From 2d71d8f679894ab49374b10784547db001bba7be Mon Sep 17 00:00:00 2001
From: 语帆 <yf352572@alibaba-inc.com>
Date: 星期三, 28 二月 2024 17:30:27 +0800
Subject: [PATCH] test
---
examples/industrial_data_pretraining/lcbnet/demo_nj.sh | 80 ++++++++++++++++++++++++++++++++++-----
1 files changed, 69 insertions(+), 11 deletions(-)
diff --git a/examples/industrial_data_pretraining/lcbnet/demo_nj.sh b/examples/industrial_data_pretraining/lcbnet/demo_nj.sh
index 9515f98..51ffad7 100755
--- a/examples/industrial_data_pretraining/lcbnet/demo_nj.sh
+++ b/examples/industrial_data_pretraining/lcbnet/demo_nj.sh
@@ -1,13 +1,71 @@
file_dir="/nfs/yufan.yf/workspace/github/FunASR/examples/industrial_data_pretraining/lcbnet/exp/speech_lcbnet_contextual_asr-en-16k-bpe-vocab5002-pytorch"
+CUDA_VISIBLE_DEVICES="0,1"
+inference_device="cuda"
-#CUDA_VISIBLE_DEVICES="" \
-python -m funasr.bin.inference \
---config-path=${file_dir} \
---config-name="config.yaml" \
-++init_param=${file_dir}/model.pb \
-++tokenizer_conf.token_list=${file_dir}/tokens.txt \
-++input=[${file_dir}/wav.scp,${file_dir}/ocr.txt] \
-+data_type='["kaldi_ark", "text"]' \
-++tokenizer_conf.bpemodel=${file_dir}/bpe.model \
-++output_dir="./outputs/debug" \
-++device="cpu" \
+if [ ${inference_device} == "cuda" ]; then
+ nj=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+else
+ inference_batch_size=1
+ CUDA_VISIBLE_DEVICES=""
+ for JOB in $(seq ${nj}); do
+ CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES"-1,"
+ done
+fi
+
+inference_dir="outputs/test"
+_logdir="${inference_dir}/logdir"
+echo "inference_dir: ${inference_dir}"
+
+mkdir -p "${_logdir}"
+key_file1=${file_dir}/wav.scp
+key_file2=${file_dir}/ocr.txt
+split_scps1=
+split_scps2=
+for JOB in $(seq "${nj}"); do
+ split_scps1+=" ${_logdir}/wav.${JOB}.scp"
+ split_scps2+=" ${_logdir}/ocr.${JOB}.txt"
+done
+utils/split_scp.pl "${key_file1}" ${split_scps1}
+utils/split_scp.pl "${key_file2}" ${split_scps2}
+
+gpuid_list_array=(${CUDA_VISIBLE_DEVICES//,/ })
+for JOB in $(seq ${nj}); do
+ {
+ id=$((JOB-1))
+ gpuid=${gpuid_list_array[$id]}
+
+ export CUDA_VISIBLE_DEVICES=${gpuid}
+
+ python -m funasr.bin.inference \
+ --config-path=${file_dir} \
+ --config-name="config.yaml" \
+ ++init_param=${file_dir}/model.pb \
+ ++tokenizer_conf.token_list=${file_dir}/tokens.txt \
+ ++input=[${_logdir}/wav.${JOB}.scp,${_logdir}/ocr.${JOB}.txt] \
+ +data_type='["kaldi_ark", "text"]' \
+ ++tokenizer_conf.bpemodel=${file_dir}/bpe.model \
+ ++output_dir="${inference_dir}/${JOB}" \
+ ++device="${inference_device}" \
+ ++ncpu=1 \
+ ++disable_log=true &> ${_logdir}/log.${JOB}.txt
+
+ }&
+done
+wait
+
+
+mkdir -p ${inference_dir}/1best_recog
+for f in token score text; do
+ if [ -f "${inference_dir}/${JOB}/1best_recog/${f}" ]; then
+ for JOB in $(seq "${nj}"); do
+ cat "${inference_dir}/${JOB}/1best_recog/${f}"
+ done | sort -k1 >"${inference_dir}/1best_recog/${f}"
+ fi
+done
+
+echo "Computing WER ..."
+echo "Computing WER ..."
+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
\ No newline at end of file
--
Gitblit v1.9.1