From 155cc81e404cc674bb94c4437e2041163184e5e7 Mon Sep 17 00:00:00 2001
From: Piotr Januszewski <piojanuszewski@gmail.com>
Date: Mon, 6 Jul 2020 05:07:27 +0200
Subject: [PATCH] Clarify training intensity configuration docstring (#9244)
 (#9306)

---
 rllib/agents/ddpg/ddpg.py      | 7 ++++---
 rllib/agents/dqn/apex.py       | 5 +++--
 rllib/agents/dqn/dqn.py        | 7 ++++---
 rllib/agents/sac/sac.py        | 7 ++++---
 rllib/contrib/maddpg/maddpg.py | 7 ++++---
 5 files changed, 19 insertions(+), 14 deletions(-)

diff --git a/rllib/agents/ddpg/ddpg.py b/rllib/agents/ddpg/ddpg.py
index a61e7778b..9d6e0b889 100644
--- a/rllib/agents/ddpg/ddpg.py
+++ b/rllib/agents/ddpg/ddpg.py
@@ -106,9 +106,10 @@ DEFAULT_CONFIG = with_common_config({
     "prioritized_replay_eps": 1e-6,
     # Whether to LZ4 compress observations
     "compress_observations": False,
-    # If set, this will fix the ratio of sampled to replayed timesteps.
-    # Otherwise, replay will proceed at the native ratio determined by
-    # (train_batch_size / rollout_fragment_length).
+    # If set, this will fix the ratio of replayed from a buffer and learned on
+    # timesteps to sampled from an environment and stored in the replay buffer
+    # timesteps. Otherwise, the replay will proceed at the native ratio
+    # determined by (train_batch_size / rollout_fragment_length).
     "training_intensity": None,
 
     # === Optimization ===
diff --git a/rllib/agents/dqn/apex.py b/rllib/agents/dqn/apex.py
index 6dfee4766..eb1bb57c6 100644
--- a/rllib/agents/dqn/apex.py
+++ b/rllib/agents/dqn/apex.py
@@ -41,8 +41,9 @@ APEX_DEFAULT_CONFIG = merge_dicts(
         "exploration_config": {"type": "PerWorkerEpsilonGreedy"},
         "worker_side_prioritization": True,
         "min_iter_time_s": 30,
-        # If set, this will fix the ratio of sampled to replayed timesteps.
-        # Otherwise, replay will proceed as fast as possible.
+        # If set, this will fix the ratio of replayed from a buffer and learned
+        # on timesteps to sampled from an environment and stored in the replay
+        # buffer timesteps. Otherwise, replay will proceed as fast as possible.
         "training_intensity": None,
     },
 )
diff --git a/rllib/agents/dqn/dqn.py b/rllib/agents/dqn/dqn.py
index e08f0c169..e94b4e2b6 100644
--- a/rllib/agents/dqn/dqn.py
+++ b/rllib/agents/dqn/dqn.py
@@ -85,9 +85,10 @@ DEFAULT_CONFIG = with_common_config({
     "compress_observations": False,
     # Callback to run before learning on a multi-agent batch of experiences.
     "before_learn_on_batch": None,
-    # If set, this will fix the ratio of sampled to replayed timesteps.
-    # Otherwise, replay will proceed at the native ratio determined by
-    # (train_batch_size / rollout_fragment_length).
+    # If set, this will fix the ratio of replayed from a buffer and learned on
+    # timesteps to sampled from an environment and stored in the replay buffer
+    # timesteps. Otherwise, the replay will proceed at the native ratio
+    # determined by (train_batch_size / rollout_fragment_length).
     "training_intensity": None,
 
     # === Optimization ===
diff --git a/rllib/agents/sac/sac.py b/rllib/agents/sac/sac.py
index ca6b0937d..43c33c639 100644
--- a/rllib/agents/sac/sac.py
+++ b/rllib/agents/sac/sac.py
@@ -68,9 +68,10 @@ DEFAULT_CONFIG = with_common_config({
     "final_prioritized_replay_beta": 0.4,
     # Whether to LZ4 compress observations
     "compress_observations": False,
-    # If set, this will fix the ratio of sampled to replayed timesteps.
-    # Otherwise, replay will proceed at the native ratio determined by
-    # (train_batch_size / rollout_fragment_length).
+    # If set, this will fix the ratio of replayed from a buffer and learned on
+    # timesteps to sampled from an environment and stored in the replay buffer
+    # timesteps. Otherwise, the replay will proceed at the native ratio
+    # determined by (train_batch_size / rollout_fragment_length).
     "training_intensity": None,
 
     # === Optimization ===
diff --git a/rllib/contrib/maddpg/maddpg.py b/rllib/contrib/maddpg/maddpg.py
index 9a0622e2f..217499a70 100644
--- a/rllib/contrib/maddpg/maddpg.py
+++ b/rllib/contrib/maddpg/maddpg.py
@@ -70,9 +70,10 @@ DEFAULT_CONFIG = with_common_config({
     # Observation compression. Note that compression makes simulation slow in
     # MPE.
     "compress_observations": False,
-    # If set, this will fix the ratio of sampled to replayed timesteps.
-    # Otherwise, replay will proceed at the native ratio determined by
-    # (train_batch_size / rollout_fragment_length).
+    # If set, this will fix the ratio of replayed from a buffer and learned on
+    # timesteps to sampled from an environment and stored in the replay buffer
+    # timesteps. Otherwise, the replay will proceed at the native ratio
+    # determined by (train_batch_size / rollout_fragment_length).
     "training_intensity": None,
     # Force lockstep replay mode for MADDPG.
     "multiagent": merge_dicts(COMMON_CONFIG["multiagent"], {