BUG: AttributeError: '_FutureWarningValidatingScratchpad' object has no attribute 'test_value'

[RichardScottOZ]

Describe the issue:

New to trying out pymc3

trying this:

Reproduceable code example:

import pymc as pm
with pm.Model() as model:
    # Define the model variables
    alpha = pm.Normal('alpha', 0, 1)
    beta = pm.Normal('beta', 0, 1)
    sigma = pm.Normal('sigma', 0, 20)
    # Define the model equation
    y_pred = np.zeros(len(returns))
    y_pred[0] = returns[0]
    for i in range(1, len(returns)):
        y_pred[i] = sigma.tag.test_value * (1-alpha.tag.test_value-beta.tag.test_value) + alpha.tag.test_value * returns[i-1]**2 + beta.tag.test_value * y_pred[i-1]
    garch_eq = pm.Deterministic('garch_eq', y_pred)
    # Specify the likelihood function
    likelihood = pm.Normal('y', mu=garch_eq, tau=1.0/sigma.tag.test_value**2, observed=returns)
    # Run the model
    mcmc = pm.MCMC(model)
    mcmc.sample(iter=50000, burn=20000)

Error message:

Traceback (most recent call last)
Cell In[127], line 10
      8 y_pred[0] = returns[0]
      9 for i in range(1, len(returns)):
---> 10     y_pred[i] = sigma.tag.test_value * (1-alpha.tag.test_value-beta.tag.test_value) + alpha.tag.test_value * returns[i-1]**2 + beta.tag.test_value * y_pred[i-1]
     11 garch_eq = pm.Deterministic('garch_eq', y_pred)
     12 # Specify the likelihood function

File ~\AppData\Local\Continuum\anaconda3\envs\copper\lib\site-packages\pymc\util.py:505, in _FutureWarningValidatingScratchpad.__getattribute__(self, name)
    500         else:
    501             warnings.warn(
    502                 f"The tag attribute {name} is deprecated. Use {alternative} instead",
    503                 FutureWarning,
    504             )
--> 505 return super().__getattribute__(name)

File ~\AppData\Local\Continuum\anaconda3\envs\copper\lib\site-packages\pytensor\graph\utils.py:284, in Scratchpad.__getattribute__(self, name)
    283 def __getattribute__(self, name):
--> 284     return super().__getattribute__(name)

PyMC version information:

pymc 5.0.1 hd8ed1ab_0 conda-forge
pymc-base 5.0.1 pyhd8ed1ab_0 conda-forge
pytensor 2.8.11 py38hd636df3_1 conda-forge
pytensor-base 2.8.11 py38haa244fe_1 conda-forge
python 3.8.15 h4de0772_0_cpython conda-forge

Windows 10
Conda

Context for the issue:

just a test: test_vallue seems to get exist

[michaelosthege]

Your example looks like pymc2 which is severely outdated.
Make sure to check the documentation at https://pymc.io

[RichardScottOZ]

Ok - said link currently not working, FYI.

[RichardScottOZ]

Right _ had looked at the 5.01 docs Friday - there's apparently a GARCH11 built-in, but didn't find any examples of such - are there any you are aware of?

[ricardoV94]

Acting on tags won't have any effect on PyMC model.

For the GARCH11, maybe the tests provide an example?

pymc/tests/distributions/test_timeseries.py

Lines 724 to 772 in 457421b

	def test_logp(self):
	# test data ~ N(0, 1)
	data = np.array(
	[
	-1.35078362,
	-0.81254164,
	0.28918551,
	-2.87043544,
	-0.94353337,
	0.83660719,
	-0.23336562,
	-0.58586298,
	-1.36856736,
	-1.60832975,
	-1.31403141,
	0.05446936,
	-0.97213128,
	-0.18928725,
	1.62011258,
	-0.95978616,
	-2.06536047,
	0.6556103,
	-0.27816645,
	-1.26413397,
	]
	)
	omega = 0.6
	alpha_1 = 0.4
	beta_1 = 0.5
	initial_vol = np.float64(0.9)
	vol = np.empty_like(data)
	vol[0] = initial_vol
	for i in range(len(data) - 1):
	vol[i + 1] = np.sqrt(omega + beta_1 * vol[i] ** 2 + alpha_1 * data[i] ** 2)
	with Model() as t:
	y = GARCH11(
	"y",
	omega=omega,
	alpha_1=alpha_1,
	beta_1=beta_1,
	initial_vol=initial_vol,
	shape=data.shape,
	)
	z = Normal("z", mu=0, sigma=vol, shape=data.shape)
	garch_like = t.compile_logp(y)({"y": data})
	reg_like = t.compile_logp(z)({"z": data})
	decimal = select_by_precision(float64=7, float32=4)
	np.testing.assert_allclose(garch_like, reg_like, 10 ** (-decimal))

Seems to be doing something similar to what you were trying? Would be good to add a docstring example though