From 9b4e9cc8a0311e5243d69b73ed073e7ea441982e Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期三, 27 三月 2024 16:05:29 +0800
Subject: [PATCH] train update
---
examples/aishell/transformer/run.sh | 34 ++++++++++++++++++++--------------
1 files changed, 20 insertions(+), 14 deletions(-)
diff --git a/examples/aishell/transformer/run.sh b/examples/aishell/transformer/run.sh
index 276152b..336c558 100755
--- a/examples/aishell/transformer/run.sh
+++ b/examples/aishell/transformer/run.sh
@@ -5,7 +5,7 @@
# general configuration
feats_dir="../DATA" #feature output dictionary
-exp_dir="."
+exp_dir=`pwd`
lang=zh
token_type=char
stage=0
@@ -15,7 +15,7 @@
nj=32
inference_device="cuda" #"cpu"
-inference_checkpoint="model.pt"
+inference_checkpoint="model.pt.avg5"
inference_scp="wav.scp"
inference_batch_size=32
@@ -39,8 +39,9 @@
valid_set=dev
test_sets="dev test"
-config=paraformer_conformer_12e_6d_2048_256.yaml
+config=transformer_12e_6d_2048_256.yaml
model_dir="baseline_$(basename "${config}" .yaml)_${lang}_${token_type}_${tag}"
+
if [ ${stage} -le -1 ] && [ ${stop_stage} -ge -1 ]; then
@@ -77,7 +78,7 @@
--config-name "${config}" \
++train_data_set_list="${feats_dir}/data/${train_set}/audio_datasets.jsonl" \
++cmvn_file="${feats_dir}/data/${train_set}/cmvn.json" \
- ++dataset_conf.num_workers=$nj
+
fi
token_list=${feats_dir}/data/${lang}_token_list/$token_type/tokens.txt
@@ -85,7 +86,7 @@
if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
echo "stage 2: Dictionary Preparation"
mkdir -p ${feats_dir}/data/${lang}_token_list/$token_type/
-
+
echo "make a dictionary"
echo "<blank>" > ${token_list}
echo "<s>" >> ${token_list}
@@ -105,10 +106,12 @@
echo "stage 4: ASR Training"
mkdir -p ${exp_dir}/exp/${model_dir}
- log_file="${exp_dir}/exp/${model_dir}/train.log.txt"
+ current_time=$(date "+%Y-%m-%d_%H-%M")
+ log_file="${exp_dir}/exp/${model_dir}/train.log.txt.${current_time}"
echo "log_file: ${log_file}"
- gpu_num=$(echo CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+ export CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES
+ gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
@@ -128,7 +131,7 @@
if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
echo "stage 5: Inference"
- if ${inference_device} == "cuda"; then
+ if [ ${inference_device} == "cuda" ]; then
nj=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
else
inference_batch_size=1
@@ -140,8 +143,9 @@
for dset in ${test_sets}; do
- inference_dir="${exp_dir}/exp/${model_dir}/${inference_checkpoint}/${dset}"
+ inference_dir="${exp_dir}/exp/${model_dir}/inference-${inference_checkpoint}/${dset}"
_logdir="${inference_dir}/logdir"
+ echo "inference_dir: ${inference_dir}"
mkdir -p "${_logdir}"
data_dir="${feats_dir}/data/${dset}"
@@ -153,7 +157,7 @@
done
utils/split_scp.pl "${key_file}" ${split_scps}
- gpuid_list_array=(${gpuid_list//,/ })
+ gpuid_list_array=(${CUDA_VISIBLE_DEVICES//,/ })
for JOB in $(seq ${nj}); do
{
id=$((JOB-1))
@@ -169,7 +173,9 @@
++input="${_logdir}/keys.${JOB}.scp" \
++output_dir="${inference_dir}/${JOB}" \
++device="${inference_device}" \
- ++batch_size="${inference_batch_size}"
+ ++ncpu=1 \
+ ++disable_log=true \
+ ++batch_size="${inference_batch_size}" &> ${_logdir}/log.${JOB}.txt
}&
done
@@ -185,10 +191,10 @@
done
echo "Computing WER ..."
- cp ${inference_dir}/1best_recog/text ${inference_dir}/1best_recog/text.proc
- cp ${data_dir}/text ${inference_dir}/1best_recog/text.ref
+ python utils/postprocess_text_zh.py ${inference_dir}/1best_recog/text ${inference_dir}/1best_recog/text.proc
+ python utils/postprocess_text_zh.py ${data_dir}/text ${inference_dir}/1best_recog/text.ref
python utils/compute_wer.py ${inference_dir}/1best_recog/text.ref ${inference_dir}/1best_recog/text.proc ${inference_dir}/1best_recog/text.cer
tail -n 3 ${inference_dir}/1best_recog/text.cer
done
-fi
+fi
\ No newline at end of file
--
Gitblit v1.9.1