报错“ValueError: invalid literal for int() with base 10: './saved_model/unet_pet/10”

[lakebodhi]

根据https://github.com/PaddlePaddle/PaddleSeg/blob/release/v0.2.0/docs/usage.md的示例操作，执行
python pdseg/train.py --do_eval \ --use_tb \ --tb_log_dir train_log \ --cfg configs/unet_pet.yaml \ BATCH_SIZE 4 \ TRAIN.PRETRAINED_MODEL_DIR pretrained_model/unet_bn_coco \ SOLVER.LR 5e-5
程序报错：

#1 visualize image path: visual\visual_results\Abyssinian_201.png
Traceback (most recent call last):
File "pdseg/train.py", line 487, in
main(args)
File "pdseg/train.py", line 474, in main
train(cfg)
File "pdseg/train.py", line 456, in train
log_writer=log_writer)
File "D:\remote_sensing\PaddleSeg\pdseg\vis.py", line 229, in visualize
epoch = int(ckpt_dir.split(os.path.sep)[-1])
ValueError: invalid literal for int() with base 10: './saved_model/unet_pet/10'

操作系统windows10 64位 cpu训练

[nepeplwu]

@lakebodhi 感谢反馈，该问题已经修复，使用最新代码试试看

[ArtemiyFirsov]

Not fixed.
Getting the same error on Ubuntu 18.04 Docker installation.
Running docker with
sudo docker run --gpus=all -it -v $(pwd)/DeepSpeech:/DeepSpeech hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu /bin/bash

Updated the code to the latest develop branch.

Error happens on infer
CUDA_VISIBLE_DEVICES=0 python infer.py --infer_manifest test_deep_speech.txt

Manifest I am trying to infer
{"audio_filepath": "/DeepSpeech/test_deep_speech.wav", "duration": 30.960, "text": ""}

So it does not even goes to the infer, it fails on data read, because the code tried to convert text to int.

UPDATE: sorry, this is in the DeepSpeech repo