Bug fix of Evaluator (Alex's fault)

gflownet/evaluator/base.py

@@ -557,7 +557,7 @@ def plot(self, x_sampled, kde_pred, kde_true, plot_kwargs, **kwargs):
             values are the figures.
         """
 
-        fig_kde_pred = fig_kde_true = fig_reward_samples = None
+        fig_kde_pred = fig_kde_true = fig_reward_samples = fig_samples_topk = None
 
         if hasattr(self.gfn.env, "plot_reward_samples") and x_sampled is not None:
             (sample_space_batch, rewards_sample_space) = (

tests/gflownet/envs/test_tetris.py

@@ -547,6 +547,7 @@ def setup(self, env):
         self.env = env
         self.repeats = {
             "test__reset__state_is_source": 10,
+            "test__gflownet_minimal_runs": 0,
         }
         self.n_states = {}  # TODO: Populate.
 
@@ -559,6 +560,7 @@ def setup(self, env_full):
         self.env = env_full
         self.repeats = {
             "test__reset__state_is_source": 10,
+            "test__gflownet_minimal_runs": 0,
         }
         self.n_states = {}  # TODO: Populate.
 

tests/gflownet/evaluator/test_base.py

@@ -237,6 +237,9 @@ def test__eval(gflownet_for_tests, parameterization):
     elif parameterization == "ctorus":
         for figname, fig in figs.items():
             assert isinstance(figname, str)
+            # plot_samples_topk not implemented in ctorus
+            if figname == "Samples TopK":
+                continue
             assert isinstance(fig, plt.Figure)
     else:
         raise ValueError(f"Unknown parameterization: {parameterization}")