zxy844288792 opened a new pull request #7951:
URL: https://github.com/apache/tvm/pull/7951
Encounter a special cases when batch_dims=1 in gather() from
centernet_hourglass_512x512_1 from tensorflow hub model zoo.
Implement the batch_dims logic according to tensorflow implementation:
ht
tmoreau89 opened a new pull request #7950:
URL: https://github.com/apache/tvm/pull/7950
Please join us to welcome @giuseros as a new reviewer. He made significant
contributions towards better supporting ARMv8 and AArch64 in TVM. He also
actively participates in code reviews and forum disc
jcf94 commented on a change in pull request #7948:
URL: https://github.com/apache/tvm/pull/7948#discussion_r623566715
##
File path: docs/conf.py
##
@@ -47,18 +47,32 @@
# General information about the project.
project = "tvm"
author = "Apache Software Foundation"
-copyright =
jcf94 commented on a change in pull request #7948:
URL: https://github.com/apache/tvm/pull/7948#discussion_r623566715
##
File path: docs/conf.py
##
@@ -47,18 +47,32 @@
# General information about the project.
project = "tvm"
author = "Apache Software Foundation"
-copyright =
Meteorix commented on pull request #7944:
URL: https://github.com/apache/tvm/pull/7944#issuecomment-829781229
@tqchen Thanks! Let's build TVM better with the community.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use
This is an automated email from the ASF dual-hosted git repository.
ziheng pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new 62309e5 [TIR][TRANSFORM] Return value support in tir.t
ZihengJiang merged pull request #7932:
URL: https://github.com/apache/tvm/pull/7932
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please
ZihengJiang commented on pull request #7831:
URL: https://github.com/apache/tvm/pull/7831#issuecomment-829742947
LGTM. Thanks @hypercubestart !
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to th
yzhliu commented on pull request #7943:
URL: https://github.com/apache/tvm/pull/7943#issuecomment-829739285
Thanks @hgt312 @sergei-grechanik
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the
This is an automated email from the ASF dual-hosted git repository.
liuyizhi pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from a057d36 Rev ci-qemu container to v0.04. (#7946)
add 49ffab1 init (#7943)
No new revisions were added by this upd
yzhliu merged pull request #7943:
URL: https://github.com/apache/tvm/pull/7943
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please cont
tqchen merged pull request #7946:
URL: https://github.com/apache/tvm/pull/7946
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please cont
This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from 7a06351 [microTVM] Refactor zephyr installation + Update Zephyr RVM
doc (#7915)
add a057d36 Rev ci-qemu container
AndrewZhaoLuo commented on pull request #7928:
URL: https://github.com/apache/tvm/pull/7928#issuecomment-829660797
Ok ONNX integration is done now, as should be most frontend tests. (We'll
see if CI agrees!)
--
This is an automated message from the Apache Git Service.
To respond to the m
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829611498
reshape only was a property about the function itself, the property have
well-defined meaning in functional land. While alias property itself does not
have a clear definition
tqchen commented on a change in pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#discussion_r623420830
##
File path: src/relay/backend/graph_executor_codegen.cc
##
@@ -380,6 +380,19 @@ class GraphExecutorCodegen : public
backend::MemoizedExprTranslatorfunc_name
tqchen commented on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829611498
reshape only was a property about the function itself, the property have
well-defined meaning in functional land. While alias property itself does not
have a clear definition in the
csullivan commented on a change in pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#discussion_r623399795
##
File path: src/relay/backend/graph_executor_codegen.cc
##
@@ -380,6 +380,19 @@ class GraphExecutorCodegen : public
backend::MemoizedExprTranslatorfunc_n
tqchen commented on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829558024
Added comments about the flat memory assumptions and TODOs about what can be
done under non-flat settings once the support is introduced
--
This is an automated message from the Ap
trevor-m opened a new pull request #7949:
URL: https://github.com/apache/tvm/pull/7949
Keras models can have nested layers, where one layer in the model is
actually another Keras model with it's own layers. With this PR, we will now
support these types of models in TVM.
