[GitHub] [incubator-tvm] apivovarov commented on issue #4456: [Relay][Legalize] Legalize conv2d cuda for NHWC

apivovarov commented on issue #4456: [Relay][Legalize] Legalize conv2d cuda for NHWC URL: https://github.com/apache/incubator-tvm/pull/4456#issuecomment-562921277 This PR is needed to compile TFLite models for CUDA. By design TFLite Relay frontend only supports NHWC target layout.

[GitHub] [incubator-tvm] masahi commented on issue #4478: [TOPI] implement pool3d op

masahi commented on issue #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#issuecomment-562912625 @optima2005 I am traveling for another few days, will review afte I am back. This is an

[GitHub] [incubator-tvm] optima2005 commented on a change in pull request #4478: [TOPI] implement pool3d op

optima2005 commented on a change in pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#discussion_r355154998 ## File path: src/relay/op/nn/pooling.cc ## @@ -720,5 +720,257 @@ RELAY_REGISTER_OP("nn.avg_pool2d_grad")

[GitHub] [incubator-tvm] optima2005 commented on a change in pull request #4478: [TOPI] implement pool3d op

optima2005 commented on a change in pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#discussion_r355154973 ## File path: topi/include/topi/nn/pooling.h ## @@ -591,6 +593,211 @@ inline Tensor global_pool(const Tensor& x,

[GitHub] [incubator-tvm] optima2005 commented on a change in pull request #4478: [TOPI] implement pool3d op

optima2005 commented on a change in pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#discussion_r355154435 ## File path: topi/src/topi.cc ## @@ -595,7 +602,7 @@ TVM_REGISTER_GLOBAL("topi.generic.schedule_injective")

[GitHub] [incubator-tvm] vinx13 merged pull request #4472: Workaround to make conv2d_transpose compilation for CUDA work

vinx13 merged pull request #4472: Workaround to make conv2d_transpose compilation for CUDA work URL: https://github.com/apache/incubator-tvm/pull/4472 This is an automated message from the Apache Git Service. To respond to

[incubator-tvm] branch master updated: Workaround to make conv2d_transpose compilation for CUDA work (#4472)

This is an automated email from the ASF dual-hosted git repository. wuwei pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git The following commit(s) were added to refs/heads/master by this push: new 3fa8097 Workaround to make

[GitHub] [incubator-tvm] zhiics commented on issue #4479: [Relay] Check function attr for hash

zhiics commented on issue #4479: [Relay] Check function attr for hash URL: https://github.com/apache/incubator-tvm/pull/4479#issuecomment-562875552 @MarisaKirisame Yes, you are right. Just changed the title. This is an

[GitHub] [incubator-tvm] zhiics opened a new pull request #4479: [Relay] Check function attr for alpha equal

zhiics opened a new pull request #4479: [Relay] Check function attr for alpha equal URL: https://github.com/apache/incubator-tvm/pull/4479 This PR adds function attrs for equality check. @jroesch @icemelon9 @wweic @MarisaKirisame

[GitHub] [incubator-tvm] masahi commented on a change in pull request #4478: [TOPI] implement pool3d op

masahi commented on a change in pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#discussion_r355132397 ## File path: src/relay/op/nn/pooling.cc ## @@ -720,5 +720,257 @@ RELAY_REGISTER_OP("nn.avg_pool2d_grad")

[GitHub] [incubator-tvm] masahi commented on a change in pull request #4478: [TOPI] implement pool3d op

masahi commented on a change in pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#discussion_r355131637 ## File path: topi/src/topi.cc ## @@ -595,7 +602,7 @@ TVM_REGISTER_GLOBAL("topi.generic.schedule_injective")

[GitHub] [incubator-tvm] masahi commented on a change in pull request #4478: [TOPI] implement pool3d op

masahi commented on a change in pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#discussion_r355131637 ## File path: topi/src/topi.cc ## @@ -595,7 +602,7 @@ TVM_REGISTER_GLOBAL("topi.generic.schedule_injective")

[GitHub] [incubator-tvm] masahi commented on a change in pull request #4465: [AutoTVM] Tune softmax CUDA schedule

masahi commented on a change in pull request #4465: [AutoTVM] Tune softmax CUDA schedule URL: https://github.com/apache/incubator-tvm/pull/4465#discussion_r355130610 ## File path: topi/python/topi/cuda/softmax.py ## @@ -52,13 +60,22 @@ def schedule_softmax(outs):

[GitHub] [incubator-tvm] masahi commented on a change in pull request #4478: [TOPI] implement pool3d op

masahi commented on a change in pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478#discussion_r355131252 ## File path: topi/include/topi/nn/pooling.h ## @@ -591,6 +593,211 @@ inline Tensor global_pool(const Tensor& x,

[GitHub] [incubator-tvm] tqchen commented on issue #4477: why time taking are added, when speed up ssd by TVM ?

tqchen commented on issue #4477: why time taking are added,when speed up ssd by TVM ? URL: https://github.com/apache/incubator-tvm/issues/4477#issuecomment-562865756 thanks for asking the question, please open a new thread on https://discuss.tvm.ai/

[GitHub] [incubator-tvm] tqchen closed issue #4477: why time taking are added, when speed up ssd by TVM ?

tqchen closed issue #4477: why time taking are added,when speed up ssd by TVM ? URL: https://github.com/apache/incubator-tvm/issues/4477 This is an automated message from the Apache Git Service. To respond to the

[GitHub] [incubator-tvm] masahi commented on a change in pull request #4465: [AutoTVM] Tune softmax CUDA schedule

masahi commented on a change in pull request #4465: [AutoTVM] Tune softmax CUDA schedule URL: https://github.com/apache/incubator-tvm/pull/4465#discussion_r355130610 ## File path: topi/python/topi/cuda/softmax.py ## @@ -52,13 +60,22 @@ def schedule_softmax(outs):

[GitHub] [incubator-tvm] optima2005 opened a new pull request #4478: [TOPI] implement pool3d op

optima2005 opened a new pull request #4478: [TOPI] implement pool3d op URL: https://github.com/apache/incubator-tvm/pull/4478 This is a part of attempts to implement #4009 The implemation in this PR is for max and avg 3d pooling. @masahi

[GitHub] [incubator-tvm] duamu opened a new issue #4477: why time taking are added, when speed up ssd by TVM ?

duamu opened a new issue #4477: why time taking are added,when speed up ssd by TVM ? URL: https://github.com/apache/incubator-tvm/issues/4477 when we speed up SSD as the tutorial（https://docs.tvm.ai/tutorials/frontend/deploy_ssd_gluoncv.html） ，time taking are added（40ms before speed