upgrade pyre version in fbcode/pearl - batch 1

Differential Revision: D60991999 fbshipit-source-id: 318e24f7f9b3a19b9e9fa07d94073728c83deb40
facebookresearch · Aug 9, 2024 · f84334c · f84334c
1 parent 6b0e81d
commit f84334c
Show file tree

Hide file tree

Showing 4 changed files with 9 additions and 0 deletions.
diff --git a/pearl/neural_networks/contextual_bandit/linear_regression.py b/pearl/neural_networks/contextual_bandit/linear_regression.py
@@ -125,6 +125,8 @@ def matrix_inv_fallback_pinv(self, A: torch.Tensor) -> torch.Tensor:
         try:
             return torch.linalg.inv(A).contiguous()
         # pyre-ignore[16]: Module `_C` has no attribute `_LinAlgError`.
+        # pyre-fixme[66]: Exception handler type annotation `unknown` must extend
+        #  BaseException.
         except torch._C._LinAlgError as e:
             logger.warning(
                 "Exception raised during A inversion, falling back to pseudo-inverse",

diff --git a/pearl/policy_learners/contextual_bandits/disjoint_bandit.py b/pearl/policy_learners/contextual_bandits/disjoint_bandit.py
@@ -102,6 +102,7 @@ def _partition_batch_by_arm(self, batch: TransitionBatch) -> List[TransitionBatc
                     state = batch.state[:, arm, :]
                 batches.append(
                     TransitionBatch(
+                        # pyre-fixme[61]: `state` is undefined, or not always defined.
                         state=state[mask],
                         reward=batch.reward[mask],
                         weight=(

diff --git a/pearl/utils/functional_utils/experimentation/create_offline_data.py b/pearl/utils/functional_utils/experimentation/create_offline_data.py
@@ -82,6 +82,7 @@ def create_offline_data(
             action = agent.act(exploit=False)
 
             action_result = env.step(action)
+            # pyre-fixme[58]: `+` is not supported for operand types `int` and `object`.
             g += action_result.reward
             agent.observe(action_result)
             transition_tuple = {

diff --git a/pearl/utils/functional_utils/train_and_eval/online_learning.py b/pearl/utils/functional_utils/train_and_eval/online_learning.py
@@ -267,6 +267,7 @@ def run_episode(
             action.cpu() if isinstance(action, torch.Tensor) else action
         )  # action can be int sometimes
         action_result = env.step(action)
+        # pyre-fixme[58]: `+` is not supported for operand types `int` and `object`.
         cum_reward += action_result.reward
         if (
             num_risky_sa is not None
@@ -296,8 +297,12 @@ def run_episode(
 
     info = {"return": cum_reward}
     if num_risky_sa is not None:
+        # pyre-fixme[6]: For 1st argument expected `SupportsKeysAndGetItem[str,
+        #  int]` but got `Dict[str, float]`.
         info.update({"risky_sa_ratio": num_risky_sa / episode_steps})
     if cum_cost is not None:
+        # pyre-fixme[6]: For 1st argument expected `SupportsKeysAndGetItem[str,
+        #  int]` but got `Dict[str, float]`.
         info.update({"return_cost": cum_cost})
 
     return info, episode_steps