FailedPreconditionError: Attempting to use uninitialized value dense_22/bias #677

dpde101 · 2020-03-18T19:14:39Z

dpde101
Mar 18, 2020

hi i see your reply of yesterday to use X = np.array([[3]]) for model.predict but when i do i have error

FailedPreconditionError: Attempting to use uninitialized value dense_22/bias

is complete erreur message

FailedPreconditionError                   Traceback (most recent call last)

/tensorflow-1.15.0/python3.6/tensorflow_core/python/client/session.py in _do_call(self, fn, *args)
   1364     try:
-> 1365       return fn(*args)
   1366     except errors.OpError as e:

9 frames

FailedPreconditionError: Attempting to use uninitialized value dense_22/bias
	 [[{{node dense_22/bias/read}}]]


During handling of the above exception, another exception occurred:

FailedPreconditionError                   Traceback (most recent call last)

/tensorflow-1.15.0/python3.6/tensorflow_core/python/client/session.py in _do_call(self, fn, *args)
   1382                     '\nsession_config.graph_options.rewrite_options.'
   1383                     'disable_meta_optimizer = True')
-> 1384       raise type(e)(node_def, op, message)
   1385 
   1386   def _extend_graph(self):

FailedPreconditionError: Attempting to use uninitialized value dense_22/bias
	 [[node dense_22/bias/read (defined at /tensorflow-1.15.0/python3.6/tensorflow_core/python/framework/ops.py:1748) ]]

Original stack trace for 'dense_22/bias/read':

Answered by lululxvi

Mar 20, 2020

Because you predict before training.

View full answer

dpde101 · 2020-03-18T20:03:18Z

dpde101
Mar 18, 2020
Author

and you can pls give me notation for initial value for exemple dy/dt=f(t,y) and y(0)=1 how to writ initial condition not boundary condition thanks

0 replies

lululxvi · 2020-03-19T15:37:37Z

lululxvi
Mar 19, 2020
Maintainer

For the error, did you initialize/train the network before prediction?

The IVP can be viewed as BVP where the boundary is at left point. check #674

0 replies

dpde101 · 2020-03-19T18:33:20Z

dpde101
Mar 19, 2020
Author

i get error when i initialize/train before and after prediction both. for initial value is there way to declare it without putting the point in the geometry like boundary value ? thank you greetings

0 replies

lululxvi · 2020-03-19T21:06:24Z

lululxvi
Mar 19, 2020
Maintainer

You can only make prediction after the network is initialized. If you don't want to do the training, you can just call train with 0 steps.

DeepXDE supports IC for at least 1D problem, e.g., Burgers https://github.com/lululxvi/deepxde/blob/master/examples/Burgers.py . But DeepXDE currently doesn't support IC for 0D problem, because it is just a special case of left boundary problem.

0 replies

dpde101 · 2020-03-19T22:22:30Z

dpde101
Mar 19, 2020
Author

i think it is because i try to use model.predict(X) in arithmetic operations (+ and -) do you have any idea how to bypass this thanks

…

On Thu, Mar 19, 2020 at 10:06 PM Lu Lu ***@***.***> wrote: You can only make prediction after the network is initialized. If you don't want to do the training, you can just call train with 0 steps. DeepXDE supports IC for at least 1D problem, e.g., Burgers https://github.com/lululxvi/deepxde/blob/master/examples/Burgers.py . But DeepXDE currently doesn't support IC for 0D problem, because it is just a special case of left boundary problem. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://github.com/lululxvi/deepxde/issues/19#issuecomment-601416813>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AO3XKEUBIJA4PT4MDVDZBDTRIKCN7ANCNFSM4LOYV6MA> .

0 replies

lululxvi · 2020-03-20T02:30:31Z

lululxvi
Mar 20, 2020
Maintainer

I am not sure what you mean. Could you show me the code?

