Dear Pyro Experts,

I am new to Pyro and probabilistic programming in general. I am trying to fit a multivariate regression with 40 covariates and 82 response variables. My simple model is as follows:

```
def model(x, y=None):
weight = pyro.sample("weight", dist.Normal(torch.zeros(40,82),torch.ones(40,82)).to_event(2))
bias = pyro.sample("bias", dist.Normal(torch.zeros(82,),torch.ones(82,)))
mu = torch.matmul(x,weight) + bias
with pyro.plate("data", x.shape[0]):
obs = pyro.sample("obs", dist.MultivariateNormal(loc=mu, scale_tril=torch.eye(82)), obs=y)
guide = AutoMultivariateNormal(model)
adam = pyro.optim.Adam({"lr": 0.0001})
svi = SVI(model, guide, adam, loss=Trace_ELBO())
for j in range(40000):
loss = svi.step(guideMatrix_trainTensor, expressionMatrix_trainTensor)
```

However my model does not converge, loss values fluctuate back and forth. When I do a simple regression with the same matrices with one response variable at a time, I get the results with no problem. May I ask what may be going wrong in here? Also is there a way to define priors over the multivariate covarince matrix?

Thanks in advance.