Example Kera
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829551076
@csullivan to followup on your points. I agree that a decision about output
aliasing can be a big undertaking to mark in the IR itself. The rationale that
brings the current P
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829551076
@csullivan to followup on your points. I agree that a decision about output
aliasing can be a big undertaking to mark in the IR itself. The rationale that
brings the current P
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829551076
@csullivan to followup on your points. I agree that a decision about output
aliasing can be a big undertaking to mark in the IR itself. The rationale that
brings the current P
tqchen commented on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829551076
@csullivan to followup on your points. I agree that a decision about output
aliasing can be a big undertaking to mark in the IR itself. The way that the
current PR structured it is a
sergei-grechanik commented on pull request #7943:
URL: https://github.com/apache/tvm/pull/7943#issuecomment-829543534
So the problem is that DivNode may be used for both integers and floats, but
truncdiv checks that its arguments are actually integers. The change looks
correct. I don't thi
tqchen commented on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829532593
Thanks @icemelon9 @altanh comments are addressed by looking for a reshapeop
property.
--
This is an automated message from the Apache Git Service.
To respond to the message, please
zxybazh commented on pull request #7936:
URL: https://github.com/apache/tvm/pull/7936#issuecomment-829465951
IMHO, the intrinsic lowering pass logic is already in `src/tir` and the
intrinsic lowering ruls are more target dependent so I think it's okay if we
keep those rules in `src/target`
altanh commented on a change in pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#discussion_r623262746
##
File path: src/relay/transforms/fuse_ops.cc
##
@@ -949,14 +949,29 @@ class FuseMutator : private MixedModeMutator {
Expr MakeNewFunction(GraphPartition
tkonolige commented on pull request #7917:
URL: https://github.com/apache/tvm/pull/7917#issuecomment-829460968
@mherkazandjian That snippet will fix the issue. `/usr/bin/cc` looks at the
environment variable `SDKROOT` and uses the `cc` in that directory. So if the
user switches their `SDKR
junrushao1994 commented on pull request #7936:
URL: https://github.com/apache/tvm/pull/7936#issuecomment-829457725
@tqchen @zxybazh I am thinking if we should move the intrinsic registeration
logic to `src/tir/intrin`, given it is part of TIR and is logically closer to
IR manipulation
--
tqchen opened a new pull request #7948:
URL: https://github.com/apache/tvm/pull/7948
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, pleas
mherkazandjian commented on pull request #7917:
URL: https://github.com/apache/tvm/pull/7917#issuecomment-829431133
> My apologies for the late docs review. Thanks for pinging me on it.
no worries. tnx for the feedback. i'll check your feedback in a bit.
--
This is an automated mes
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829426059
Thanks @icemelon9 let me take a look and see what is going on there
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to G
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829426059
Thanks @icemelon9 I am not too sure the implication to the VM memalloc
optimizations(as the assumption can differ). Perhaps you or others can send
that as a followup?
--
Th
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829426059
Thanks @icemelon9 I am not too sure the implication to the VM memalloc
optimizations(as the assumption can differ), would be good to make that as a
followup
--
This is an a
tqchen commented on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829426059
@icemelon9 I am not too sure the implication to the VM memalloc
optimizations(as the assumption can differ), would be good to make that as a
followup
--
This is an automated messa
tqchen commented on pull request #2:
URL: https://github.com/apache/tvm-rfcs/pull/2#issuecomment-829422213
cc @apache/tvm-committers final call for feedbacks for 36 hours before we
merge and start the formal voting process
--
This is an automated message from the Apache Git Service.
