Error: no MLCTrainingGraph has been found about tensorflow_macos

[anna-tikhonova]

@nikdata Thank you very much for reporting this issue! We've investigated this failure and have a fix. We will include the fix in the next update.

[hcancelik]

I also get Compute: Operation received an exception: Compute: No MLCTrainingGraph has been found. when using
disable_eager_execution().

[nikdata]

So I reran everything, and now I get a different error message after running "mlcompute.set_mlc_device(device_name='gpu')":

AttributeError: module 'tensorflow.python._pywrap_util_port' has no attribute 'IsAppleMLCEnabled'

Is this suggesting that the install of the TensorFlow package did not work?

[dkgaraujo]

Hi! Thanks for sharing your experience. What I would do is the following:

1. try running another code, such as this one in #10 and see if it works.
2. if it also does not work, re-install the whole thing as per the README instructions.
3. after successfully installing, don't run any code directly. Do the following steps to check you have a functioning tensorflow installation.
   3.1) in the Terminal, type ~/tensorflow_macos_venv/bin/python3
   3.2) when the python sign >>> appears, type:

import tensorflow as tf
from tensorflow.python.compiler.mlcompute import mlcompute
mlcompute.set_mlc_device(device_name = 'gpu')

4. The code above should run OK. You may even get a warning that TF is slow in eager mode when using the GPU. This is a good sign. It should mean you are good to go.

---

I hope the above helps you and others out!

[nikdata]

@dkgaraujo Thanks for the suggestions. Trying the commands through terminal instead of an IDE did work for me. However, the error still persisted. It seems the error of no MLCTrainingGraph is tied to the dropout layers. If I remove the dropout layers, the model works with no issues.

I'd say this issue and issue #23 are related since the root cause seems to be the dropout layer. The code posted in issue #25 works just fine (note that no dropout layer is used).

[sarmadali20]

Hi! Thanks for sharing your experience. What I would do is the following:

1. try running another code, such as this one in #10 and see if it works.
2. if it also does not work, re-install the whole thing as per the README instructions.
3. after successfully installing, don't run any code directly. Do the following steps to check you have a functioning tensorflow installation.
   3.1) in the Terminal, type ~/tensorflow_macos_venv/bin/python3
   3.2) when the python sign >>> appears, type:

import tensorflow as tf
from tensorflow.python.compiler.mlcompute import mlcompute
mlcompute.set_mlc_device(device_name = 'gpu')

1. The code above should run OK. You may even get a warning that TF is slow in eager mode when using the GPU. This is a good sign. It should mean you are good to go.

I hope the above helps you and others out!

I had to uninstall using pip and pip3 the tensorflow and re-installed the tensorflow_macos then all the jupyter stuff to make it work!!

[anna-tikhonova]

@nikdata Could you please try out the updated release and let us know if you still see the issue you reported? Thank you!

[nikdata]

@anna-tikhonova Works great! I didn't have any errors while running the script. While updating the package, I did have to manually reinstall a few of the python libraries to specific versions (e.g., gast, grpcio, numpy, protobuf) since the versions I had were newer than the ones required.

[traderpedroso]

Hi! Thanks for sharing your experience. What I would do is the following:

1. try running another code, such as this one in #10 and see if it works.
2. if it also does not work, re-install the whole thing as per the README instructions.
3. after successfully installing, don't run any code directly. Do the following steps to check you have a functioning tensorflow installation.
   3.1) in the Terminal, type ~/tensorflow_macos_venv/bin/python3
   3.2) when the python sign >>> appears, type:

import tensorflow as tf
from tensorflow.python.compiler.mlcompute import mlcompute
mlcompute.set_mlc_device(device_name = 'gpu')

1. The code above should run OK. You may even get a warning that TF is slow in eager mode when using the GPU. This is a good sign. It should mean you are good to go.

I hope the above helps you and others out!

I had to uninstall using pip and pip3 the tensorflow and re-installed the tensorflow_macos then all the jupyter stuff to make it work!!

when you do not use disable_eagler_execution() the learning take longer the apple version even the cpu works faster only when you using eager disabled and gpu tranning in parallels only works when eagler is disabled for any video card on version 2.4! so still getting error

Operation received an exception: Cannot find cached MLCTrainingGraph for the LSTM.