when i run the eval_only example ,ie. -e True, something wrong

[ross-Hr]

AssertionError: Some Python objects were not bound to checkpointed values, likely due to changes in the Python program: [<tf.Variable 'attention_layer_2/q:0' shape=(512, 8, 64) dtype=float32, numpy=
...
there is my config
args = ap.parse_args(['-data_path','../dataset/great',
'-vocabulary_path','../vocab.txt',
'-config','../config.yml',
'-m','../models/preTModels/ggnn-checkpoint',
'-l','../models/preTModels/ggnn-checkpoint/ggnn-log.txt',
'-e','True'])

[ross-Hr]

i just use the default config.yml

[VHellendoorn]

Hi! It looks like you are importing the ggnn checkpoint, but since you never alter the config.yml file it is trying to load this into a GREAT model. Please note line 2 of the config, where you can specify the model type. By default it is set to "great", so for you you'd want to change that to "ggnn". If you were trying to restore the GREAT model, please point it to the appropriate pretrained model directory and log file.

Hope this helps,
-Vincent

[ross-Hr]

Hi! It looks like you are importing the ggnn checkpoint, but since you never alter the config.yml file it is trying to load this into a GREAT model. Please note line 2 of the config, where you can specify the model type. By default it is set to "great", so for you you'd want to change that to "ggnn". If you were trying to restore the GREAT model, please point it to the appropriate pretrained model directory and log file.

Hope this helps,
-Vincent

That works!
Thank you very much for your contribution！