From a8725a8dcc96bdfea359d280abd909675573e5f9 Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期三, 22 三月 2023 19:01:04 +0800
Subject: [PATCH] cer tool
---
funasr/runtime/python/utils/infer.py | 4 ++--
funasr/runtime/python/utils/infer.sh | 14 ++++++++------
2 files changed, 10 insertions(+), 8 deletions(-)
diff --git a/funasr/runtime/python/utils/infer.py b/funasr/runtime/python/utils/infer.py
index f44a884..f5b2233 100644
--- a/funasr/runtime/python/utils/infer.py
+++ b/funasr/runtime/python/utils/infer.py
@@ -37,8 +37,8 @@
for i, wav_path_i in enumerate(wav_files):
wav_name, wav_path = wav_path_i.strip().split()
result = model(wav_path)
- text_i = "{} {}\n".format(wav_name, result[0])
- token_i = "{} {}\n".format(wav_name, result[1])
+ text_i = "{} {}\n".format(wav_name, result[0]['preds'][0])
+ token_i = "{} {}\n".format(wav_name, result[0]['preds'][1])
text_f.write(text_i)
text_f.flush()
token_f.write(token_i)
diff --git a/funasr/runtime/python/utils/infer.sh b/funasr/runtime/python/utils/infer.sh
index f5012da..c434f4a 100644
--- a/funasr/runtime/python/utils/infer.sh
+++ b/funasr/runtime/python/utils/infer.sh
@@ -1,4 +1,9 @@
+split_scps_tool=split_scp.pl
+inference_tool=infer.py
+proce_text_tool=proce_text.py
+compute_wer_tool=compute_wer.py
+
nj=32
stage=0
stop_stage=2
@@ -6,10 +11,7 @@
scp="/nfs/haoneng.lhn/funasr_data/aishell-1/data/test/wav.scp"
label_text="/nfs/haoneng.lhn/funasr_data/aishell-1/data/test/text"
export_root="/nfs/zhifu.gzf/export"
-split_scps_tool=split_scp.pl
-inference_tool=infer.py
-proce_text_tool=proce_text.py
-compute_wer_tool=compute_wer.py
+
#:<<!
model_name="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
@@ -44,7 +46,7 @@
for JOB in $(seq ${nj}); do
{
core_id=`expr $JOB - 1`
- taskset -c ${core_id} python ${rtf_tool} --backend ${backend} --model_dir ${model_dir} --wav_file ${output_dir}/wav.$JOB.scp --quantize ${quantize} --output_dir ${output_dir}/${JOB} &> ${output_dir}/log.$JOB.txt
+ taskset -c ${core_id} python ${inference_tool} --backend ${backend} --model_dir ${model_dir} --wav_file ${output_dir}/wav.$JOB.scp --quantize ${quantize} --output_dir ${output_dir}/${JOB} &> ${output_dir}/log.$JOB.txt
}&
done
@@ -54,7 +56,7 @@
for f in token text; do
if [ -f "${output_dir}/1/${f}" ]; then
for JOB in $(seq "${nj}"); do
- cat "${output_dir}/${JOB}/1best_recog/${f}"
+ cat "${output_dir}/${JOB}/${f}"
done | sort -k1 >"${output_dir}/1best_recog/${f}"
fi
done
--
Gitblit v1.9.1