Tao automl

[behna.rahimi]

Please provide the following information when requesting support.

• Hardware : GTX 3090
• Network Type : Classification
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here): TAO API
• Training spec file(If have, please share here)
• How to reproduce the issue ? (This is for errors. Please share the command line and the detailed log here.)

Hello,
I am using the client notbook for AUTOML on classification (multiclass), When I want to use the effientnet B0 or B1, I have problems. Resnets and darknet seems to be working.
On EfficinetnetB1, The trainng stuck at epoch zero and dont show any error message.
On EfficinetnetB0, it can not find any pretrain model of classification models.
Could you help with this?

[Morganh]

@behna.rahimi

On EfficinetnetB1, The trainng stuck at epoch zero and dont show any error message.
On EfficinetnetB0, it can not find any pretrain model of classification models.

Could you share the full log or save the .ipynb file to html file, and then upload here?

[behna.rahimi]

for B0, it is not finding the pretrained model and gived none: I trid different versiones of B0( relu,swish or just B0. no luck!

[behna.rahimi]

And this is for B1, the full file with deleted key cell. It Stuck at zero epoch.
old-classification(1).html (709.3 KB)

[Morganh]

@behna.rahimi

I trid different versiones of B0( relu,swish or just B0. no luck!

Please get the link from Open Images Pre-trained EfficientDet - NVIDIA Docs → TAO Pretrained EfficientDet | NVIDIA NGC

@behna.rahimi

And this is for B1, the full file with deleted key cell. It Stuck at zero epoch.

Did you download the notebook via the step in

TAO Toolkit Quick Start Guide - NVIDIA Docs

wget --content-disposition https://api.ngc.nvidia.com/v2/resources/nvidia/tao/tao-getting-started/versions/4.0.0/zip -O getting_started_v4.0.0.zip
unzip -u getting_started_v4.0.0.zip -d ./getting_started_v4.0.0 && rm -rf getting_started_v4.0.0.zip && cd ./getting_started_v4.0.0

Seems that your .ipynb is different.

[behna.rahimi]

yes, download link is same. the file is in folder: 'getting_started_v4.0.0/notebooks/tao_api_starter_kit/client/automl"
I suspect there some modification on the train config file thatshould be done (number of lyers or activation). did you be able to run with B1?

For B0, it is looking for ptm item number in the notebook, where would I put the link from your suggestion in the notebook?

[behna.rahimi]

Changin the “Find pretrained model” section codes in the notebook will solve the problem:
1-
pretrained_map = {“classification” : “pretrained_EfficientDet:efficientnet_b0”,
“multitask_classification” : “pretrained_classification:resnet10”}
2-
if metadata_network_arch == network_arch:
ptm_id = ptm_metadata[“id”]
break

[Morganh]

Thanks for the info. Appreciate it.