Fix #811 (#817)

setup.cfg

@@ -8,7 +8,7 @@ exclude =
 	dist
 	*.egg-info
 max-line-length = 87
-ignore = B305,W504,B006,B008,B024,W503
+ignore = B305,W504,B006,B008,B024,W503,B028
 
 [yapf]
 based_on_style = pep8

setup.py

@@ -22,7 +22,6 @@ def get_install_requires() -> str:
         "torch>=1.4.0",
         "numba>=0.51.0",
         "h5py>=2.10.0",  # to match tensorflow's minimal requirements
-        "protobuf~=3.19.0",  # breaking change, sphinx fail
         "packaging",
     ]
 
@@ -30,9 +29,9 @@ def get_install_requires() -> str:
 def get_extras_require() -> str:
     req = {
         "dev": [
-            "sphinx<4",
+            "sphinx",
             "sphinx_rtd_theme",
-            "jinja2<3.1",  # temporary fix
+            "jinja2",
             "sphinxcontrib-bibtex",
             "flake8",
             "flake8-bugbear",

test/base/test_buffer.py

@@ -437,6 +437,38 @@ def test_herreplaybuffer(size=10, bufsize=100, sample_sz=4):
     assert np.all(buf[10:].obs.desired_goal == buf[0].obs.desired_goal)  # (same ep)
     assert np.all(buf[0].obs.desired_goal != buf[5].obs.desired_goal)  # (diff ep)
 
+    # Another test case for cycled indices
+    env_size = 99
+    bufsize = 15
+    env = MyGoalEnv(env_size, array_state=False)
+    buf = HERReplayBuffer(
+        bufsize, compute_reward_fn=compute_reward_fn, horizon=30, future_k=8
+    )
+    buf.future_p = 1
+    for x, ep_len in enumerate([10, 20]):
+        obs, _ = env.reset()
+        for i in range(ep_len):
+            act = 1
+            obs_next, rew, terminated, truncated, info = env.step(act)
+            batch = Batch(
+                obs=obs,
+                act=[act],
+                rew=rew,
+                terminated=(i == ep_len - 1),
+                truncated=(i == ep_len - 1),
+                obs_next=obs_next,
+                info=info
+            )
+            if x == 1 and obs["observation"] < 10:
+                obs = obs_next
+                continue
+            buf.add(batch)
+            obs = obs_next
+    buf._restore_cache()
+    sample_indices = np.array([10])  # Suppose the sampled indices is [10]
+    buf.rewrite_transitions(sample_indices)
+    assert int(buf.obs.desired_goal[10][0]) in [11, 12, 13, 14, 15, 16, 17, 18, 19, 20]
+
 
 def test_update():
     buf1 = ReplayBuffer(4, stack_num=2)

tianshou/data/buffer/her.py

@@ -120,9 +120,10 @@ class HERReplayBuffer(ReplayBuffer):
         # Calculate future timestep to use
         current = indices[0]
         terminal = indices[-1]
-        future_offset = np.random.uniform(size=len(indices[0])) * (terminal - current)
-        future_offset = future_offset.astype(int)
-        future_t = (current + future_offset)
+        episodes_len = (terminal - current + self.maxsize) % self.maxsize
+        future_offset = np.random.uniform(size=len(indices[0])) * episodes_len
+        future_offset = np.round(future_offset).astype(int)
+        future_t = (current + future_offset) % self.maxsize
 
         # Compute indices
         #   open indices are used to find longest, unique trajectories among