diff --git a/recipes/intel_ndns/README.md b/recipes/intel_ndns/README.md
index 2913a5b..3932444 100644
--- a/recipes/intel_ndns/README.md
+++ b/recipes/intel_ndns/README.md
@@ -61,3 +61,19 @@ There are some 0s files. You may need to filter them out.
 - `russian_speech/M-AILABS_Speech_Dataset/ru_RU_47hrs_48k/male/minaev/oblomov/wavs/oblomov_s002765.wav` is `[1, 0]`.
 - `russian_speech/M-AILABS_Speech_Dataset/ru_RU_47hrs_48k/male/minaev/oblomov/wavs/oblomov_s003811.wav` is `[1, 0]`.
 - Number of non-mono wav files: 0
+
+## Inference Example
+
+Please change the following arguments according to your environment.
+
+```shell
+# For denoising
+cd recipes/intel_ndns/spiking_fullsubnet
+# For dereverberation
+cd recipes/reverb/spiking_fullsubnet
+# For separation
+cd recipes/wsj0-mix/spiking_fullsubnet
+```
+
+accelerate launch --multi_gpu --num_processes=4 --gpu_ids 4,5,6,7 --main_process_port 46529 run.py -C baseline_m.toml -M test --ckpt_path best
+```
\ No newline at end of file
diff --git a/recipes/intel_ndns/spiking_fullsubnet/trainer.py b/recipes/intel_ndns/spiking_fullsubnet/trainer.py
index 7857e95..e62abe2 100644
--- a/recipes/intel_ndns/spiking_fullsubnet/trainer.py
+++ b/recipes/intel_ndns/spiking_fullsubnet/trainer.py
@@ -104,4 +104,4 @@ def test_step(self, batch, batch_idx, dataloader_idx=0):
         return self.validation_step(batch, batch_idx, dataloader_idx)
 
     def test_epoch_end(self, outputs, log_to_tensorboard=True):
-        return self.test_epoch_end(outputs, log_to_tensorboard=False)
+        return self.validation_epoch_end(outputs, log_to_tensorboard=False)