Tao toolkit on aws cloud

[user85597]

Hi Team,

we followed following instruction on installation of tao toolkit on AWS instance. however after installation when we run tao info… we are not getting mask_rcnn option withing the option.

https://docs.nvidia.com/tao/tao-toolkit/text/running_in_cloud/running_tao_toolkit_on_aws.html

Configuration of the TAO Toolkit Instance

task_group: [‘model’, ‘dataset’, ‘deploy’]

format_version: 3.0

toolkit_version: 5.0.0

published_date: 07/14/2023

how do we configure tao toolkit to get this option?

[Morganh]

On this page Running TAO Toolkit on an AWS VM - NVIDIA Docs ,

Below is an old version. Sorry for not updating it.

wget --content-disposition https://api.ngc.nvidia.com/v2/resources/nvidia/tao/cv_samples/versions/v1.2.0/zip -O cv_samples_v1.2.0.zip
unzip -u cv_samples_v1.2.0.zip -d ./cv_samples_v1.2.0 && cd ./cv_samples_v1.2.0

Please refer to https://docs.nvidia.com/tao/tao-toolkit/text/tao_toolkit_quick_start_guide.html???
Below will download latest 5.0 notebook.

wget --content-disposition https://api.ngc.nvidia.com/v2/resources/nvidia/tao/tao-getting-started/versions/5.0.0/zip -O getting_started_v5.0.0.zip
unzip -u getting_started_v5.0.0.zip -d ./getting_started_v5.0.0 && rm -rf getting_started_v5.0.0.zip && cd ./getting_started_v5.0.0

[user85597]

Thank you for the above update. it’s working now.

we are trying to run an Action recognition using TAO ActionRecognitionNet … first testing with a sample dataset to make sure everything works… however, stuck on the following step :

any help why its giving this error?

[Morganh]

Did you follow the default notebook? For this error, please double check the dataset. Please try to run the default notebook and default dataset.

[user85597]

Yes, there is no modification done we just followed the default notebook and it downloaded the default dataset. however, it can not convert video to a png file it gives the above error.

[Morganh]

Please try to check why the shell script failed to run well against the dataset.
The source code is in

NVIDIA-AI-IOT/tao_toolkit_recipes/blob/main/tao_action_recognition/data_generation/preprocess_HMDB_RGB.sh

I recall that other user did not meet this kind of issue.

I think “preprocess README” is not expected. Maybe it is related to the path.

[user85597]

we modify the path and then it started working…

import os
os.chdir(‘tao_toolkit_recipes/tao_action_recognition/data_generation’)

import os
current_working_directory = os.getcwd()
print(current_working_directory)

so what could be the error in the default notebook.?

[Morganh]

Could you save the notebook as .html file and share with us?
Thanks.

[user85597]

After modification we still stuck on training error.

actionrecognitionnet (1).html (832 KB)

[Morganh]

Please delete https://github.com/NVIDIA/tao_tutorials/blob/main/notebooks/tao_launcher_starter_kit/action_recognition_net/specs/train_rgb_3d_finetune.yaml#L1 and retry.

[user85597]

Thank you … however we are using this file to train model… following tutorial from

https://www.youtube.com/watch?v=9cSyNtpIqLc

print(“Train RGB only model with PTM”)

!tao model action_recognition train \

-e $SPECS_DIR/train_rgb_3d_finetune.yaml \

#-r $RESULTS_DIR/rgb_3d_ptm \

results_dir=$RESULTS_DIR/rgb_3d_ptm \

-k $KEY \

model.rgb_pretrained_model_path=$RESULTS_DIR/pretrained/actionrecognitionnet_vtrainable_v1.0/resnet18_3d_rgb_hmdb5_32.tlt \

model.rgb_pretrained_num_classes=5

should we be using different file ?

[Morganh]

@user85597

train_rgb_3d_finetune.yaml

Can you share the yaml file?
This file is actually output_dir: /results/rgb_3d_ptm.

For workaround, please change
output_dir: /results/rgb_3d_ptm

to
results_dir: /results/rgb_3d_ptm