Fix typos

python/ray/_private/accelerators/amd_gpu.py

@@ -53,7 +53,7 @@ def get_visible_accelerator_ids_env_var() -> str:
                 env_var = CUDA_VISIBLE_DEVICES_ENV_VAR
             elif hip_val != cuda_val:
                 raise ValueError(
-                    f"Inconsistant values found. Please use either {HIP_VISIBLE_DEVICES_ENV_VAR} or {CUDA_VISIBLE_DEVICES_ENV_VAR}."
+                    f"Inconsistent values found. Please use either {HIP_VISIBLE_DEVICES_ENV_VAR} or {CUDA_VISIBLE_DEVICES_ENV_VAR}."
                 )
 
         return env_var

python/ray/autoscaler/_private/resource_demand_scheduler.py

@@ -648,7 +648,7 @@ def _add_min_workers_nodes(
     request_resources() constraints.
 
     Args:
-        node_resources: Resources of exisiting nodes already launched/pending.
+        node_resources: Resources of existing nodes already launched/pending.
         node_type_counts: Counts of existing nodes already launched/pending.
         node_types: Node types config.
         max_workers: global max_workers constaint.

python/ray/dashboard/modules/reporter/profile_manager.py

@@ -287,7 +287,7 @@ async def attach_profiler(
 
         Returns:
             Tuple[bool, str]: A tuple containing a boolean indicating the success
-                of the operation and a string of a sucess message or an error message.
+                of the operation and a string of a success message or an error message.
         """
         memray = shutil.which(self.profiler_name)
         if memray is None:

python/ray/data/_internal/cluster_autoscaler/default_cluster_autoscaler.py

@@ -36,7 +36,7 @@ def __init__(
         self._last_request_time = 0
 
     def try_trigger_scaling(self):
-        """Try to scale up the cluster to accomodate the provided in-progress workload.
+        """Try to scale up the cluster to accommodate the provided in-progress workload.
 
         This makes a resource request to Ray's autoscaler consisting of the current,
         aggregate usage of all operators in the DAG + the incremental usage of all

python/ray/experimental/tqdm_ray.py

@@ -251,7 +251,7 @@ def close_bar(self, state: ProgressBarState) -> None:
         instance().unhide_bars()
 
     def slots_required(self):
-        """Return the number of pos slots we need to accomodate bars in this group."""
+        """Return the number of pos slots we need to accommodate bars in this group."""
         if not self.bars_by_uuid:
             return 0
         return 1 + max(bar.state["pos"] for bar in self.bars_by_uuid.values())

python/ray/tune/search/variant_generator.py

@@ -344,7 +344,7 @@ def assign_value(spec: Dict, path: Tuple, value: Any):
     """Assigns a value to a nested dictionary.
 
     Handles the special case of tuples, in which case the tuples
-    will be re-constructed to accomodate the updated value.
+    will be re-constructed to accommodate the updated value.
     """
     parent_spec = None
     parent_key = None

python/ray/tune/utils/file_transfer.py

@@ -383,7 +383,7 @@ def _unpack_dir(stream: io.BytesIO, target_dir: str, *, _retry: bool = True) ->
     target_dir = os.path.normpath(target_dir)
     try:
         # Timeout 0 means there will be only one attempt to acquire
-        # the file lock. If it cannot be aquired, a TimeoutError
+        # the file lock. If it cannot be acquired, a TimeoutError
         # will be thrown.
         with TempFileLock(f"{target_dir}.lock", timeout=0):
             with tarfile.open(fileobj=stream) as tar:
@@ -426,7 +426,7 @@ def _copy_dir(
     target_dir = os.path.normpath(target_dir)
     try:
         # Timeout 0 means there will be only one attempt to acquire
-        # the file lock. If it cannot be aquired, a TimeoutError
+        # the file lock. If it cannot be acquired, a TimeoutError
         # will be thrown.
         with TempFileLock(f"{target_dir}.lock", timeout=0):
             _delete_path_unsafe(target_dir)

release/tune_tests/fault_tolerance_tests/workloads/terminate_node_aws.py

@@ -102,15 +102,15 @@ def kill(self):
         failures = 0
         max_failures = 3
         node = None
-        terminated_succesfully = False
-        while not terminated_succesfully and failures < max_failures:
+        terminated_successfully = False
+        while not terminated_successfully and failures < max_failures:
             try:
                 node = get_random_node()
                 if not node:
                     logger.info("No alive worker nodes")
                     continue
                 terminate_node(node["NodeID"])
-                terminated_succesfully = True
+                terminated_successfully = True
                 logger.info(
                     f"Killed node {node['NodeID']} with IP {node['NodeManagerAddress']}"
                 )
@@ -125,7 +125,7 @@ def kill(self):
             {
                 "timestamp": time.time(),
                 "node": node,
-                "terminated_succesfully": terminated_succesfully,
+                "terminated_successfully": terminated_successfully,
             }
         )
         # safe_write_to_results_json(self.history)

release/tune_tests/fault_tolerance_tests/workloads/test_tune_worker_fault_tolerance.py

@@ -10,7 +10,7 @@
 
 If a trial is restored, it should restart from the last checkpointed iteration.
 
