InterDigitalInc / CompressAI

Thank you for your work. I found the default batch size is 10 times the input batch size.
Is this your setting, and what's the point of doing so? Where can I find it?

What are you referring to by the "default batch size" and "input batch size"?

The examples/train.py specifies a training batch size of 16:

CompressAI/examples/train.py

Lines 198 to 200 in ee91d53

    
           parser.add_argument( 
        
               "--batch-size", type=int, default=16, help="Batch size (default: %(default)s)" 
        
           )

The documentation states:

Models were trained with a batch size of 16 or 32

I mean, when I'm training, the batch size I specify is 16, but when I print out a file, it's 160 times. Is this your setting?

That's probably just the default interval for printing the progress bar.

https://github.com/tqdm/tqdm#documentation

The actual batch size is still 16, which you can check by printing out x.shape.

Thank you very much for your answer. I wish you a happy life and work

	parser.add_argument(
	"--batch-size", type=int, default=16, help="Batch size (default: %(default)s)"
	)

The default batch size is 10 times the input batch size