rpy to tensor

[divagora]

rpy to tensor

[scarlehoff]

I don't think for this case in particular you will see a gain of performance by using VegasFlow. VegasFlow is most useful when you whole problem can be cast as a set of tensorflow primitives (or more generally, tensors). In your case I see you are using an R package so I believe you won't see any advantages with respect to vegas (unless you aim to use the parallelization over function returns see docs)

But here you have some answers to your questions:

In order to change any object to a tensor, you can follow the same strategy you are following but instead of using np.array you can use tf.constant. Another possibility is to enable eager mode so that you can integrate an arbitrary numpy function. See docs for that: https://vegasflow.readthedocs.io/en/latest/examples.html#integrating-a-numpy-function

In order to change the integration limits (see docs https://vegasflow.readthedocs.io/en/latest/how_to.html#changing-the-integration-limits) you will need to use the xmin and xmax limits, for instance if I want to integrate between (0,1) the first dimension and between (-4, 10) the second one I would do:

from vegasflow import VegasFlow

dimensions = 2
vegas_instance = VegasFlow(dimensions, n_calls, xmin=[0, -4], xmax=[1, 10])