From 0cbf26e0c7899731f09b5c758ee0324d5d6000ce Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=B8=B8=E9=9B=81?= <zhifu.gzf@alibaba-inc.com>
Date: Sat, 23 May 2026 21:21:02 +0800
Subject: [PATCH] fix: increase inference_batch_size from 1 to 32 in aishell
 recipes

The conformer, transformer, branchformer, and e_branchformer recipes
had inference_batch_size=1 which made GPU decoding extremely slow
(processing one utterance at a time). Changed to 32 to match the
paraformer recipe and utilize GPU parallelism during inference.
---
 examples/aishell/branchformer/run.sh   | 2 +-
 examples/aishell/conformer/run.sh      | 4 ++--
 examples/aishell/e_branchformer/run.sh | 2 +-
 examples/aishell/transformer/run.sh    | 4 ++--
 4 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/examples/aishell/branchformer/run.sh b/examples/aishell/branchformer/run.sh
index 5b649546f..722955777 100755
--- a/examples/aishell/branchformer/run.sh
+++ b/examples/aishell/branchformer/run.sh
@@ -17,7 +17,7 @@ nj=32
 inference_device="cuda" #"cpu"
 inference_checkpoint="model.pt.avg10"
 inference_scp="wav.scp"
-inference_batch_size=1
+inference_batch_size=32
 
 # data
 raw_data=../raw_data
diff --git a/examples/aishell/conformer/run.sh b/examples/aishell/conformer/run.sh
index 0c8ab508a..9fd386850 100755
--- a/examples/aishell/conformer/run.sh
+++ b/examples/aishell/conformer/run.sh
@@ -17,7 +17,7 @@ nj=32
 inference_device="cuda" #"cpu", "cuda:0", "cuda:1"
 inference_checkpoint="model.pt.avg10"
 inference_scp="wav.scp"
-inference_batch_size=1
+inference_batch_size=32
 
 # data
 raw_data=../raw_data
@@ -136,7 +136,7 @@ if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
   if [ ${inference_device} == "cuda" ]; then
       nj=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   else
-      inference_batch_size=1
+      inference_batch_size=32
       CUDA_VISIBLE_DEVICES=""
       for JOB in $(seq ${nj}); do
           CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES"-1,"
diff --git a/examples/aishell/e_branchformer/run.sh b/examples/aishell/e_branchformer/run.sh
index 452ec80ba..0eee34a55 100755
--- a/examples/aishell/e_branchformer/run.sh
+++ b/examples/aishell/e_branchformer/run.sh
@@ -17,7 +17,7 @@ nj=32
 inference_device="cuda" #"cpu"
 inference_checkpoint="model.pt.avg10"
 inference_scp="wav.scp"
-inference_batch_size=1
+inference_batch_size=32
 
 # data
 raw_data=../raw_data
diff --git a/examples/aishell/transformer/run.sh b/examples/aishell/transformer/run.sh
index 3fb846519..618bbd150 100755
--- a/examples/aishell/transformer/run.sh
+++ b/examples/aishell/transformer/run.sh
@@ -17,7 +17,7 @@ nj=32
 inference_device="cuda" #"cpu"
 inference_checkpoint="model.pt.avg10"
 inference_scp="wav.scp"
-inference_batch_size=1
+inference_batch_size=32
 
 # data
 raw_data=../raw_data
@@ -137,7 +137,7 @@ if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
   if [ ${inference_device} == "cuda" ]; then
       nj=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
   else
-      inference_batch_size=1
+      inference_batch_size=32
       CUDA_VISIBLE_DEVICES=""
       for JOB in $(seq ${nj}); do
           CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES"-1,"