Write a function with a for loop with tensorflow.compat.v1 to implement a integral

Hi! I'm working with deepxde using tensorflow.compat.v1 as a backend and I need to approximate an integral using tensorflow but I haven't able to figure it out. As context the integral is

The difficulty here is that the variable t is the upper limit of the integral but also is in the function q. So for each t I need to create a tensor and iterate it.

This is my code so far

def tf_integral(t_tensor, a):
    dt = tf.constant(0.1)
    ta = tf.TensorArray(tf.float32, size=0, dynamic_size=True)
    ta = ta.unstack([0., 1.])
    for i in range(n_days):
        t = t_tensor[i]
        # breakpoint()
        eta_tensor = tf.range(dt, t + dt, dt)
        value = tf.reduce_sum(tf_q(t - eta_tensor, b, c) * tf.exp(eta_tensor, d)) * dx
        ta = ta.write(i, value)
    return ta.stack()

where t_tensor is a tensor of shape [None, 1] (a column), n_days is a integer since I wasn't able to get something like len(t_tensor). I'm getting the following error.

ValueError: in user code:

    File "/mnt/d/Documents/git/insight/notebooks/china project/china-v-fitted.py", line 130, in tf_integral_dv_dphi  *
        nu_tensor = tf.range(dx, t + dx, dx)

    ValueError: Shape must be rank 0 but is rank 1
         for 'limit' for ' = Range[Tidx=DT_FLOAT](Const, add, Const)' with input shapes: [], [1], [].

I'm not sure if I can use @tf.function decorator with tensorflow.compat.v1 backend since it is my first time working with tensorflow.

I've been looking for alternatives and it seems I could use tf.compat.v1.while_loop and/or tf.compat.v1.wrap_function but honestly I'm not sure.

As a reference, the code with scipy works well

def integral(t, a):
    dt = 0.1
    eta_array = np.arange(dt, t + dt, dt)
    value = np.sum(q(t - eta_array) * np.exp(a * eta_array)) * dx
    return value

Thanks in advance!

source https://stackoverflow.com/questions/75961673/write-a-function-with-a-for-loop-with-tensorflow-compat-v1-to-implement-a-integr