Dear Knimers
I eventually found a workaround to solve this Enigma =>
On a Conda command-line window :
-
delete all directories related to tensorflow => C:\Users\Worker\AppData\Roaming\Python\Python36\site-packages\tensorflow\
-
Execute the following Conda commands :
134 conda env create --file C:\Users\Worker\py36_knime.yml --name py36_knime_tf_cpu
135 conda activate py36_knime_tf_cpu
136 python
137 > import numpy
138 quit() # No complain 
139 conda install tensorflow-base=1.12 -c anaconda --freeze-installed
140 python
141 > import tensorflow
142 quit() # No complain
143 conda install keras=2.2.4 -c anaconda --freeze-installed
144 python
145 > import keras
146 quit() # No complain
147 conda install h5py=2.8 -c anaconda --freeze-installed
148 python
# same import tests as before to check again everything works as before
149 >>> import tensorflow as tf
150 >>> from tensorflow.python.client import device_lib
151 >>> device_lib.list_local_devices()
2019-11-15 13:39:40.964411: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
[name: “/device:CPU:0”
device_type: “CPU”
memory_limit: 268435456
locality {
}
incarnation: 16165568108189707881
]
152 >>> import keras
153 Using TensorFlow backend.
153 >>> import pandas
153 >>> import matplotlib
153 >>> quit() # No complain
149 => run KNIME & activate py36_knime_tf_cpu environment. it worked at this point !
======================================
IF GPU installation then =>
134 conda env create --file C:\Users\Worker\py36_knime.yml --name py36_knime_tf_gpu
135 conda activate py36_knime_tf_gpu
136 python
137 > import numpy
138 quit() # No complain
139 conda install tensorflow-gpu=1.12 -c anaconda --freeze-installed
140 python
141 > import tensorflow
142 quit() # No complain
143 conda install keras=2.2.4 -c anaconda --freeze-installed
144 python
145 > import keras
146 conda install h5py=2.8 -c anaconda --freeze-installed
147 quit() # No complain
148 python
# same import tests as before to check again everything works fine
149 >>> import tensorflow as tf
150 >>> from tensorflow.python.client import device_lib
151 >>> device_lib.list_local_devices()
2019-11-15 17:37:49.714012: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
2019-11-15 17:37:49.894883: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1432] Found device 0 with properties:
name: Quadro P2000 major: 6 minor: 1 memoryClockRate(GHz): 1.607
pciBusID: 0000:01:00.0
totalMemory: 4.00GiB freeMemory: 1.71GiB
2019-11-15 17:37:49.902258: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0
2019-11-15 17:37:50.408704: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-11-15 17:37:50.412570: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988] 0
2019-11-15 17:37:50.415250: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0: N
2019-11-15 17:37:50.417994: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/device:GPU:0 with 1425 MB memory) -> physical GPU (device: 0, name: Quadro P2000, pci bus id: 0000:01:00.0, compute capability: 6.1)
[name: “/device:CPU:0”
device_type: “CPU”
memory_limit: 268435456
locality {
}
incarnation: 1313116399824212057
, name: “/device:GPU:0”
device_type: “GPU”
memory_limit: 1495046552
locality {
bus_id: 1
links {
}
}
incarnation: 8001539658469800458
physical_device_desc: “device: 0, name: Quadro P2000, pci bus id: 0000:01:00.0, compute capability: 6.1”
]
152 >>> import keras
153 Using TensorFlow backend.
153 >>> import pandas
153 >>> import matplotlib
153 >>> quit() # No complains
149 => run KNIME & activate py36_knime_tf_gpu environment. it worked at this point !
==========================================
My goal was to do DL + cheminformatics on these environments so just for information I installed scikit-learn & RDKIT too as follows :
conda install rdkit -c conda-forge --freeze-installed
conda install scikit-learn -c anaconda --freeze-installed
After trying on different workflows and jupyter notebooks, making use of all these packages, everything worked fine, so I think the problem is solved unless proven otherwise.
Thanks @mlauber71 & @MarcelW for your help and thanks to KNIME !
Ael