Wang Chenguang
Wang Chenguang
tf 1.11，ok
hello，I also have this problem, I want to know, how to solve?
When i set use_character_lstm in src/parameters.ini to false. After i train a model，when i use the model to do some prediction,it has some wrong，i just set use_pretrained_model in src/parameters.ini to...
I've solved this problem, parameter.ini is in the wrong place
When I comment out this line, this error is gone.
when i up the tensorflow to 1.3，this code is ok。
use: > from datasets import load_dataset > ds = load_dataset("Hello-SimpleAI/HC3-Chinese") > ds.save_to_disk('Hello-SimpleAI/HC3-Chinese') then: use the saved directory, it works for me
谢谢回复 ; 当前问题已解决，后续升级为您说的推理方式： 根据报错再阅读了官方文档发现，当前转换的参数的二进制格式和推理所需要的二进制不符合 > configs (dict，可选) – 其他配置选项，目前支持以下选项：（1）use_binary_format（bool）- 如果被保存的对象是静态图的 Tensor，你可以指定这个参数。如果被指定为 True，这个 Tensor 会被保存为由 paddle 定义的二进制格式的文件；否则这个 Tensor 被保存为 pickle 格式。默认为 False 。 原始拆分代码： ``` new_dict[new_name] = value value_all.append(value) paddle.save(new_dict,model_path_main +...
when this line commented out，it works for me； `# --enable_hybrid_engine \`
it works for me , set model.model to model.module ; line : print_throughput(model.module, args, end - start,args.global_rank)