training

[strWuYaoJu]

running: python keras-retinanet/keras_retinanet/bin/train.py --snapshot-path models csv image_labels/train_data.csv image_labels/class.csv --val-annotations image_labels/val_data.csv

error: TypeError: type object got multiple values for keyword argument 'training'

[gosha20777]

I've get the same error

[GlitchBox]

I've got the same error. Why did it occur?

[gosha20777]

@GlitchBox I think that this implementation is not compartable with tf 2.x and keras 2.3.1

You can use my fork of it witch supports the latest tf and keras versions https://github.com/lacmus-foundation/lacmus

[GlitchBox]

@gosha20777 thank you so much!

[hgaiser]

A PR to fix compatibility would be welcome :)

[wenwenwenwenshei]

use keras==2.3.1

[GeorgeVJose]

It was working for me in the beginning, but the problem started when I changed the anchors in the config file. Also tried keras==2.3.1 but the problem still exists.

[akramaskar]

Any updates here? Tried running it with keras==2.3.1 but another error came up AttributeError: module 'tensorflow.python.framework.ops' has no attribute '_TensorLike'

[Zhejing-Chin]

I am using: !pip install tensorflow==1.14.0 !pip install keras==2.3.0 !pip install numpy==1.16.1

and now my !keras_retinanet/bin/train.py --freeze-backbone --random-transform --weights {'/content/keras-retinanet/_pretrained_model.h5'} --batch-size 8 --steps 500 --epochs 10 csv annotations.csv classes.csv is finally training epoch

[udaylunawat]

Any way to make it work with tensorflow 2.3.0? I've managed to make it work with 2.2.0 though.

[udaylunawat]

Any updates here? Tried running it with keras==2.3.1 but another error came up AttributeError: module 'tensorflow.python.framework.ops' has no attribute '_TensorLike'

Got the same error with tf 2.3 and keras 2.3.1

[vehansayvazi]

changing to keras 2.4.0 fixed this for me

[tleyden]

pip install tensorflow==2.3.1 && pip install keras==2.4.0 worked for me and got past the AttributeError: module 'tensorflow.python.framework.ops' has no attribute '_TensorLike' error I was getting with tf 2.3 + keras 2.3.1

[stale[bot]]

This issue has been automatically marked as stale due to the lack of recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[sharathm451]

@GlitchBox I think that this implementation is not compartable with tf 2.x and keras 2.3.1

You can use my fork of it witch supports the latest tf and keras versions https://github.com/lacmus-foundation/lacmus

bro could you elaborate plz. I am using imageai.objectdetection and got this error. what does it mean like fork your repo, is it to download Retinanet model and give a try? coz I have no idea of downloading tf and keras from git fork.