Issue with Identity Emissions

[YUJINHAN93]

Hi,
This is related to #88 , I think there is still an issue with identity emissions.
When I set emissions="studentst_id" or "gaussian_id", the initialization works fine, but I encounter the error below during model fitting.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-50-77c812f71179> in <module>
      1 q_elbos_lem, q_lem = rslds.fit(y, method="laplace_em",
      2                                variational_posterior="structured_meanfield",
----> 3                                initialize=False, num_iters=100, alpha=0.5)

~/.local/lib/python3.7/site-packages/ssm/util.py in wrapper(self, datas, inputs, masks, tags, **kwargs)
    107             tags = [tags]
    108 
--> 109         return f(self, datas, inputs=inputs, masks=masks, tags=tags, **kwargs)
    110 
    111     return wrapper

~/.local/lib/python3.7/site-packages/ssm/lds.py in fit(self, datas, inputs, masks, tags, method, variational_posterior, initialize, **kwargs)
    853         # Initialize the variational posterior
    854         posterior = self._make_variational_posterior(variational_posterior, datas, inputs, masks, tags, method)
--> 855         elbos = _fitting_methods[method](posterior, datas, inputs, masks, tags, learning=True, **kwargs)
    856         return elbos, posterior
    857 

~/.local/lib/python3.7/site-packages/ssm/lds.py in _fit_laplace_em(self, variational_posterior, datas, inputs, masks, tags, num_iters, num_samples, continuous_optimizer, continuous_tolerance, continuous_maxiter, emission_optimizer, emission_optimizer_maxiter, parameters_update, alpha, learning)
    730         Assume q(z) is a chain-structured discrete graphical model.
    731         """
--> 732         elbos = [self._laplace_em_elbo(variational_posterior, datas, inputs, masks, tags)]
    733         pbar = trange(num_iters)
    734         pbar.set_description("ELBO: {:.1f}".format(elbos[-1]))

~/.local/lib/python3.7/site-packages/ssm/lds.py in _laplace_em_elbo(self, variational_posterior, datas, inputs, masks, tags)
    699             log_Ps = self.transitions.log_transition_matrices(x, input, x_mask, tag)
    700             log_likes = self.dynamics.log_likelihoods(x, input, x_mask, tag)
--> 701             log_likes += self.emissions.log_likelihoods(data, input, mask, tag, x)
    702 
    703             # Compute the expected log probability

~/.local/lib/python3.7/site-packages/ssm/emissions.py in log_likelihoods(self, data, input, mask, tag, x)
    448     def log_likelihoods(self, data, input, mask, tag, x):
    449         N, etas, nus = self.N, np.exp(self.inv_etas), np.exp(self.inv_nus)
--> 450         mus = self.forward(x, input, tag)
    451 
    452         resid = data[:, None, :] - mus

TypeError: forward() takes 3 positional arguments but 4 were given

I also experience the same error as in #88 under these conditions. I'm not sure why this error keeps happening, given that it was updated in #88 . Does anyone have any ideas?

---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
<ipython-input-51-b963545280d0> in <module>
      1 q_elbos_lem, q_lem = rslds.fit(y, method="bbvi",
      2                                variational_posterior="meanfield",
----> 3                                initialize=False, num_iters=100, alpha=0.5)

~/.local/lib/python3.7/site-packages/ssm/util.py in wrapper(self, datas, inputs, masks, tags, **kwargs)
    107             tags = [tags]
    108 
--> 109         return f(self, datas, inputs=inputs, masks=masks, tags=tags, **kwargs)
    110 
    111     return wrapper

~/.local/lib/python3.7/site-packages/ssm/lds.py in fit(self, datas, inputs, masks, tags, method, variational_posterior, initialize, **kwargs)
    853         # Initialize the variational posterior
    854         posterior = self._make_variational_posterior(variational_posterior, datas, inputs, masks, tags, method)
--> 855         elbos = _fitting_methods[method](posterior, datas, inputs, masks, tags, learning=True, **kwargs)
    856         return elbos, posterior
    857 

~/.local/lib/python3.7/site-packages/ssm/lds.py in _fit_svi(self, variational_posterior, datas, inputs, masks, tags, learning, optimizer, num_iters, **kwargs)
    414         # Initialize the parameters
    415         if learning:
--> 416             params = (self.params, variational_posterior.params)
    417         else:
    418             params = variational_posterior.params

~/.local/lib/python3.7/site-packages/ssm/lds.py in params(self)
    157                self.transitions.params, \
    158                self.dynamics.params, \
--> 159                self.emissions.params
    160 
    161     @params.setter

~/.local/lib/python3.7/site-packages/ssm/emissions.py in params(self)
    434     @property
    435     def params(self):
--> 436         return super(_StudentsTEmissionsMixin, self).params + (self.inv_etas, self.inv_nus)
    437 
    438     @params.setter

~/.local/lib/python3.7/site-packages/ssm/emissions.py in params(self)
     20     @property
     21     def params(self):
---> 22         raise NotImplementedError
     23 
     24     @params.setter

NotImplementedError: