Add Implicit Quantile Network #371
Conversation
Codecov Report
@@ Coverage Diff @@
## master #371 +/- ##
==========================================
+ Coverage 94.53% 94.64% +0.11%
==========================================
Files 55 56 +1
Lines 3569 3644 +75
==========================================
+ Hits 3374 3449 +75
Misses 195 195
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
tianshou/utils/net/discrete.py
Outdated
| def forward( # type: ignore | ||
| self, s: Union[np.ndarray, torch.Tensor], sample_size: int, **kwargs: Any | ||
| ) -> Tuple[torch.Tensor, Any]: |
There was a problem hiding this comment.
hmm does this change the definition of network?
https://tianshou.readthedocs.io/en/master/tutorials/dqn.html#build-the-network
The first return object can be any form but the second one should be the hidden state
There was a problem hiding this comment.
I see. I'll follow the convention.
tianshou/policy/modelfree/iqn.py
Outdated
| sample_size = self._online_sample_size | ||
| else: | ||
| sample_size = self._sample_size | ||
| logits, h = model(obs_, sample_size, state=state, info=batch.info) |
There was a problem hiding this comment.
better using sample_size=sample_size
I will try to get some results over the weekend but our GPU machines are a bit busy recently.