Tensorflow 1.5

[scigeek72]

Hi, I am interested to use your model for our project. But we have the current tensorflow (1.5). But from reading the README, it seems that the code is optimized for Tensorflow version 1.2.1. I want to know if this is going to work on our version of Tensorflow or do we need to downgrade to that specific version.

Thanks,

Tridib

[victorherbemontagne]

Hi @scigeek72 , If you are still looking for a TF 1.5 version of this code, you can check my fork, where I upgrade the code to work on TF 1.5 and Python 3, to test it on new dataset (so working with the same model proposed in the readme file)

[borgr]

Just for convenience https://github.com/victorherbemontagne/pointer-generator

[kailasdayanandan]

The main change for Tensorflow 1.5 is below.

https://stackoverflow.com/questions/49206318/typeerror-unsupported-operand-types-for-int-and-flag

However when trying to train with coverage enabled in Python 3 with becxer fork and also in the above mentioned fork it hangs after this.

INFO:tensorflow:Starting standard services. INFO:tensorflow:Saving checkpoint to path ./pg_log/pg_coverage/train/model.ckpt INFO:tensorflow:global_step/sec: 0 INFO:tensorflow:Starting queue runners. INFO:tensorflow:Created session. INFO:tensorflow:starting run_training INFO:tensorflow:running training step...

[nozomi233]

The main change for Tensorflow 1.5 is below.

https://stackoverflow.com/questions/49206318/typeerror-unsupported-operand-types-for-int-and-flag

However when trying to train with coverage enabled in Python 3 with becxer fork and also in the above mentioned fork it hangs after this.

INFO:tensorflow:Starting standard services. INFO:tensorflow:Saving checkpoint to path ./pg_log/pg_coverage/train/model.ckpt INFO:tensorflow:global_step/sec: 0 INFO:tensorflow:Starting queue runners. INFO:tensorflow:Created session. INFO:tensorflow:starting run_training INFO:tensorflow:running training step...

excuse me,have you solve the problem?

[ZhichaoOuyang]

The main change for Tensorflow 1.5 is below.

https://stackoverflow.com/questions/49206318/typeerror-unsupported-operand-types-for-int-and-flag

However when trying to train with coverage enabled in Python 3 with becxer fork and also in the above mentioned fork it hangs after this.

INFO:tensorflow:Starting standard services. INFO:tensorflow:Saving checkpoint to path ./pg_log/pg_coverage/train/model.ckpt INFO:tensorflow:global_step/sec: 0 INFO:tensorflow:Starting queue runners. INFO:tensorflow:Created session. INFO:tensorflow:starting run_training INFO:tensorflow:running training step...

I have the same problem, have you solve the problem?

[yashkumaratri]

If you want to run it then you can install TF 1.2.1 using this in a venv

pip install --ignore-installed --upgrade https://github.com/mind/wheels/releases/download/tf1.2.1-gpu/tensorflow-1.2.1-cp36-cp36m-linux_x86_64.whl

and it will run TF 1.2.1 on GPU with

nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2017 NVIDIA Corporation Built on Fri_Sep__1_21:08:03_CDT_2017 Cuda compilation tools, release 9.0, V9.0.176