Update examples #2254
Update examples #2254
Conversation
junpenglao
commented
Jun 1, 2017
junpenglao
commented
- use new sampling api (advi init, etc)
- unify style
* use new sampling api (advi init, etc) * unify style
|
Following the discussion in #2236 |
pymc3/examples/baseball.py
Outdated
| data = np.loadtxt(pm.get_data('efron-morris-75-data.tsv'), delimiter="\t", | ||
| skiprows=1, usecols=(2,3)) | ||
|
|
||
| atBats = pm.floatX(data[:,0]) |
| with model: | ||
| # initialize NUTS() with ADVI under the hood | ||
| trace = pm.sample( n ) | ||
| trace = pm.sample(n, nuts_kwargs={'target_accept':.99}) |
There was a problem hiding this comment.
Wow, does it require this to converge?
There was a problem hiding this comment.
it's fine without, but there is a divergent warning.
* add example into pm.Bound warning, also improve pm.Bound docstring
|
Remain to be done:
|
|
Trying to fix |
|
Before Merging: @springcoil could you please test it in GPU again #2033? |
pymc3/distributions/distribution.py
Outdated
| @@ -476,8 +476,14 @@ class Bound(object): | |||
|
|
|||
| Example | |||
| ------- | |||
| boundedNormal = pymc3.Bound(pymc3.Normal, lower=0.0) | |||
| par = boundedNormal(mu=0.0, sd=1.0, testval=1.0) | |||
| # In general Bounded distribution need to be defined befor model context | |||
There was a problem hiding this comment.
@fonnesbeck is this accurate?
There was a problem hiding this comment.
I've defined them inside a context before. I don't think it matters (it shouldn't).
Also, before has an "e" at the end.
depends on #2277 by @aseyboldt
|
I'll have a look this weekend maybe (swamped with work before then). Looks good work though I'll check it on a GPU though. |
