Predictions cannot be generated by specifying TF name #5
Description
Problem: I can generate predictions only if I explicitly provide the full path to the trained model in the call to maxATAC predict. I cannot generate predictions if I use the syntax “–tf $TF_NAME”, which is what is described in the vignette.
Error: When I try to generate predictions by indicating the name of the TF (e.g., CTCF), I get an error message that reads: “OSError: SavedModel file does not exist at: CBX2/{saved_model.pbtxt|saved_model.pb}”.
Failed Fixes: I tried explicitly setting the DATA_PATH variable in predict.py (lines 53 and 55) to the local location of all model files, but this was not successful and gave me the same error message.
Additional Information: I used the versions of keras/tensorflow included in the requirements.txt file for the quant-maxATAC repo. I wonder if line 345 in prediction_tools.py might be the issue? I found the following discussion that might be useful.
Please let me know if you have any questions!
Akshata Rudrapatna