[Relay][Op][TF] Complete tensor array unstack with all ranks support#4309
[Relay][Op][TF] Complete tensor array unstack with all ranks support#4309zhiics merged 1 commit intoapache:masterfrom
Conversation
| if any(['Rank' in param for param in params]): | ||
| limit = params.pop('Rank').asnumpy()[0] | ||
| else: | ||
| limit = _infer_value_simulated(inputs[1], params).asnumpy()[0] |
There was a problem hiding this comment.
Thanks a lot for adding this. While you're here, would you mind adding the fix to _transpose() too?
...
except (IndexError, KeyError):
axes = _infer_value_simulated(inputs[1], params)
There was a problem hiding this comment.
@soiferj How can I reproduce the failure? I can add it if I can add a test case for it.
There was a problem hiding this comment.
I think it should reproduce when the perm input is a variable, instead of a constant. For a test case, I suppose you could make it a placeholder.
There was a problem hiding this comment.
There was a problem hiding this comment.
@soiferj Thank you! I thought about it a bit, it doesn't seem to relate to my pr though the fix looks similar. We can definitely send another pr to fix later. I'd suggest we keep the pr single purpose. What do you think?
|
Thanks everyone. This is now merged. |
5 participants
Thanks @soiferj for providing
Rangefix in his branch, I need it to fix the unit test, so included in the PR.cc @icemelon9 @kevinthesun @zhiics @yongwww @srkreddy1238 @soiferj