Merge branch 'master' into dee_cfr_jax_refactor

Author: alexunderch

open_spiel/colabs/CFR_and_REINFORCE.ipynb

@@ -134,7 +134,7 @@
         "\n",
         "    # Compute regrets at this state.\n",
         "    cfr_prob = np.prod(reach[:player]) * np.prod(reach[player+1:])\n",
-        "    value = np.einsum('ap,a-\u003ep', utility, curr_policy[index])\n",
+        "    value = np.einsum('ap,a->p', utility, curr_policy[index])\n",
         "    for action in state.legal_actions():\n",
         "      regrets[index][action] += cfr_prob * (utility[action][player] - value[player])\n",
         "\n",
@@ -173,7 +173,7 @@
         "  policy.action_probability_array += curr_policy * lr\n",
         "\n",
         "  # Evaluate the average policy\n",
-        "  if step \u0026 (step-1) == 0:\n",
+        "  if step & (step-1) == 0:\n",
         "    nc = exploitability.nash_conv(game, policy)\n",
         "    eval_steps.append(step)\n",
         "    eval_nash_conv.append(nc)\n",

open_spiel/colabs/crowd_modelling_4rooms_MFGsurvey.ipynb

@@ -169,7 +169,7 @@
         "\n",
         "def decode_distribution(game: pyspiel.Game,\n",
         "                        dist: Dict[str, float],\n",
-        "                        nans: bool = True) -\u003e np.ndarray:\n",
+        "                        nans: bool = True) -> np.ndarray:\n",
         "  \"\"\"Decodes the distribution of a 2D crowd modelling game from a dictionary.\"\"\"\n",
         "  # Extract the size of the distribution from the game parameters. Time, i.e.\n",
         "  # horizon is the leading dimension so that we can easily present the temporal\n",
@@ -179,7 +179,7 @@
         "  decoded = np.zeros(dist_size)\n",
         "\n",
         "  for key, value in dist.items():\n",
-        "    m = re.fullmatch(r'\\((?P\u003cx\u003e\\d+),\\s*(?P\u003cy\u003e\\d+),\\s*(?P\u003ct\u003e\\d+)\\)', key)\n",
+        "    m = re.fullmatch(r'\\((?P<x>\\d+),\\s*(?P<y>\\d+),\\s*(?P<t>\\d+)\\)', key)\n",
         "    if m:\n",
         "      g = m.group\n",
         "      decoded[(int(g('t')), int(g('y')), int(g('x')))] = value\n",
@@ -188,14 +188,14 @@
         "\n",
         "\n",
         "def get_policy_distribution(game: pyspiel.Game,\n",
-        "                            policy: policy_std.Policy) -\u003e np.ndarray:\n",
+        "                            policy: policy_std.Policy) -> np.ndarray:\n",
         "  \"\"\"Returns the distribution of the policy.\"\"\"\n",
         "  dist_policy = distribution.DistributionPolicy(game, policy)\n",
         "  return decode_distribution(game, dist_policy.distribution)\n",
         "\n",
         "\n",
         "def animate_distributions(dists: np.ndarray,\n",
-        "                          fixed_cbar: bool = False) -\u003e animation.FuncAnimation:\n",
+        "                          fixed_cbar: bool = False) -> animation.FuncAnimation:\n",
         "  \"\"\"Animates the given distributions.\n",
         "\n",
         "  Args:\n",