Fix some bug in detect.py

[YMCAlan]

• modify the parameter attempt_download of the function.
• modify non_max_suppression function.

But I don't know the actual output of the model, so this modify is just a temporary.

[lsutardja]

yolov9/val_dual.py

Lines 190 to 199 in 33fc75e

	with dt[1]:
	preds, train_out = model(im) if compute_loss else (model(im, augment=augment), None)
	# Loss
	if compute_loss:
	preds = preds[1]
	#train_out = train_out[1]
	#loss += compute_loss(train_out, targets)[1] # box, obj, cls
	else:
	preds = preds[0][1]

Here is the validation script. The correct change to NMS should be prediction = prediction[0][1].

Also see #51 (comment)

[cobaner3]

good bro

[YMCAlan]

Thank you for your response, I have modify the code, and provide the dual flag for detect.py.
I consider the pt file should provide the some attribute for us to check the model type.
@lsutardja @cobaner3