Inferring parameters of SDEs using a Euler-Maruyama scheme

Questions
1 
with pm.Model() as model:
    τh = pm.Uniform('τh', lower=0.1, upper=5.0)
    ah = pm.Uniform('ah', lower=0.5, upper=1.5)
    mh = pm.Uniform('mh', lower=0.0, upper=1.0)
    xyh = EulerMaruyama('xyh', dt, osc_sde, (τh, ah), shape=xys.shape, testval=xys)
    zh = pm.Normal('zh', mu=mh * xyh[:, 0] + (1 - mh) * xyh[:, 1], sigma=0.1, observed=zs)
with model:
    trace = pm.sample(2000, tune=1000)

def osc_sde(xy, τ, a):
    x, y = xy[:, 0], xy[:, 1]
    dx = τ * (x - x**3.0/3.0 + y)
    dy = (1.0 / τ) * (a - x)
    dxy = tt.stack([dx, dy], axis=0).T
    return dxy, σ2

Hi,
I am very new to PyMC, I want to implement above code with my own osc_sde function, That i want to replace with some model like LSTM , but I am unable to get value of T and a in desire datatype as it has datatype pymc3.model.TransformedRV, Suggest someway to proceed further