Feat/autodiff/rnn#1723
Conversation
|
This pull request introduces 35 alerts and fixes 2 when merging a5bff96 into 07af715 - view on LGTM.com new alerts:
fixed alerts:
|
|
retest this please |
b0e025a to
60cd1d4
Compare
|
retest this please |
2d09de1 to
d20fa6f
Compare
|
This pull request introduces 2 alerts and fixes 2 when merging d20fa6f into 07af715 - view on LGTM.com new alerts:
fixed alerts:
|
jvesely
left a comment
There was a problem hiding this comment.
The mxm and outer_product parts look.
The rest of the compiled parts needs significant rework to use standard tools (or add comments why that is not possible)
To be more specific here. I think the LSTM function should have its own class rather than a rather hacky approach of using UDF, since you're writing the important parts anyway it shouldn't be too difficult and it will lead to clearer structure (including LLVM implementation). It should also have standalone tests. Custom |
|
This pull request introduces 2 alerts and fixes 2 when merging 5a083e6 into c7afc35 - view on LGTM.com new alerts:
fixed alerts:
|
42e66e8 to
45c5d1d
Compare
|
This pull request introduces 3 alerts and fixes 2 when merging 45c5d1d into c7afc35 - view on LGTM.com new alerts:
fixed alerts:
|
|
This pull request introduces 3 alerts and fixes 2 when merging b176980 into c7afc35 - view on LGTM.com new alerts:
fixed alerts:
|
|
This pull request introduces 2 alerts and fixes 2 when merging fe18306 into c7afc35 - view on LGTM.com new alerts:
fixed alerts:
|
Required to support single node learning
|
This pull request introduces 2 alerts and fixes 2 when merging b99cd4b into e31dc28 - view on LGTM.com new alerts:
fixed alerts:
|
No description provided.