Reduce Ray / RLlib startup messages (#5368)

python/ray/node.py

@@ -493,7 +493,7 @@ class Node(object):
 
     def start_head_processes(self):
         """Start head processes on the node."""
-        logger.info(
+        logger.debug(
             "Process STDOUT and STDERR is being redirected to {}.".format(
                 self._logs_dir))
         assert self._redis_address is None
@@ -507,7 +507,7 @@ class Node(object):
 
     def start_ray_processes(self):
         """Start all of the processes on the node."""
-        logger.info(
+        logger.debug(
             "Process STDOUT and STDERR is being redirected to {}.".format(
                 self._logs_dir))
 

python/ray/rllib/agents/ppo/ppo.py

@@ -142,10 +142,6 @@ def validate_config(config):
         logger.warning(
             "Using the simple non-minibatch optimizer. This will greatly "
             "reduce performance, consider simple_optimizer=False.")
-    if not config["vf_share_layers"]:
-        logger.warning(
-            "FYI: By default, the value function will not share layers "
-            "with the policy model ('vf_share_layers': False).")
 
 
 PPOTrainer = build_trainer(

python/ray/rllib/evaluation/rollout_worker.py

@@ -317,9 +317,9 @@ class RolloutWorker(EvaluatorInterface):
             if (ray.is_initialized()
                     and ray.worker._mode() != ray.worker.LOCAL_MODE
                     and not ray.get_gpu_ids()):
-                logger.info("Creating policy evaluation worker {}".format(
+                logger.debug("Creating policy evaluation worker {}".format(
                     worker_index) +
-                            " on CPU (please ignore any CUDA init errors)")
+                             " on CPU (please ignore any CUDA init errors)")
             if not tf:
                 raise ImportError("Could not import tensorflow")
             with tf.Graph().as_default():

python/ray/rllib/policy/dynamic_tf_policy.py

@@ -321,7 +321,7 @@ class DynamicTFPolicy(TFPolicy):
             batch_tensors[k] = placeholder
 
         if log_once("loss_init"):
-            logger.info(
+            logger.debug(
                 "Initializing loss function with dummy input:\n\n{}\n".format(
                     summarize(batch_tensors)))
 

python/ray/rllib/utils/__init__.py

@@ -65,7 +65,10 @@ def try_import_tf():
         return None
 
     try:
+        if "TF_CPP_MIN_LOG_LEVEL" not in os.environ:
+            os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
         import tensorflow.compat.v1 as tf
+        tf.logging.set_verbosity(tf.logging.ERROR)
         tf.disable_v2_behavior()
         return tf
     except ImportError:

python/ray/services.py

@@ -404,7 +404,7 @@ def wait_for_redis_to_start(redis_ip_address,
     while counter < num_retries:
         try:
             # Run some random command and see if it worked.
-            logger.info(
+            logger.debug(
                 "Waiting for redis server at {}:{} to respond...".format(
                     redis_ip_address, redis_port))
             redis_client.client_list()
@@ -805,7 +805,7 @@ def _start_redis_instance(executable,
         redis_client.config_set("maxmemory", str(redis_max_memory))
         redis_client.config_set("maxmemory-policy", "allkeys-lru")
         redis_client.config_set("maxmemory-samples", "10")
-        logger.info("Starting Redis shard with {} GB max memory.".format(
+        logger.debug("Starting Redis shard with {} GB max memory.".format(
             round(redis_max_memory / 1e9, 2)))
 
     # If redis_max_clients is provided, attempt to raise the number of maximum
@@ -1470,9 +1470,9 @@ def start_plasma_store(stdout_file=None,
 
     # Print the object store memory using two decimal places.
     object_store_memory_str = (object_store_memory / 10**7) / 10**2
-    logger.info("Starting the Plasma object store with {} GB memory "
-                "using {}.".format(
-                    round(object_store_memory_str, 2), plasma_directory))
+    logger.debug("Starting the Plasma object store with {} GB memory "
+                 "using {}.".format(
+                     round(object_store_memory_str, 2), plasma_directory))
     # Start the Plasma store.
     process_info = _start_plasma_store(
         object_store_memory,

python/ray/signature.py

@@ -105,7 +105,7 @@ def check_signature_supported(func, warn=False):
         message = ("The function {} has a **kwargs argument, which is "
                    "currently not supported.".format(function_name))
         if warn:
-            logger.warning(message)
+            logger.debug(message)
         else:
             raise Exception(message)
 
@@ -114,7 +114,7 @@ def check_signature_supported(func, warn=False):
                    "(defined after * or *args), which is currently "
                    "not supported.".format(function_name))
         if warn:
-            logger.warning(message)
+            logger.debug(message)
         else:
             raise Exception(message)
 

python/ray/tune/log_sync.py

@@ -31,7 +31,7 @@ def log_sync_template():
     ssh_key = get_ssh_key()
     if ssh_key is None:
         if not _log_sync_warned:
-            logger.error("Log sync requires cluster to be setup with "
+            logger.debug("Log sync requires cluster to be setup with "
                          "`ray up`.")
             _log_sync_warned = True
         return

python/ray/tune/trial_runner.py

@@ -173,7 +173,7 @@ class TrialRunner(object):
                 logger.exception(
                     "Runner restore failed. Restarting experiment.")
         else:
-            logger.info("Starting a new experiment.")
+            logger.debug("Starting a new experiment.")
 
         self._start_time = time.time()
         self._last_checkpoint_time = -float("inf")

python/ray/tune/util.py

@@ -29,6 +29,7 @@ except ImportError:
 
 _pinned_objects = []
 PINNED_OBJECT_PREFIX = "ray.tune.PinnedObject:"
+START_OF_TIME = time.time()
 
 
 class UtilMonitor(Thread):
@@ -139,7 +140,7 @@ class warn_if_slow(object):
 
     def __exit__(self, type, value, traceback):
         now = time.time()
-        if now - self.start > 0.1:
+        if now - self.start > 0.1 and now - START_OF_TIME > 60.0:
             logger.warning("The `{}` operation took {} seconds to complete, ".
                            format(self.name, now - self.start) +
                            "which may be a performance bottleneck.")