0 replies

dpde101 · 2020-03-20T11:22:03Z

dpde101
Mar 20, 2020
Author

    layer_size = [1] + [35] * 3 + [1]
    activation = "tanh"
    initializer = "Glorot uniform"
    geom = dde.geometry.Interval(1, 20)
    net = dde.maps.FNN(layer_size, activation, initializer)

    t0=0
    i=0
    X=np.array([[i+1]])




    while (i<20):
           bc = dde.DirichletBC(geom, lambda x:0*x+t0, boundary)
           data = dde.data.PDE(geom, 1, ode_system, bc, 30, 2, num_test=100)
           model = dde.Model(data, net)
           t0=(t0+model.predict(x))
           model.compile("adam", lr=0.001, metrics=["l2 relative error"])
           losshistory, train_state = model.train(epochs=20000); 
           i=i+1

    dde.saveplot(losshistory, train_state, issave=True, isplot=True)

edit: i still get same error after putting t0=(t0+model.predict(x)) after model.train so i think issues is in model.predict

0 replies

lululxvi · 2020-03-20T18:00:20Z

lululxvi
Mar 20, 2020
Maintainer

Because you predict before training.

0 replies

dpde101 · 2020-03-20T19:14:15Z

dpde101
Mar 20, 2020
Author

absolutely correct, i make mistake early in the code i forget copy paste a line of code thank s

…

On Fri, Mar 20, 2020 at 7:00 PM Lu Lu ***@***.***> wrote: Because you predict before training. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://github.com/lululxvi/deepxde/issues/19#issuecomment-601837417>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AO3XKEWZMFKD2Z3CISRHKK3RIOVMHANCNFSM4LOYV6MA> .

0 replies

lululxvi · 2020-03-24T21:42:54Z

lululxvi
Mar 24, 2020
Maintainer

@dpde101 In fact I find that DeepXDE supports using IC for IVP, see https://github.com/lululxvi/deepxde/blob/master/examples/ode_system.py

0 replies

dpde101 · 2020-03-25T02:22:23Z

dpde101
Mar 25, 2020
Author

thanks for tell me can you give example for syntax for example dy1/dt=f1(t,y), dy2/dt=f2(t,y) and y1(0)=1 and y2(0)=3 what are the expressions for boundary and ic1 and ic2 and data thank you

…

On Tue, Mar 24, 2020 at 10:43 PM Lu Lu ***@***.***> wrote: @dpde101 <https://github.com/dpde101> In fact I find that DeepXDE supports using IC for IVP, see https://github.com/lululxvi/deepxde/blob/master/examples/ode_system.py — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <https://github.com/lululxvi/deepxde/issues/19#issuecomment-603520286>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AO3XKEUSD2FJGULFXGXQDA3RJESO7ANCNFSM4LOYV6MA> .

0 replies

lululxvi · 2020-03-25T16:20:18Z

lululxvi
Mar 25, 2020
Maintainer

I am not sure what you mean. The ODE system example should have what you need.

0 replies

dpde101 · 2020-03-25T19:26:22Z

dpde101
Mar 25, 2020
Author

not really because they use a function (np.sin and np.cos) to define ic1 and ic2 but i want to use lambda x this is exctly like #673 where you gave a well detailed solution #673 (comment), what i want is the same thing but for initial conditions instead of boundary conditions thankyou

…

On Wed, Mar 25, 2020 at 5:20 PM Lu Lu ***@***.***> wrote: I am not sure what you mean. The ODE system example should have what you need. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <https://github.com/lululxvi/deepxde/issues/19#issuecomment-603937587>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AO3XKESKG3IULZ4YGGYNG6TRJIVNFANCNFSM4LOYV6MA> .

0 replies

lululxvi · 2020-03-25T20:37:36Z

lululxvi
Mar 25, 2020
Maintainer

np.sin is the same as lambda expression, right? This is all about Python grammar.
IC and BC have almost the same usage.
You can find more examples, e.g., https://github.com/lululxvi/deepxde/blob/master/examples/Lorenz_inverse.py

0 replies

athtareq · 2020-03-25T20:41:31Z

athtareq
Mar 25, 2020

I have been following this, thanks for the reply in example ic1 = dde.IC(geom, lambda X: -8 * np.ones(X.shape), boundary, component=0) , is this referring to the IC: y1(-8)=0 or y1(0)=-8 ? Le mer. 25 mars 2020 à 21:37, Lu Lu <notifications@github.com> a écrit :

…

- np.sin is the same as lambda expression, right? This is all about Python grammar. - IC and BC have almost the same usage. - You can find more examples, e.g., https://github.com/lululxvi/deepxde/blob/master/examples/Lorenz_inverse.py — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <https://github.com/lululxvi/deepxde/issues/19#issuecomment-604074115>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOWG24O34K7BUV5WF55LVOTRJJTR5ANCNFSM4LOYV6MA> .

