isobar-us / multilabel-image-classification-tensorflow Goto Github PK

Hi, great project. I was wandering is there a way to restore last given checkpoint and continue training without downloading module from TFhub?
Can you give some example for Mobilenet maybe?

Thanks

Thank you for putting together the object detection code for sagemaker.

When running a training job, we are getting No such file or directory: '/opt/ml/input/data/checkpoint/' not round error
from File "/opt/ml/code/train", line 95.

Can you please confirm if this location need to be explicitly created or is it something that should be mounted by sagemaker

Also, do we provide the directory name (like extracted "ssd_mobilenet_v2_quantized_300x300_coco_2019_01_03") under checkpoint setting in training job or provide ssd_mobilenet_v2_quantized_300x300_coco_2019_01_03.tar.gz under model configuration?

Appreciate your feedback.

HI

Nice work !!

i am getting an error while running the training job. Any idea?

AlgorithmError: Exception during training: [Errno 2] No such file or directory: '/opt/ml/input/data/checkpoint/' Traceback (most recent call last): File "/opt/ml/code/train", line 95, in file_list = os.listdir(tfrecord_pretrained_checkpoint_path) FileNotFoundError: [Errno 2] No such file or directory: '/opt/ml/input/data/checkpoint/'

Hi, I'm fairly new to Tensorflow object detection and SageMaker, but like others I have to say thanks for making this! I did all the steps in the README, plus I'm using the same configuration.config file, and I believe I configured the input channels correctly. However, The job always fails and I only get a single log from Cloudwatch:

Any help would be appreciated, thanks!

I would like to use Tensorboard to monitor training loss and accuracy. I have been using TensorFlow API to train model directly, but I'm not sure how this would happen through AWS Sagemaker training.

Also, how would you go about continuing training if the model isn't converged in the defined number of steps? Like is saving the checkpoints and continuing training from there possible?

I have just started using AWS Sagemaker, so excuse me if this is a silly question.

Hello

Firstly, thanks for putting this work out in the wild! Im curious to get an environment set up to run a multi label classification training session - and from experience know some times there's particularities on versions.

Was this built / run against a particular version of Tensorflow? Any other requirements?

Thank you!

I've successfully trained the model, but when I'm trying to run label_image.py it throws
ValueError: Cannot feed value of shape (1, 229, 229, 3) for Tensor 'import/Placeholder:0', which has shape '(?, 299, 299, 3)'
the command im running is
python label_image.py --input_height=229 --input_width=229 --graph=/tmp/output_graph.pb --labels=labels.txt --input_layer=Placeholder --output_layer=final_result --image=test.jpg