display the gaph of a .ckpt model

[Tailor2019]

Hello!
@ChWick
@andbue
I'm trying to display the architecture of a .ckpt model
I'm using this code
import tensorflow.compat.v1 as tf tf.disable_v2_behavior() tf.train.import_meta_graph("./0.ckpt.h5") for n in tf.get_default_graph().as_graph_def().node: print(n) with tf.Session() as sess: writer = tf.summary.FileWriter("./output/", sess.graph) writer.close()
But it not working
Please what code to use with these options
--model.layers.0.name --model.layers.0.filters --model.layers.0.kernel_size --model.layers.0.strides --model.layers.0.padding --model.layers.0.activation --model.layers.1.name --model.layers.1.name ................... to display the architecture of a particular model

[Andreas Büttner]

Have a look at

calamari/calamari_ocr/ocr/predict/predictor.py

Lines 20 to 32 in a378469

 def from_checkpoint(params: PredictorParams, checkpoint: str, auto_update_checkpoints=True): 
     DeviceConfig(params.device)  # Device must be specified first 
     ckpt = SavedCalamariModel(checkpoint, auto_update=auto_update_checkpoints) 
     scenario_params = CalamariScenario.params_from_dict(ckpt.dict) 
     scenario = CalamariScenario(scenario_params) 
     predictor = Predictor(params, scenario.create_data()) 
     predictor.set_model( 
         keras.models.load_model( 
             ckpt.ckpt_path + ".h5", 
             custom_objects=CalamariScenario.model_cls().all_custom_objects(), 
         ) 
     ) 
     return predictor 

on how to load a saved model.

[Tailor2019]

Hello!
@andbue
Thanks for your help
But the solution returns me this error :
Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
What can I use as restore operator please to solve this problem?
Thanks a lot!

[Andreas Büttner]

Without your code I can only guess here: maybe the file path is wrong? It has to be something like "path/to/model/0.ckpt" without any further file endings.

[Tailor2019]

Thanks I resolve this problem.