Runing Error

[lkytal]

Error when trying to do prediction, seems like the graph is disconnected? Clone the last version and use your model from figshare.

File "/root/prosit/server.py", line 28, in predict
result = prediction.predict(tensor, model, model_config)
File "/root/prosit/prediction.py", line 14, in predict
model.compile(optimizer="adam", loss="mse")

[omitted.....]

ValueError: Tensor("out_target:0", shape=(?, ?), dtype=float32) must be from the same graph as Tensor("out/Reshape:0", shape=(?, ?), dtype=float32)

[gessulat]

I am not able to reproduce this error. Can you give me details about your system?

Prosit needs nvidia-docker and CUDA installed. Then prediction should work with
make predict DATA=/path/to/your.hdf5 MODEL=/our/model/provided/at/figshare/
When using Prosit without make and Docker please make sure all dependencies are in line with the docker file.

Be sure to have the model directory has all the necessary files (model.yml, config.yml and weights.hdf5) all can be found at https://figshare.com/projects/Prosit/35582

[tobigithub]

Ff you run "make jump ..." it will not produce the 'ValueError: Tensor" error.
Its basically an issue with the current prosit makefile or prosit python scripts or prosit setups.
The current prosit version supports CUDA 9 and CUDA 10.
We are still trying to figure that out.

[gessulat]

See #8