How to use ParametricDMD with BOPDMD?

[zonexo]

Hi,

I have managed to get ParametricDMD working with the vanilla DMD.

However, when I tried to use it with BOPDMD, I got an error:

pdmd_monolithic.fit(training_snapshots, training_params)

TypeError: fit() missing 1 required positional argument: 't'

I understand that BOPDMD requires an additional argument:

optdmd = BOPDMD(svd_rank=svd_rank_bop, num_trials=0)
        optdmd.fit(snapshots, snapshots_time)

So how should I add snapshots_time to pdmd_monolithic.fit to get it working?

Thanks!

[fandreuz]

Hi @zonexo, thanks for raising this problem. I've opened a PR to fix this by allowing passing additional parameters to the constructor of ParametricDMD to tweak the call to the underlying DMD(s) instance(s): #463

You will be able to call the constructor as follows:

p = ParametricDMD(
        dmds,
        POD(rank=5),
        RBF(),
        # Positional and keyword arguments
        dmd_fit_args=(1, 2, None),
        dmd_fit_kwargs={"label": "PIPPO", "n": 2},
    )

Then each fit() call will be called as follows:

sub_dmd.fit(..., 1, 2, None, label="PIPPO", n=2)

Maybe you could give it a try and see if it fixes your problem? You just need to clone https://github.com/PyDMD/PyDMD/tree/param-dmd-git-args and then pip install -e . from the directory. Let me know if you need more info.

[fandreuz]

@zonexo The fix was merged a few hours ago, therefore you can simply git clone the main repository. Note that it may take some time to have the change in PyPi

[zonexo]

So I can just use:
pip install pydmd ?

[fandreuz]

The addition is not yet in PyPi, therefore you need to clone the repository and install PyDMD from the source:

$ https://github.com/PyDMD/PyDMD.git
$ cd PyDMD
$ python -m pip install -e .

[zonexo]

Hi @zonexo, thanks for raising this problem. I've opened a PR to fix this by allowing passing additional parameters to the constructor of ParametricDMD to tweak the call to the underlying DMD(s) instance(s): #463

You will be able to call the constructor as follows:

p = ParametricDMD(
        dmds,
        POD(rank=5),
        RBF(),
        # Positional and keyword arguments
        dmd_fit_args=(1, 2, None),
        dmd_fit_kwargs={"label": "PIPPO", "n": 2},
    )

Then each fit() call will be called as follows:

sub_dmd.fit(..., 1, 2, None, label="PIPPO", n=2)

Maybe you could give it a try and see if it fixes your problem? You just need to clone https://github.com/PyDMD/PyDMD/tree/param-dmd-git-args and then pip install -e . from the directory. Let me know if you need more info.

Sorry, I don't understand fully. In tutoial 14:

optdmd = BOPDMD(svd_rank=1, num_trials=0)
optdmd.fit(snapshots_1d, time)
time = np.arange(0, 6 + 0.4, 0.4)

In tutorial 10,

dmd = DMD(svd_rank=-1)
rom = POD(rank=20)
interpolator = RBF()

pdmd_monolithic = ParametricDMD(dmd, rom, interpolator)

pdmd_monolithic.fit(
    training_snapshots, training_params
)  # same for pdmd_partitioned

So now if I replace:
dmd = DMD(svd_rank=-1)
with
dmd = BOPDMD(svd_rank=-1, num_trials=0)

Now:

p = ParametricDMD(
        dmds,
        POD(rank=5),
        RBF(),
        # Positional and keyword arguments
        dmd_fit_args=(1, 2, None),
        dmd_fit_kwargs={"label": "PIPPO", "n": 2},
    )

How do I insert the time data into the new arguement?

Thanks

[fandreuz]

dmd_fit_args and dmd_fit_kwargs are respectively positional and keywork arguments which will be attached to each call to the internal DMD instance in the parametric DMD. Therefore you need to create an instance of ParametricDMD with the argument dmd_fit_args=(time,).

[zonexo]

I tried git clone https://github.com/PyDMD/PyDMD/tree/param-dmd-git-args but I got the error:

$ git clone https://github.com/PyDMD/PyDMD/tree/param-dmd-git-args
Cloning into 'param-dmd-git-args'...
fatal: repository 'https://github.com/PyDMD/PyDMD/tree/param-dmd-git-args/' not found

Is there a spelling mistake?

I then tried git clone https://github.com/PyDMD/PyDMD and pip install -e .

Is this the same?

However, I then modify part of tutorial 10:

t = np.linspace(0, 4 * np.pi, n_time)
dmd = BOPDMD(svd_rank=0, num_trials=0)
rom = POD(rank=20)
interpolator = RBF()

pdmd_monolithic = ParametricDMD(dmd, rom, interpolator,dmd_fit_args=(t,))

but got the error:

TypeError: __init__() got an unexpected keyword argument 'dmd_fit_args'

Thanks for the help!

[fandreuz]

git clone https://github.com/PyDMD/PyDMD/tree/param-dmd-git-args

You can use the main repository since the PR is already merged.

For the error, I propose you try pip uninstall pydmd and then reinstalling it from the source (pip install -e . from the PyDMD directory). Maybe you're still using an old version.

[zonexo]

I tried to uninstall and git clone everything again and it's working now! Thanks for the help!
Btw, can you explain a bit what:
dmd_fit_kwargs={"label": "PIPPO", "n": 2}
is for?

[fandreuz]

It's just a common way to display how to use a specific syntax, but it does not mean anything, sorry for confusing you!

You can omit dmd_fit_kwargs