Error when reusing already trained model - bad magic number 0
Hi several weeks ago I've asked for reuse an existing model to transfer the classification model from a year to another:
Here the steps I followed:
To re-use an existing model
- At first launch iota2 once to generate a model, thanks to configuration_file_A.cfg
- Then re-launch iota2 until sampling processes thanks to configuration_file_B.cfg
- copy the model generated by configuration_file_A.cfg in
/output/config_B/model
- re-launch configuration_file_B.cfg from the classification stage.
This trick can only work if configuration_file_A.cfg and configuration_file_B.cfg will generate the same number of features to feed the model.
Here the results of the steps:
1/2 - The first init/Sampling seems working well with no error.
3 - In order to run the next step classification/validation I've copied the file 'model_1_seed_0.txt' in the directory "/output/config_B/model" and the file 'configModel.cfg' in the directory "/output/config_B/config_model".
4 - I re-launch with the new configuration file and I get an error: "Running step 15: Generate classifications (1 tasks) ERROR 1: ~/IOTA2_TEST_S2/sensor_data/T19UEQ/S2A_MSIL2A_20160823T153602_N9999_R111_T19UEQ_20200915T143648.SAFE/GRANULE/L2A_T19UEQ_A006114_20160823T153558/IMG_DATA/R10m/T19UEQ_20160823T153602_B03_10m.jp2:Not a TIFF or MDI file, bad magic number 0 (0x0)"
The file is on the directory tree. Don't understand the "bad magic number" error. Thanks