Why is the model file generated in .hdf5 format, and can a .tlt format file be generated? yolov4_tiny , cspdarknet_tiny

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1 
Why is the model file generated in .hdf5 format, and can a .tlt format file be generated?

print("To run with multigpu, please change --gpus based on the number of available GPUs in your machine.")
!tao yolo_v4_tiny train -e $SPECS_DIR/yolo_v4_tiny_train_kitti.txt \
                   -r $USER_EXPERIMENT_DIR/experiment_dir_unpruned \
                   -k $KEY \
                   --gpus 1
2

Since TAO5.0 version, the yolo_v4_tiny model will generate .hdf5 format file.
In previous TAO versions, it is .tlt format which is actually an encrypted file of .hdf5 file.
May I know why you need a .tlt file?

3

I need to deploy a .tlt file. How can I get the .tlt file?

4

Can the previous version of TAO still be used?

5

Do you mean you want to deploy .etlt file, right? The .etlt is actually an encrypted version of onnx file.
In TAO5.0, you will get the onnx file. Then you can deploy it directly.

Yes, you can docker pull the old dockers from https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/containers/tao-toolkit/tags → nvcr.io/nvidia/tao/tao-toolkit:4.0.1-tf1.15.5

6

If I already have a newer version of the image, will there be any conflict after pulling the older version of the image?

7

Will it default to using the older Docker image?

8

No conflict.

If you use “tao launcher”, it will point to the docker when you install it. See “tao info --verbose”.

If you use “docker run” method, there is no default.

9

thank you

10

How to use an old image for TAO training

11

You can run
docker run --runtime=nvidia -it --rm nvcr.io/nvidia/tao/tao-toolkit:4.0.1-tf1.15.5 /bin/bash

Then inside the docker,
# yolo_v4_tiny train xxx

12

I have downloaded a YOLOv4_tiny notebook, how should I use an old Docker image with it?

13

Refer to TAO Toolkit Quick Start Guide - NVIDIA Docs.
Run below inside the docker.
jupyter notebook --ip 0.0.0.0 --port 8888 --allow-root

14

Which TAO notebook instance should I run with ‘nvcr.io/nvidia/tao/tao-toolkit:4.0.1-tf1.15.5’?

15

Can you create a new forum topic?

16

thank you

17

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.