update

嘉渊

2023-04-25 0f06fc04c0dcd20c5a3c07976195f9ff3f95fada

update

 egs/aishell/conformer/conf/train_asr_conformer.yaml

@@ -29,16 +29,22 @@
    self_attention_dropout_rate: 0.0
    src_attention_dropout_rate: 0.0

# frontend related
frontend: wav_frontend
frontend_conf:
    fs: 16000
    window: hamming
    n_mels: 80
    frame_length: 25
    frame_shift: 10
    lfr_m: 1
    lfr_n: 1

# hybrid CTC/attention
model_conf:
    ctc_weight: 0.3
    lsm_weight: 0.1     # label smoothing option
    length_normalized_loss: false

# minibatch related
batch_type: length
batch_bins: 25000
num_workers: 16

# optimization related
accum_grad: 1
@@ -76,5 +82,15 @@
    - 40
    num_time_mask: 2

dataset_conf:
    shuffle: True
    shuffle_conf:
        shuffle_size: 2048
        sort_size: 500
    batch_conf:
        batch_type: token
        batch_size: 25000
    num_workers: 8

log_interval: 50
normalize: None

 egs/aishell/conformer/run.sh

@@ -3,7 +3,7 @@
. ./path.sh || exit 1;

# machines configuration
CUDA_VISIBLE_DEVICES="0,1"
CUDA_VISIBLE_DEVICES="2,3"
gpu_num=2
count=1
gpu_inference=true  # Whether to perform gpu decoding, set false for cpu decoding
@@ -13,7 +13,7 @@
infer_cmd=utils/run.pl

# general configuration
feats_dir="../DATA" #feature output dictionary
feats_dir="/nfs/wangjiaming.wjm/Funasr_data/aishell-1-fix-cmvn" #feature output dictionary
exp_dir="."
lang=zh
dumpdir=dump/fbank
@@ -21,7 +21,7 @@
token_type=char
scp=feats.scp
type=kaldi_ark
stage=0
stage=3
stop_stage=4

# feature configuration
@@ -161,7 +161,8 @@
            rank=$i
            local_rank=$i
            gpu_id=$(echo $CUDA_VISIBLE_DEVICES | cut -d',' -f$[$i+1])
            asr_train.py \
            train.py \
                --task_name asr \
                --gpu_id $gpu_id \
                --use_preprocessor true \
                --token_type char \
@@ -177,7 +178,6 @@
                --resume true \
                --output_dir ${exp_dir}/exp/${model_dir} \
                --config $asr_config \
                --input_size $feats_dim \
                --ngpu $gpu_num \
                --num_worker_count $count \
                --multiprocessing_distributed true \