Explanation of parameter max_len in tokenizer.encode ？

[Kiris-tingna]

It seems that this parameter should be capable with the pretrained BERT model , but where to find the setting in donwleded bert_config?

[CyberZHG]

keras-bert/demo/load_model/load_and_extract.py

Line 9 in 5380fa4

print('CONFIG_PATH: $UNZIPPED_MODEL_PATH/bert_config.json')

[houchangtao]

In the load and extract demo, which model you are loading? I tried the model: uncased_L-12_H-768_A-12 but get a different result as you got.

[CyberZHG]

@houchangtao chinese_L-12_H-768_A-12

[houchangtao]

@CyberZHG Thanks!