# FloatingPointError: NaN occurred in optimization in HMM

**URL:** <https://discourse.pymc.io/t/floatingpointerror-nan-occurred-in-optimization-in-hmm/8331>\
**Category:** Questions\
**Created:** [November 17, 2021, 2:00pm UTC](https://discourse.pymc.io/t/floatingpointerror-nan-occurred-in-optimization-in-hmm/8331 "2021-11-17T14:00:05Z")\
**Posts on this page:** 1\
**Showing post:** 5

<div class="post-metadata">

**Author:** ![NickN](https://avatars.discourse-cdn.com/v4/letter/n/ed8c4c/32.png) [@NickN](https://discourse.pymc.io/u/NickN)\
**Post date:** [November 18, 2021, 9:41am UTC](https://discourse.pymc.io/t/floatingpointerror-nan-occurred-in-optimization-in-hmm/8331/5 "2021-11-18T09:41:24Z")

</div>

Thanks Junpenglao,

when I use the pm.sample() default function (like in the previous post: [[Hidden Markov Model - Estimating Transition and Emission CPDs from multiple sequences - not working](https://discourse.pymc.io/t/hidden-markov-model-estimating-transition-and-emission-cpds-from-multiple-sequences-not-working/6308)])  
I obtain the following error:

Auto-assigning NUTS sampler… Initializing NUTS using jitter+adapt\_diag… Sequential sampling (2 chains in 1 job) NUTS: [B, A, P\_transition]

Interrupted

* * *

SamplingError Traceback (most recent call last)

in () 33 34 #trace = pm.sample(3, init=‘advi+adapt\_diag’)# tune=1000 —\> 35 trace = pm.sample(3)

7 frames

[/usr/local/lib/python3.7/dist-packages/pymc3/step\_methods/hmc/base\_hmc.py](https://datifb0aq2w-496ff2e9c6d22116-0-colab.googleusercontent.com/outputframe.html?vrz=colab-20211116-060056-RC00_410210951#) in astep(self, q0) 157 ) 158 self.\_warnings.append(warning) → 159 raise SamplingError(“Bad initial energy”) 160 161 adapt\_step = self.tune and self.adapt\_step\_size

SamplingError: Bad initial energy

---

_[View the full topic](https://discourse.pymc.io/t/floatingpointerror-nan-occurred-in-optimization-in-hmm/8331)._