Eval batch size always 0

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

hello i’m using deepstream 5.0, and tlt

i’m trying to train my model
but evaluation fail every time

please help

3

Is there something wrong in above tfrecords path?

4

thanks for reply
unfortunately path is correct
because training is running before evaluation

5

Could you double check why there is double “//” ?

/workspace//tlt-experiments//train//tfrecord//

6

it’s just my trying result
because some path using single ‘/’ cause path error
is it cause error???
but it was working before evaluation

7

How many images of your val dataset?
If possible, please share your full log when you run tlt-dataset-convert.

8

sure
i attached log file!

Thanks for helping me

9

According to your tfrecord log, there are only 3 classes.
b’person’: b’person’
b’car’: b’car’
b’target’: b’target’

But you set wrong target_class_mapping in the training spec.
Could you change to below and retry?
target_class_mapping {
key: “person”
value: “person”
}
target_class_mapping {
key: “car”
value: “car”
}
target_class_mapping {
key: “target”
value: “target”
}