ntypes error

[Xi-yuanWang]

I set the typemap in param.json as ["Co","Ni"], and use a dataset of bulk Ni. Then ntypes error happens！

Traceback (most recent call last):
  File "/nfs-share/home/1800011848/miniconda3/bin/dp", line 10, in <module>
    sys.exit(main())
  File "/nfs-share/home/1800011848/miniconda3/lib/python3.8/site-packages/deepmd/main.py", line 66, in main
    train(args)
  File "/nfs-share/home/1800011848/miniconda3/lib/python3.8/site-packages/deepmd/train.py", line 81, in train
    _do_work(jdata, run_opt)
  File "/nfs-share/home/1800011848/miniconda3/lib/python3.8/site-packages/deepmd/train.py", line 134, in _do_work
    model.build (data, stop_batch)
  File "/nfs-share/home/1800011848/miniconda3/lib/python3.8/site-packages/deepmd/Trainer.py", line 209, in build
    assert (self.ntypes == data.get_ntypes()), "ntypes should match that found in data"
AssertionError: ntypes should match that found in data

And the same param works well when using a alloy dataset( all type.raw have two elements)

[njzjz]

Does sel in your parameters contain 2 values?

deepmd-kit/examples/water/train/water_se_a.json

Lines 4 to 7 in 59d780c

	"type_map": ["O", "H"],
	"descriptor" :{
	"type": "se_a",
	"sel": [46, 92],

[Xi-yuanWang]

Yes, and here is my param.json and dataset.
error.zip

[njzjz]

Creating a file error/0/system/type_map.raw can solve the problem:

Co
Ni

I am marking this issue as a bug and would like to discuss whether we need to solve the problem without type_map.raw. Thanks!