test-other; do
+    {
+        python3 ${MAIN_ROOT}/utils/format_data.py \
+        --cmvn_path "data/mean_std.json" \
+        --unit_type ${unit_type} \
+        --vocab_path="${dict_dir}/vocab.txt" \
+        --manifest_path="data/manifest.${set}.raw" \
+        --output_path="data/manifest.${set}"
+
+        if [ $? -ne 0 ]; then
+            echo "Formt mnaifest.${set} failed. Terminated."
+            exit 1
+        fi
+    }&
+    done
+    wait
+fi
+
+echo "LibriSpeech Data preparation done."
+
+if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
+    mkdir -p exp/hubert
+    echo "Pretrained hubert model download"
+    wget -P exp/hubert https://paddlespeech.bj.bcebos.com/hubert/hubert-large-lv60.pdparams
+fi
+
+exit 0
\ No newline at end of file
diff --git a/examples/librispeech/asr4/local/test.sh b/examples/librispeech/asr4/local/test.sh
new file mode 100755
index 000000000..dfbd56ac2
--- /dev/null
+++ b/examples/librispeech/asr4/local/test.sh
@@ -0,0 +1,83 @@
+#!/bin/bash
+
+set -e
+
+ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+echo "using $ngpu gpus..."
+
+expdir=exp
+datadir=data
+
+recog_set="test-clean test-other dev-clean dev-other"
+recog_set="test-clean"
+
+config_path=$1
+decode_config_path=$2
+ckpt_prefix=$3
+
+source ${MAIN_ROOT}/utils/parse_options.sh || exit 1;
+
+# download language model
+#bash local/download_lm_en.sh
+#if [ $? -ne 0 ]; then
+#    exit 1
+#fi
+
+python3 utils/format_rsl.py \
+    --origin_ref data/manifest.test-clean.raw \
+    --trans_ref data/manifest.test-clean.text
+
+
+for type in ctc_greedy_search; do
+    echo "decoding ${type}"
+    batch_size=16
+    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"
+
+exit 0
diff --git a/examples/librispeech/asr4/local/test_wav.sh b/examples/librispeech/asr4/local/test_wav.sh
new file mode 100755
index 000000000..fdf3589f4
--- /dev/null
+++ b/examples/librispeech/asr4/local/test_wav.sh
@@ -0,0 +1,58 @@
+#!/bin/bash
+
+if [ $# != 4 ];then
+    echo "usage: ${0} config_path decode_config_path ckpt_path_prefix audio_file"
+    exit -1
+fi
+
+ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+echo "using $ngpu gpus..."
+
+config_path=$1
+decode_config_path=$2
+ckpt_prefix=$3
+audio_file=$4
+
+mkdir -p data
+wget -nc https://paddlespeech.bj.bcebos.com/datasets/single_wav/en/demo_002_en.wav -P data/
+if [ $? -ne 0 ]; then
+   exit 1
+fi
+
+if [ ! -f ${audio_file} ]; then
+    echo "Plase input the right audio_file path"
+    exit 1
+fi
+
+chunk_mode=false
+if [[ ${config_path} =~ ^.*chunk_.*yaml$ ]];then
+    chunk_mode=true
+fi
+
+# download language model
+#bash local/download_lm_ch.sh
+#if [ $? -ne 0 ]; then
+#    exit 1
+#fi
+
+for type in ctc_greedy_search; do
+    echo "decoding ${type}"
+    batch_size=1
+    output_dir=${ckpt_prefix}
+    mkdir -p ${output_dir}
+    python3 -u ${BIN_DIR}/test_wav.py \
+    --ngpu ${ngpu} \
+    --config ${config_path} \
+    --decode_cfg ${decode_config_path} \
+    --result_file ${output_dir}/${type}.rsl \
+    --checkpoint_path ${ckpt_prefix} \
+    --opts decode.decoding_method ${type} \
+    --opts decode.decode_batch_size ${batch_size} \
+    --audio_file ${audio_file}
+
+    if [ $? -ne 0 ]; then
+        echo "Failed in evaluation!"