-The test is succesfull if all trials finish with the expected number of iterations,
+The test is successful if all trials finish with the expected number of iterations,
 and that a checkpoint is always available when restoring.
 
 This test only works on AWS as it uses AWS CLI to terminate nodes.

rllib/core/rl_module/tests/test_rl_module_specs.py

@@ -88,7 +88,7 @@ def test_customized_multi_agent_module(self):
         spec.build()
 
     def test_get_spec_from_module_multi_agent(self):
-        """Tests wether MultiRLModuleSpec.from_module() works."""
+        """Tests whether MultiRLModuleSpec.from_module() works."""
         env = gym.make("CartPole-v1")
         num_agents = 2
         module_specs = {}
@@ -107,7 +107,7 @@ def test_get_spec_from_module_multi_agent(self):
         self.assertEqual(spec, spec_from_module)
 
     def test_get_spec_from_module_single_agent(self):
-        """Tests wether RLModuleSpec.from_module() works."""
+        """Tests whether RLModuleSpec.from_module() works."""
         env = gym.make("CartPole-v1")
         spec = RLModuleSpec(
             module_class=VPGTorchRLModule,
@@ -121,7 +121,7 @@ def test_get_spec_from_module_single_agent(self):
         self.assertEqual(spec, spec_from_module)
 
     def test_update_specs(self):
-        """Tests wether RLModuleSpec.update() works."""
+        """Tests whether RLModuleSpec.update() works."""
         env = gym.make("CartPole-v0")
 
         # Test if RLModuleSpec.update() works.

rllib/offline/estimators/tests/test_ope_math.py

@@ -51,7 +51,7 @@ def action_distribution_fn(self, model, obs_batch=None, **kwargs):
 
         # add 0.5 to the action that gave a good reward (2) and subtract 0.5 from the
         # action that gave a bad reward (1)
-        # to acheive this I can just subtract 1.5 from old_reward
+        # to achieve this I can just subtract 1.5 from old_reward
         delta = old_rewards - 1.5
         if not self.improved:
             # reverse the logic for a worse policy

src/ray/common/cgroup2/sysfs_cgroup_driver.h

@@ -168,7 +168,7 @@ class SysFsCgroupDriver : public CgroupDriverInterface {
 
   /**
     Reads the cgroup.procs of "from" and writes them out to the given file.
-    The cgroup.procs file is newline seperated. The current user must have
+    The cgroup.procs file is newline separated. The current user must have
     read-write permissions to both cgroup.procs file as well as the common ancestor
     of the source and destination cgroups.
 

src/ray/gcs/gcs_placement_group_scheduler.h

@@ -467,7 +467,7 @@ class GcsPlacementGroupScheduler : public GcsPlacementGroupSchedulerInterface {
   ///
   /// \param bundle The node to which the bundle is scheduled and the bundle's
   /// specification.
-  /// \return True if the bundle is succesfully released. False otherwise.
+  /// \return True if the bundle is successfully released. False otherwise.
   bool TryReleasingBundleResources(
       const std::pair<NodeID, std::shared_ptr<const BundleSpecification>> &bundle);
 

src/ray/gcs/gcs_worker_manager.h

@@ -81,10 +81,10 @@ class GcsWorkerManager : public rpc::WorkerInfoGcsServiceHandler {
   std::vector<std::function<void(std::shared_ptr<rpc::WorkerTableData>)>>
       worker_dead_listeners_;
 
-  /// Tracks the number of occurences of worker crash due to system error
+  /// Tracks the number of occurrences of worker crash due to system error
   int32_t worker_crash_system_error_count_ = 0;
 
-  /// Tracks the number of occurences of worker crash due to OOM
+  /// Tracks the number of occurrences of worker crash due to OOM
   int32_t worker_crash_oom_count_ = 0;
 
   /// Ray metrics

src/ray/raylet/local_object_manager.cc

@@ -355,7 +355,7 @@ void LocalObjectManager::SpillObjectsInternal(
           io_worker_pool_.PushSpillWorker(io_worker);
           size_t num_objects_spilled = status.ok() ? r.spilled_objects_url_size() : 0;
           // Object spilling is always done in the order of the request.
-          // For example, if an object succeeded, it'll guarentee that all objects
+          // For example, if an object succeeded, it'll guarantee that all objects
           // before this will succeed.
           RAY_CHECK(num_objects_spilled <= requested_objects_to_spill.size());
           for (size_t i = num_objects_spilled; i != requested_objects_to_spill.size();