Models: eval.py problem

[xxktju]

INFO:tensorflow:Evaluating on val set
INFO:tensorflow:Performing single-scale test.
INFO:tensorflow:Eval num images 1449
INFO:tensorflow:Eval batch size 1 and num batch 1449
INFO:tensorflow:Waiting for new checkpoint at ./model/model.ckpt

when i run the eval.py, it stoped at the "Waiting for new checkpoint at ./model/model.ckpt",and no go further evaluation

[robieta]

I apologize, but I am having a hard time understanding what the problem is, where the problem is, and what version it affects. Please resubmit and pay attention to the issue template (https://github.com/tensorflow/tensorflow/issues/new). Please provide all the information it asks. Thank you.

[AnameZT]

I crossed with the same problem several days ago，and have solved it. I hope we have made the same problem,so you may try the following method.
I think you may check your checkpoint_dir and specify it to a folder which contains both a checkpoint document and the three other model-ckpt files(meta/index/data).Then you should add the follow text to your checkpoint document : model_checkpoint_path: "model.ckpt-xxxxx"
all_model_checkpoint_paths: "model.ckpt-xxxxx"
I hope it can be helpful.

[lizleo]

@AnameZT
Hello, I have three files in the checkpoints_dir as frozen_inference_graph.pb , model.ckpt.data-00000-of-00001 and model.ckpt.index, and I met the same problem, what should I do to load the checkpoint? Could you please give me a hand , many thanks for your help~

[AnameZT]

@lizleo
The files you put into the checkpoint_dir is the base model you need use during the training phase.When you evaluation your dataset,the checkpoint_dir should be specified to path as the train_logdir does(the path is necessary during training),just to check it out.
hope i'm not misunderstand your problem.

[lizleo]

@AnameZT
Hello, I want to evaluate DeepLab on Cityscapes Semantic Segmentation Dataset according to https://github.com/tensorflow/models/blob/master/research/deeplab/g3doc/cityscapes.md, so I download the checkpoints that have been pretrained on Cityscapes train_fine set in the https://github.com/tensorflow/models/blob/master/research/deeplab/g3doc/model_zoo.md, and then I meet the "waiting for new checkpoint "problem. I didn't train the model, just want to evaluate the results to compare my previous net.

[aspenstarss]

you should add
--max_number_of_evaluations=1 for eval.py
and
--max_number_of_iterations=1 for vis.py
it's a feature, not is a bug.

[LemonYear]

you should add
--max_number_of_evaluations=1 for eval.py
and
--max_number_of_iterations=1 for vis.py
it's a feature, not is a bug.

sorry,it doesn't work for me

[Saptakatha]

@aspenstarss Sorry, it didn't work for me also.

[YknZhu]

+1 to @AnameZT's solution. Usually checkpoint_dir would direct to the logging directory generated during training. If you would like to try on published models in model zoo, please add a new file named checkpoint in the same directory with ckpt files, with content

model_checkpoint_path: "<CKPT_FILES>"
all_model_checkpoint_paths: "<CKPT_FILES>"

where <CKPT_FILES> should be the checkpoint prefix (shared name before '.index' / '.data*', '.meta').

Hope this helps.

[RajatGarg45]

AnameZT

I have all the three files specified by you along with the checkpoint file which is automatically generated when I run training (containing required information) but still eval.py and vis.py gets stuck at Waiting for checkpoint.

[lolitsgab]

@RajatGarg45 Did you figure it out?