0 replies

dpde101 · 2020-03-25T20:48:03Z

dpde101
Mar 25, 2020
Author

@tareqath haha you are following without telling me wach asahbi hetta nta kheddam hna katsowel, khasna nemchi l Casablanca chi nhar @luluxvi thankyou smuch i understand it now

…

On Wed, Mar 25, 2020 at 9:41 PM tareqath ***@***.***> wrote: Thanks for reply in example ic1 = dde.IC(geom, lambda X: -8 * np.ones(X.shape), boundary, component=0) , is this referring to the IC: y1(-8)=0 or y1(0)=-8 ? Le mer. 25 mars 2020 à 21:37, Lu Lu ***@***.***> a écrit : > > - np.sin is the same as lambda expression, right? This is all about > Python grammar. > - IC and BC have almost the same usage. > - You can find more examples, e.g., > https://github.com/lululxvi/deepxde/blob/master/examples/Lorenz_inverse.py > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <https://github.com/lululxvi/deepxde/issues/19#issuecomment-604074115>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AOWG24O34K7BUV5WF55LVOTRJJTR5ANCNFSM4LOYV6MA > > . > — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <https://github.com/lululxvi/deepxde/issues/19#issuecomment-604076131>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AO3XKEWW226PRL7OPP7WOATRJJUAVANCNFSM4LOYV6MA> .

0 replies

lululxvi · 2020-03-25T22:01:34Z

lululxvi
Mar 25, 2020
Maintainer

@tareqath

lambda X: -8 * np.ones(X.shape) is f(x) = -8
component=0 means that this IC/BC is for the first component, i.e., y(0)

So, it is y(0) = -8.

0 replies

athtareq · 2020-03-25T22:52:02Z

athtareq
Mar 25, 2020

I can't stress this enough: thank you so much ! Le mer. 25 mars 2020 à 23:01, Lu Lu <notifications@github.com> a écrit :

…

@tareqath <https://github.com/tareqath> - lambda X: -8 * np.ones(X.shape) is f(x) = -8 - component=0 means that this IC/BC is for the first component, i.e., y(0) So, it is y(0) = -8. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <https://github.com/lululxvi/deepxde/issues/19#issuecomment-604110508>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOWG24ILWKT2U2M5SBIZ4X3RJJ5MZANCNFSM4LOYV6MA> .

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FailedPreconditionError: Attempting to use uninitialized value dense_22/bias #677

{{title}}

Replies: 18 comments

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

Select a reply

FailedPreconditionError: Attempting to use uninitialized value dense_22/bias #677

dpde101 Mar 18, 2020

Replies: 18 comments

dpde101 Mar 18, 2020 Author

lululxvi Mar 19, 2020 Maintainer

dpde101 Mar 19, 2020 Author

lululxvi Mar 19, 2020 Maintainer

dpde101 Mar 19, 2020 Author

lululxvi Mar 20, 2020 Maintainer

dpde101 Mar 20, 2020 Author

lululxvi Mar 20, 2020 Maintainer

dpde101 Mar 20, 2020 Author

lululxvi Mar 24, 2020 Maintainer

dpde101 Mar 25, 2020 Author

lululxvi Mar 25, 2020 Maintainer

dpde101 Mar 25, 2020 Author

lululxvi Mar 25, 2020 Maintainer

athtareq Mar 25, 2020

dpde101 Mar 25, 2020 Author

lululxvi Mar 25, 2020 Maintainer

athtareq Mar 25, 2020

dpde101
Mar 18, 2020

dpde101
Mar 18, 2020
Author

lululxvi
Mar 19, 2020
Maintainer

dpde101
Mar 19, 2020
Author

lululxvi
Mar 19, 2020
Maintainer

dpde101
Mar 19, 2020
Author

lululxvi
Mar 20, 2020
Maintainer

dpde101
Mar 20, 2020
Author

lululxvi
Mar 20, 2020
Maintainer

dpde101
Mar 20, 2020
Author

lululxvi
Mar 24, 2020
Maintainer

dpde101
Mar 25, 2020
Author

lululxvi
Mar 25, 2020
Maintainer

dpde101
Mar 25, 2020
Author

lululxvi
Mar 25, 2020
Maintainer

athtareq
Mar 25, 2020

dpde101
Mar 25, 2020
Author

lululxvi
Mar 25, 2020
Maintainer

athtareq
Mar 25, 2020