94bfb32cc1
1. add policy.eval() in all test scripts' "watch performance" 2. remove dict return support for collector preprocess_fn 3. add `__contains__` and `pop` in batch: `key in batch`, `batch.pop(key, deft)` 4. exact n_episode for a list of n_episode limitation and save fake data in cache_buffer when self.buffer is None (#184) 5. fix tensorboard logging: h-axis stands for env step instead of gradient step; add test results into tensorboard 6. add test_returns (both GAE and nstep) 7. change the type-checking order in batch.py and converter.py in order to meet the most often case first 8. fix shape inconsistency for torch.Tensor in replay buffer 9. remove `**kwargs` in ReplayBuffer 10. remove default value in batch.split() and add merge_last argument (#185) 11. improve nstep efficiency 12. add max_batchsize in onpolicy algorithms 13. potential bugfix for subproc.wait 14. fix RecurrentActorProb 15. improve the code-coverage (from 90% to 95%) and remove the dead code 16. fix some incorrect type annotation The above improvement also increases the training FPS: on my computer, the previous version is only ~1800 FPS and after that, it can reach ~2050 (faster than v0.2.4.post1).
90 lines
2.2 KiB
Python
90 lines
2.2 KiB
Python
import pytest
|
|
import numpy as np
|
|
|
|
from tianshou.data import (ListReplayBuffer, PrioritizedReplayBuffer,
|
|
ReplayBuffer, SegmentTree)
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def data():
|
|
np.random.seed(0)
|
|
obs = {'observable': np.random.rand(100, 100),
|
|
'hidden': np.random.randint(1000, size=200)}
|
|
info = {'policy': "dqn", 'base': np.arange(10)}
|
|
add_data = {'obs': obs, 'rew': 1., 'act': np.random.rand(30),
|
|
'done': False, 'obs_next': obs, 'info': info}
|
|
buffer = ReplayBuffer(int(1e3), stack_num=100)
|
|
buffer2 = ReplayBuffer(int(1e4), stack_num=100)
|
|
indexes = np.random.choice(int(1e3), size=3, replace=False)
|
|
return {
|
|
'add_data': add_data,
|
|
'buffer': buffer,
|
|
'buffer2': buffer2,
|
|
'slice': slice(-3000, -1000, 2),
|
|
'indexes': indexes,
|
|
}
|
|
|
|
|
|
def test_init():
|
|
for _ in np.arange(1e5):
|
|
_ = ReplayBuffer(1e5)
|
|
_ = PrioritizedReplayBuffer(size=int(1e5), alpha=0.5, beta=0.5)
|
|
_ = ListReplayBuffer()
|
|
|
|
|
|
def test_add(data):
|
|
buffer = data['buffer']
|
|
for _ in np.arange(1e5):
|
|
buffer.add(**data['add_data'])
|
|
|
|
|
|
def test_update(data):
|
|
buffer = data['buffer']
|
|
buffer2 = data['buffer2']
|
|
for _ in np.arange(1e2):
|
|
buffer2.update(buffer)
|
|
|
|
|
|
def test_getitem_slice(data):
|
|
Slice = data['slice']
|
|
buffer = data['buffer']
|
|
for _ in np.arange(1e3):
|
|
_ = buffer[Slice]
|
|
|
|
|
|
def test_getitem_indexes(data):
|
|
indexes = data['indexes']
|
|
buffer = data['buffer']
|
|
for _ in np.arange(1e2):
|
|
_ = buffer[indexes]
|
|
|
|
|
|
def test_get(data):
|
|
indexes = data['indexes']
|
|
buffer = data['buffer']
|
|
for _ in np.arange(3e2):
|
|
buffer.get(indexes, 'obs')
|
|
buffer.get(indexes, 'rew')
|
|
buffer.get(indexes, 'done')
|
|
buffer.get(indexes, 'info')
|
|
|
|
|
|
def test_sample(data):
|
|
buffer = data['buffer']
|
|
for _ in np.arange(1e1):
|
|
buffer.sample(int(1e2))
|
|
|
|
|
|
def test_segtree(data):
|
|
size = 100000
|
|
tree = SegmentTree(size)
|
|
tree[np.arange(size)] = np.random.rand(size)
|
|
|
|
for i in np.arange(1e5):
|
|
scalar = np.random.rand(64) * tree.reduce()
|
|
tree.get_prefix_sum_idx(scalar)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
pytest.main(["-s", "-k buffer_profile", "--durations=0", "-v"])
|