+        exit 1
+    fi
+done
+exit 0
diff --git a/examples/librispeech/asr4/local/train.sh b/examples/librispeech/asr4/local/train.sh
new file mode 100755
index 000000000..24776fd17
--- /dev/null
+++ b/examples/librispeech/asr4/local/train.sh
@@ -0,0 +1,58 @@
+#!/bin/bash
+
+if [ $# -lt 2 ] && [ $# -gt 3 ];then
+    echo "usage: CUDA_VISIBLE_DEVICES=0 ${0} config_path ckpt_name ips(optional)"
+    exit -1
+fi
+
+ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+echo "using $ngpu gpus..."
+
+config_path=$1
+ckpt_name=$2
+resume=$3
+ips=$4
+
+if [ ! $ips ];then
+  ips_config=
+else
+  ips_config="--ips="${ips}
+fi
+
+mkdir -p exp
+
+# seed may break model convergence
+seed=1988
+if [ ${seed} != 0 ]; then
+    export FLAGS_cudnn_deterministic=True
+fi
+
+# export FLAGS_cudnn_exhaustive_search=true
+# export FLAGS_conv_workspace_size_limit=4000
+export FLAGS_allocator_strategy=naive_best_fit
+if [ ${ngpu} == 0 ]; then
+python3 -u ${BIN_DIR}/train.py \
+--ngpu ${ngpu} \
+--config ${config_path} \
+--output exp/${ckpt_name} \
+--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} \
+--resume ${resume}
+fi
+
+if [ ${seed} != 0 ]; then
+    unset FLAGS_cudnn_deterministic
+fi
+
+if [ $? -ne 0 ]; then
+    echo "Failed in training!"
+    exit 1
+fi
+
+exit 0
diff --git a/examples/librispeech/asr4/run.sh b/examples/librispeech/asr4/run.sh
old mode 100644
new mode 100755
index c880c9cbf..47e71d60f
--- a/examples/librispeech/asr4/run.sh
+++ b/examples/librispeech/asr4/run.sh
@@ -7,7 +7,7 @@ set -e
 gpus=0
 stage=0
 stop_stage=4
-conf_path=conf/wav2vec2ASR.yaml
+conf_path=conf/hubertASR.yaml
 ips=            #xx.xx.xx.xx,xx.xx.xx.xx
 decode_conf_path=conf/tuning/decode.yaml
 avg_num=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} ${resume} ${ips} 
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/train.sh ${conf_path} ${ckpt} ${resume} ${ips}
 fi
 
 if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
@@ -44,4 +44,4 @@ fi
 if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
     # test a single .wav file
     CUDA_VISIBLE_DEVICES=0 ./local/test_wav.sh ${conf_path} ${decode_conf_path} exp/${ckpt}/checkpoints/${avg_ckpt} ${audio_file} || exit -1
-fi
\ No newline at end of file
+fi
diff --git a/examples/librispeech/asr4/utils b/examples/librispeech/asr4/utils
new file mode 120000
index 000000000..973afe674
--- /dev/null
+++ b/examples/librispeech/asr4/utils
@@ -0,0 +1 @@
+../../../utils
\ No newline at end of file
diff --git a/paddlespeech/s2t/models/hubert/__init__.py b/paddlespeech/s2t/models/hubert/__init__.py
index 4df88bd34..87887a4ce 100644
--- a/paddlespeech/s2t/models/hubert/__init__.py
+++ b/paddlespeech/s2t/models/hubert/__init__.py
@@ -12,6 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 from .hubert_ASR import HubertASR
-from .wav2vec2_ASR import Wav2vec2Base
+from .hubert_ASR import HubertBase
 
-__all__ = ["Wav2vec2ASR", "Wav2vec2Base"]
+__all__ = ["HubertASR", "HubertBase"]