Graph visualization failed, in Graph mode [TF 2.2, TB 2.2.1]

[OverLordGoldDragon]

Eager works, though partially - Graph doesn't work at all. Suggested thread doesn't reveal much.

import shutil
import tempfile
import numpy as np
import tensorflow as tf
tf.compat.v1.disable_eager_execution()

from tensorflow.keras.layers import Input, Dense
from tensorflow.keras.models import Model
from tensorflow.keras.callbacks import TensorBoard

#%%############################################
ipt = Input(shape=(16,))
out = Dense(16)(ipt)
model = Model(ipt, out)
model.compile('adam', 'mse')

logdir = tempfile.mkdtemp()
print('tensorboard --logdir="%s"' % logdir)
tb = TensorBoard(logdir, write_graph=True)

#%%############################################
x = y = np.random.randn(160, 16)
model.fit(x, y, batch_size=32, callbacks=[tb])

# shutil.rmtree(logdir)

---

Environment info:

• Browser: Google Chrome v83.0.4103.61 x64 [also tried Firefox Developer Edition]
• System: Windows 10 x64, GTX 1070, i7-7700HQ 2.8 GHz

[stephanwlee]

I looked at it a bit and it looks like you are using a Functional style of Keras (

tensorflow/tensorflow/python/keras/engine/training.py

Lines 236 to 237 in a3aec8d

	from tensorflow.python.keras.engine import functional # pylint: disable=g-import-not-at-top
	return functional.Functional(*args, **kwargs)

).

While I am confident that Keras will generate op graph when using Sequential model (a special case of Functional), I am not too certain op graphs will be generated in all cases of Functional. It is inconclusive with my cursory reading of https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/keras/engine/functional.py as it is quite dense. @omalleyt12 could you enlighten us a bit?

[XiaotianZou]

Also having the same issue. Did you solve this problem eventually?

[garryyan2]

Greetings! I am very new to Python and Tensorflow. I am just trying to find out where is a good place to start. It seems Keras is a good starting point. However, I ran into the same issue. I can see the SCALARS' plots. However, I can't see the model in GRAPHS. I just used a very simple Sequential model as given in a web page tutorial. I understand that Tensorflow Version 2.0 made some changes. I was hoping it's a simple fix in my code to your experts. Thank you in advance.

train_2b.txt

[hy2610]

I also have the same issue. When I used sequential model and model.fit(), there is a tab for graph, but it shows" Graph visualization failed". Also, the example code for @tf.function does not works for me either.

[acarl005]

Also having this issue with my sub-classed model. I'm using TF 2.4 and Python 3.7.

[hitskyer]

Also having the same issue. Did you solve this problem eventually?

[nikitamaia]

Hi @OverLordGoldDragon, is there a particular reason you are disabling eager execution? In TF2 tf.compat.v1.disable_eager_execution() is generally not recommended.

[OverLordGoldDragon]

@nikitamaia Back then it was recommended for speed and memory advantages; I also ensured Graph & Eager compatibility of my packages for users.

I've switched to Pytorch and no longer seek to resolve any TF bugs. Keeping issue open for others.

[sushreebarsa]

Was able to reproduce the issue in TF v2.5 ,please find the gist here..Thanks !

[tensorflowbutler]

Hi There,

This is a stale issue. As you are using an older version of tensorflow, we are checking to see if you still need help on this issue. Please test the issue with the latest TensorFlow (TF2.7 and tf-nightly). If the issue still persists with the newer versions of TF, please feel free to open it in keras-team/keras repository by providing details about the issue and a standalone code to reproduce the issue. Thanks!

Please note that Keras development has moved to a separate Keras-team/keras repository to focus entirely on only Keras. Thanks!

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No