add mossformer inference recipe

egs_modelscope/speech_separation/speech_separation_mossformer_8k_pytorch/demo.py

@@ -0,0 +1,15 @@
+import numpy
+import soundfile as sf
+from modelscope.pipelines import pipeline
+from modelscope.utils.constant import Tasks
+
+input = 'https://modelscope.cn/api/v1/models/damo/speech_separation_mossformer_8k_pytorch/repo?Revision=master&FilePath=examples/mix_speech1.wav'
+separation = pipeline(
+   Tasks.funasr_speech_separation,
+   model='damo/speech_separation_mossformer_8k_pytorch',
+   output_dir='./',
+   model_revision='v1.0.1')
+result = separation(audio_in=input)
+for i, signal in enumerate(result):
+    save_file = f'output_spk_{i+1}.wav'
+    sf.write(save_file, signal, 8000)