You are viewing a plain text version of this content. The canonical link for it is here.

Posted to discuss-archive@tvm.apache.org by Katherine Zamudio via Apache TVM Discuss <no...@discuss.tvm.ai> on 2022/02/22 18:17:46 UTC

[Apache TVM Discuss] [Questions] [VTA] Questions about VTA packed format


Also got an error at line `assert dshape[1] % cfactor == 0` of *_pack_weight* with a MobileNetV2-based NN made with PyTorch. When playing with *start_name* and *stop_name* and their indexes, it appears that some layers have Tensor sizes of [a, **1**, b, c], causing the assert to fail. Those layers seem to be the bottleneck layers' middle convolution layers...

Would implementing a `assert dshape[1] % cfactor != 0` case really be the only way to fix this issue?

Has anyone else tried MobileNetV2 with a VTA *graph_pack* by any chance?

Thanks.





---
[Visit Topic](https://discuss.tvm.apache.org/t/vta-questions-about-vta-packed-format/1878/8) to respond.

You are receiving this because you enabled mailing list mode.

To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/4ba93fb4841d7a833d366fafbf6da2de16e7857123a47e579d0300edfc90a79a).

[Apache TVM Discuss] [Questions] [VTA] Questions about VTA packed format

Posted by Julien Posso via Apache TVM Discuss <no...@discuss.tvm.ai>.


The problem raised by @KZamudio is solved. The 1 in the tensor shape [a, **1**, b, c] corresponds to the depthwise convolution (convolution is only computed on 1 channel at a time). The solution was to transform the depthwise convolution into a grouped convolution with groups of size 16 as VTA only supports convolutions with a number of channels which is a multiple of 16.





---
[Visit Topic](https://discuss.tvm.apache.org/t/vta-questions-about-vta-packed-format/1878/10) to respond.

You are receiving this because you enabled mailing list mode.

To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/83ba11e721f17eaaa7c42cc08931b9613e41b687580d84c8a4b469c1d624104a).

[Apache TVM Discuss] [Questions] [VTA] Questions about VTA packed format

Posted by okkk via Apache TVM Discuss <no...@discuss.tvm.ai>.


Hi, has your problem been solved?





---
[Visit Topic](https://discuss.tvm.apache.org/t/vta-questions-about-vta-packed-format/1878/9) to respond.

You are receiving this because you enabled mailing list mode.

To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/56bbfd0d5acbd1d06e52fe72685dacb09e34bcbb47b73360c88799249b4b788b).