You should be able to use the op-based annotation now. The PR we merged today
provides a pass to combine supported ops in a single subgraph:
https://github.com/apache/incubator-tvm/pull/5134
---
[Visit
Topic](https://discuss.tvm.ai/t/external-codegen-status-of-annotating-composite-functio
Hi, I have an LSTM autoencoder model which I wanted to convert with input shape
: (1,12,6). However, I am getting this error
```
Traceback (most recent call last):
[bt] (3) /usr/tvm/build/libtvm.so(TVMFuncCall+0x61) [0x7f2fda57dad1]
[bt] (2) /usr/tvm/build/libtvm.so(+0x45e98d) [0x7f2fd9d50
I was playing around with the new external codegen & composite functions and
noticed this flow fails inside of AnnotateTarget. Wondering if it is because of
this
[TODO](https://github.com/apache/incubator-tvm/blob/021213832cb98703dda54f631215ac17fbabff7b/src/relay/transforms/annotate_target.cc
I found the reason of such undefined symbol error. In
CMakeFiles/tvm_runtime.dir/link.txt, it use `/usr/bin/gcc` rather than
`/usr/bin/c++`. So the libtvm_runtime.so cannot find libstdc++.so to link. I
manually modify link.txt, and make again. It worked. But I wonder why link.txt
use gcc inst
I find the 'while/GatherV2' is the input of 'NextIteration',It will continue to
be passed in the 'while_loop‘.
```
name: "while/NextIteration_26"
op: "NextIteration"
input: "while/GatherV2"
...
name: "while/Merge_26"
op: "Merge"
input: "while/Enter_26"
input: "while/NextIteration_26"
..
I try to transform the NLP model using from_tensorflow, then get an error.
```
..
%1817 = add(%35, %1816);
%1818 = reshape(%1817, newshape(-1,[28500]))
%1819 = topk(%1818, k=3, dtype='int32')
%1820 = %1819.1
%1821 = cast(%1820, dtype='float64')
%1822 = cast(9500, dtype='float64')
%1823 = divi
https://github.com/apache/incubator-tvm/issues/5172
---
[Visit
Topic](https://discuss.tvm.ai/t/frontend-on-tensorflow-py-the-stridedslice-have-some-problems/6146/1)
to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click
here](h
