Model with 5 atom types

[PabloPiaggi]

Hello,

I am trying to train a model with 5 atom types. I am using DeePMD-kit v2.1.1 installed with conda. If I run "dp train" using data with 4 atom types and the input:

{
    "model": {
        "type_map":     ["A","B","C","D"],
        "descriptor": {
            "type": "se_a",
            "sel": [50,50,50,50],
            "rcut_smth": 3.0,
            "rcut": 6.0,
            "neuron": [
                25,
                50,
                100
            ],
            "axis_neuron": 16,
            "seed": 1,
            "_comment": " that's all"
        },

I see no problem. Whereas if I use 5 atom types and the input:

{
    "_comment": "that's all",
    "model": {
        "type_map":     ["A","B","C","D","E"],
        "descriptor": {
            "type": "se_a",
            "sel": [50,50,50,50,50],
            "rcut_smth": 3.0,
            "rcut": 6.0,
            "neuron": [
                25,
                50,
                100
            ],
            "axis_neuron": 16,
            "seed": 1,
            "_comment": " that's all"
        },

I get the error

DEEPMD INFO    initialize model from scratch
DEEPMD INFO    start training at lr 2.00e-03 (== 2.00e-03), decay_step 5000, decay_rate 0.969946, final lr will be 1.00e-08
Traceback (most recent call last):
  File "/home/ppiaggi/.conda/envs/deepmd/bin/dp", line 10, in <module>
    sys.exit(main())
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/entrypoints/main.py", line 473, in main
    train_dp(**dict_args)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/entrypoints/train.py", line 106, in train
    _do_work(jdata, run_opt, is_compress)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/entrypoints/train.py", line 167, in _do_work
    model.train(train_data, valid_data)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/train/trainer.py", line 502, in train
    self.valid_on_the_fly(fp, [train_batch], valid_batches, print_header=True)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/train/trainer.py", line 604, in valid_on_the_fly
    valid_results = self.get_evaluation_results(valid_batches)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/train/trainer.py", line 655, in get_evaluation_results
    results = self.loss.eval(self.sess, feed_dict, natoms)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/loss/ener.py", line 141, in eval
    error, error_e, error_f, error_v, error_ae, error_pf = run_sess(sess, run_data, feed_dict=feed_dict)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/deepmd/utils/sess.py", line 21, in run_sess
    return sess.run(*args, **kwargs)
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/tensorflow/python/client/session.py", line 970, in run
    result = self._run(None, fetches, feed_dict, options_ptr,
  File "/home/ppiaggi/.conda/envs/deepmd/lib/python3.9/site-packages/tensorflow/python/client/session.py", line 1167, in _run
    raise ValueError(
ValueError: Cannot feed value of shape (6,) for Tensor t_natoms:0, which has shape (7,)

I have two data sets - one with 4 atom types (A,B,C,D) and another with 5 atom types (A,B,C,D,E). I am using both for training.

Is this a silly mistake on my side or a bug?

Thank you very much!

All the best,

Pablo

[PabloPiaggi]

I found the mistake. My training set contained all atom types A,B,C,D,E and my validation set contained only A,B,C,D. This is an error on my side but it would be good to have an error message that explains this.

Best,

Pablo

[njzjz]

#1677 will fix the error.

[PabloPiaggi]

Perfect! Thank you.