Comments (6)
Yeah, it is a limitation of cond
. JAX requires two branches to have the same pytree structure. I don't have a good solution for it. Depending on the algorithm, something like this (suitable for MCMC) might be required
def lazy_branch():
amount_slept = numpyro.sample("amount_slept", dist.Uniform(-100., 100.))
numpyro.factor("amount_slept_factor", dist.Normal(loc=8+2*ignore_alarm, scale=1).log_prob(amount_slept))
return amount_slept
Similar for the other branch
numpyro.factor("amount_slept_factor", 0.)
from numpyro.
I think you can move ignore_alarm
outside of the lazy branch so that two branches have the same latent variables.
from numpyro.
That works in this particular case since both branches sample from a normal distribution but not in general. For example, what if one branch has a normal distribution and one has a uniform distribution? This throws the same error. However it is possible to do in Pyro
def sleep_model(data):
feeling_lazy = numpyro.sample("feeling_lazy", dist.Bernoulli(probs=0.9))
ignore_alarm = numpyro.sample("ignore_alarm", dist.Bernoulli(probs=0.8))
def lazy_branch():
amount_slept = numpyro.sample("amount_slept", dist.Normal(loc=8+2*ignore_alarm, scale=1))
return amount_slept
def not_lazy_branch():
# amount_slept numpyro.sample("amount_slept", dist.Normal(loc=6, scale=1)) # this works
amount_slept = numpyro.sample("amount_slept", dist.Uniform(low=6., high=10.)) # this doesn't
return amount_slept
amount_slept = cond(feeling_lazy == 1,
lambda x: lazy_branch(),
lambda x: not_lazy_branch(),
data
)
return amount_slept
This works using Pyro.
def sleep_model():
feeling_lazy = pyro.sample("feeling_lazy", dist.Bernoulli(0.9))
if feeling_lazy:
ignore_alarm = pyro.sample("ignore_alarm", dist.Bernoulli(0.8))
amount_slept = pyro.sample("amount_slept",
dist.Normal(8 + 2 * ignore_alarm, 1))
else:
amount_slept = pyro.sample("amount_slept", dist.Uniform(1, 6))
return amount_slept
from numpyro.
Unfortunate, but thank you.
from numpyro.
Just curious on what is the issue that you got?
from numpyro.
Related Issues (20)
- Using deprecated `jax.core.safe_map` in `ops/provenance.py` HOT 7
- Normalizing Flow MCMC HOT 4
- `AutoNormal`, `AutoDelta`, and `AutoGuideList` do not support subsamples of variable size. HOT 2
- Speeding up NUTS and MCMC in tests HOT 3
- random_flax_module broken HOT 4
- Can't pickle MCMC object (MixedHMC kernel) when `chain_method="parallel"` HOT 3
- Adding HMCECS proxy functions HOT 2
- AttributeError: Can't pickle local object 'ESS.DifferentialMove.<locals>.make_differential_move.<locals>.differential_move' HOT 2
- [FR] MaskTransform HOT 2
- tracer error in blocked AutoGuide HOT 12
- Custom sampling per site for non-HMC approaches HOT 3
- `TransformedDistribution` support too broad when using `AffineTransform` transformation? HOT 2
- Factor Analysis/PPCA Tutorial HOT 1
- Censoring Example HOT 1
- numpyro.deterministic static on infer.Predictive HOT 13
- [FR] Support for different supports in component distributions for mixture models HOT 5
- ImportError: cannot import name 'CAR' from "numpyro.distrubutions.continuous' HOT 2
- Use biased autocorrelation estimate by default HOT 1
- mean_accept_prob significantly different after warmup HOT 8
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from numpyro.