Pyro: Trace_dist must be trace posterior distribution object when i ran code in tutorial.

[lwgkzl]

Description

I'm having problems with the code I'm running in tutorial:
https://bookdown.org/robertness/causalml/docs/tutorial-on-deep-probabilitic-modeling-with-pyro.html
codes:

error:

thanks for any help!

Environment

For any bugs, please provide the following:

• OS and python version. windows 10, Python 3.8.3
• PyTorch version 1.6.0
• Pyro version: 1.5.0

[fehiepsi]

@lwgkzl The API of MCMC has changed since that tutorial is written. To get posterior samples, you can use

mcmc = MCMC(...)
mcmc.run(...)
posterior = mcmc.get_samples()

[lwgkzl]

@lwgkzl The API of MCMC has changed since that tutorial is written. To get posterior samples, you can use

mcmc = MCMC(...)
mcmc.run(...)
posterior = mcmc.get_samples()

Many thanks : )

[lwgkzl]

@lwgkzl The API of MCMC has changed since that tutorial is written. To get posterior samples, you can use

mcmc = MCMC(...)
mcmc.run(...)
posterior = mcmc.get_samples()

After this line of code is run：
samples_posterior = mcmc.get_samples()
The type of variable samples_posterior is still not TracePosterior, so when i run the code:
marginal = EmpiricalMarginal(samples_posterior , "weight")
The error still exists, did I not understand what you meant?

[fehiepsi]

Hi @lwgkzl, you don't need to use EmpiricalMarginal to plot histogram. You can plot samples in samples_posterior directly.

[lwgkzl]

Hi @lwgkzl, you don't need to use EmpiricalMarginal to plot histogram. You can plot samples in samples_posterior directly.

ありがとうございました