Added reset teset in supply chain scenarios #550
Conversation
tests/supply_chain/test_env_reset.py
Outdated
|
|
||
| # ##################################### Before reset ##################################### | ||
|
|
||
| order_1 = Order( |
There was a problem hiding this comment.
why do we place specific order_1, order_2, order_3, and also orders in random ticks?
There was a problem hiding this comment.
The idea I set up like this is that order_1, order_2, order_3, are their own unit functions. Then the random ticks later is because, I think the current order needs to be passed in the current env.tick, which will affect the expected_finish_tick. So the purpose is to randomly tick several orders in the set rounds to make the test more perfect.
tests/supply_chain/test_env_reset.py
Outdated
| env.step(None) | ||
| env_metric_1[i] = env.metrics | ||
| states_1[i] = list(storage_nodes[i:storage_node_index:features].flatten().astype(np.int)) | ||
| states_1[i].append(storage_nodes[i:storage_node_index:"product_id_list"].flatten().astype(np.int).sum()) |
There was a problem hiding this comment.
I understand it this way, because many attributes of the snapshot inch of storage are NodeAttributeAccessor objects, which are of ndarray type after being taken out from the snapshot. I was afraid that it would be difficult to compare, so I added sum().
tests/supply_chain/test_env_reset.py
Outdated
|
|
||
| class MyTestCase(unittest.TestCase): | ||
| """ | ||
| . consumer unit test |
There was a problem hiding this comment.
no need to separate the snapshots checking of different units into different tests.
Maybe you can organize the tests to:
- test env reset with None action
- test env reset with ManufactureAction only
- test env reset with ConsumerAction only
- test env reset with both ManufactureAction and ConsumerAction
and also enrich the action onwer.
There was a problem hiding this comment.
Okay, I will modify it in the direction you said.
Add reset test in sc scene and fix the problem found by reset test
Modify test according to comments
Modify test according to comments ++
| env_metric_1[i] = env.metrics | ||
|
|
||
| for idx in range(len(consumer_nodes)): | ||
| states_1_consumer[i][idx] = consumer_nodes[i:idx:consumer_features].flatten().astype(np.int) |
There was a problem hiding this comment.
some are int, some are float
|
|
||
| # snapshot should reset after env.reset(). | ||
| for idx in range(len(manufacture_nodes)): | ||
| states = manufacture_nodes[1:idx:manufacture_features].flatten().astype(np.int) |
| states = seller_nodes[1:idx:seller_features].flatten().astype(np.int) | ||
| self.assertEqual([0, 0, 0, 0, 0, 0, 0, 0], list(states)) | ||
|
|
||
| expect_tick = 10 |
There was a problem hiding this comment.
no need to re-claim expect_tick
| manufacture_nodes = env.snapshot_list["manufacture"] | ||
| distribution_nodes = env.snapshot_list["distribution"] | ||
|
|
||
| consumer_features = ( |
There was a problem hiding this comment.
you can put these features in common, no need to re-claim them in each test function
| states_2_manufacture: Dict[int, dict] = defaultdict(dict) | ||
| states_2_distribution: Dict[int, dict] = defaultdict(dict) | ||
|
|
||
| for i in range(expect_tick): |
There was a problem hiding this comment.
-
you can wrap line 110-130 as a function, so as to reduce the redundant code(current line 170-190, similar parts in other test functions)
-
snapshot can be accessed (all ticks) only once at the end of the simulation.
| @@ -0,0 +1,790 @@ | |||
| # Copyright (c) Microsoft Corporation. | |||
There was a problem hiding this comment.
Both file test_action_reset and test_env_reset in this PR, prefer the naming test_env_reset
Modify test according to comments +++
Description
Added reset teset and fixed distribution unit reset issue in supply chain scenarios
Type of Change
Related Component
Has Been Tested
Needs Follow Up Actions
Checklist