gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-18 11:40:18 +01:00
Code Issues Projects Releases Wiki Activity

RL in Large Discrete Action Spaces - Wolpertinger Agent (#394)

Browse Source 
* Currently this is specific to the case of discretizing a continuous action space. Can easily be adapted to other case by feeding the kNN otherwise, and removing the usage of a discretizing output action filter
This commit is contained in:
Gal Leibovich
2019-09-08 12:53:49 +03:00
committed by GitHub
parent fc50398544
commit 138ced23ba
46 changed files with 1193 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
docs/_modules/rl_coach/exploration_policies/additive_noise.html
View File

@@ -245,7 +245,9 @@
<span class="bp">self</span><span class="o">.</span><span class="n">evaluation_noise</span> <span class="o">=</span> <span class="n">evaluation_noise</span>
<span class="bp">self</span><span class="o">.</span><span class="n">noise_as_percentage_from_action_space</span> <span class="o">=</span> <span class="n">noise_as_percentage_from_action_space</span>
<span class="k">if</span> <span class="ow">not</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="n">BoxActionSpace</span><span class="p">):</span>
<span class="k">if</span> <span class="ow">not</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="n">BoxActionSpace</span><span class="p">)</span> <span class="ow">and</span> \
<span class="p">(</span><span class="nb">hasattr</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="s1">&#39;filtered_action_space&#39;</span><span class="p">)</span> <span class="ow">and</span> <span class="ow">not</span>
<span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="o">.</span><span class="n">filtered_action_space</span><span class="p">,</span> <span class="n">BoxActionSpace</span><span class="p">)):</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">&quot;Additive noise exploration works only for continuous controls.&quot;</span>
<span class="s2">&quot;The given action space is of type: </span><span class="si">{}</span><span class="s2">&quot;</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">action_space</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span><span class="p">))</span>

5
docs/_modules/rl_coach/exploration_policies/exploration_policy.html
View File

@@ -298,7 +298,10 @@
<span class="sd">&quot;&quot;&quot;</span>
<span class="sd"> :param action_space: the action space used by the environment</span>
<span class="sd"> &quot;&quot;&quot;</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="n">BoxActionSpace</span><span class="p">)</span> <span class="ow">or</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="n">GoalsSpace</span><span class="p">)</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="n">BoxActionSpace</span><span class="p">)</span> <span class="ow">or</span> \
<span class="p">(</span><span class="nb">hasattr</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="s1">&#39;filtered_action_space&#39;</span><span class="p">)</span> <span class="ow">and</span>
<span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="o">.</span><span class="n">filtered_action_space</span><span class="p">,</span> <span class="n">BoxActionSpace</span><span class="p">))</span> <span class="ow">or</span> \
<span class="nb">isinstance</span><span class="p">(</span><span class="n">action_space</span><span class="p">,</span> <span class="n">GoalsSpace</span><span class="p">)</span>
<span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">action_space</span><span class="p">)</span>
</pre></div>

3
docs/_modules/rl_coach/exploration_policies/truncated_normal.html
View File

@@ -271,9 +271,6 @@
<span class="k">else</span><span class="p">:</span>
<span class="n">action_values_std</span> <span class="o">=</span> <span class="n">current_noise</span>
<span class="c1"># scale the noise to the action space range</span>
<span class="n">action_values_std</span> <span class="o">=</span> <span class="n">current_noise</span> <span class="o">*</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">action_space</span><span class="o">.</span><span class="n">high</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">action_space</span><span class="o">.</span><span class="n">low</span><span class="p">)</span>
<span class="c1"># extract the mean values</span>
<span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">action_values</span><span class="p">,</span> <span class="nb">list</span><span class="p">):</span>
<span class="c1"># the action values are expected to be a list with the action mean and optionally the action stdev</span>