To r
hogepodge commented on pull request #7917:
URL: https://github.com/apache/tvm/pull/7917#issuecomment-829399526
My apologies for the late docs review. Thanks for pinging me on it.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHu
hogepodge commented on a change in pull request #7917:
URL: https://github.com/apache/tvm/pull/7917#discussion_r623212796
##
File path: docs/deploy/index.rst
##
@@ -46,20 +54,105 @@ on a Linux based embedded system such as Raspberry Pi:
cmake ..
make runtime
-Note t
hogepodge commented on a change in pull request #7917:
URL: https://github.com/apache/tvm/pull/7917#discussion_r623211513
##
File path: docs/deploy/index.rst
##
@@ -46,20 +54,105 @@ on a Linux based embedded system such as Raspberry Pi:
cmake ..
make runtime
-Note t
hogepodge commented on a change in pull request #7917:
URL: https://github.com/apache/tvm/pull/7917#discussion_r623211513
##
File path: docs/deploy/index.rst
##
@@ -46,20 +54,105 @@ on a Linux based embedded system such as Raspberry Pi:
cmake ..
make runtime
-Note t
jwfromm commented on pull request #7937:
URL: https://github.com/apache/tvm/pull/7937#issuecomment-829385657
Yeah i just wanted to echo @mbrookhart that it would be great to include a
test that operates on a real quantized model. Ideally you could start at a
prequantized tf or pytorch mode
mherkazandjian commented on pull request #7917:
URL: https://github.com/apache/tvm/pull/7917#issuecomment-829383992
> @mherkazandjian i think the CI is having some trouble now, feel free to
push an empty commit to retriever
cool. i did a dummy commit to trigger the build, the CI chec
hogepodge commented on pull request #2:
URL: https://github.com/apache/tvm-rfcs/pull/2#issuecomment-829380288
Doesn't look like there's much conversation around this. I'm ready to move
forward on it when you think it's appropriate.
https://discuss.tvm.apache.org/t/rfc-preparing-to-la
This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a commit to branch v0.7
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/v0.7 by this push:
new 701d2c3 [CI] v0.7 CI updates (#7947)
701d2c3 is descri
tqchen merged pull request #7947:
URL: https://github.com/apache/tvm/pull/7947
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please cont
tqchen opened a new pull request #7947:
URL: https://github.com/apache/tvm/pull/7947
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, pleas
areusch opened a new pull request #7946:
URL: https://github.com/apache/tvm/pull/7946
ci-qemu v0.04 now tracks Zephyr v2.5-branch, which allows us to pickup
bugfixes.
passed ci-jenkins-staging enough:
https://ci.tlcpack.ai/blue/organizations/jenkins/tvm/detail/PR-7936/6/pipeline/46
tqchen removed a comment on pull request #7942:
URL: https://github.com/apache/tvm/pull/7942#issuecomment-829313650
You can run `tests/lint/git-clang-format -i upstream/main` (assuming
upstream/main points to the latest main). The error is due to clang format
--
This is an automated mess
tqchen commented on pull request #7942:
URL: https://github.com/apache/tvm/pull/7942#issuecomment-829313650
You can run `tests/lint/git-clang-format -i upstream/main` (assuming
upstream/main points to the latest main). The error is due to clang format
--
This is an automated message from
mbrookhart commented on pull request #7937:
URL: https://github.com/apache/tvm/pull/7937#issuecomment-829295087
Hi @huochaitiantang ,
That all makes sense, I'm really just wondering if we've ever seen an
example of this in the wild, i.e. q->conv->dq instead of q->convinteger->dq. Do
tqchen edited a comment on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829275660
cc @altanh @jroesch @zhiics @icemelon9 @Meteorix @ZihengJiang @junrushao1994
--
This is an automated message from the Apache Git Service.
To respond to the message, please l
tqchen commented on pull request #7945:
URL: https://github.com/apache/tvm/pull/7945#issuecomment-829275660
cc @altanh @jroesch @zhiics @icemelon9 @Meteorix @ZihengJiang
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and us
tqchen opened a new pull request #7945:
URL: https://github.com/apache/tvm/pull/7945
Previously we are generating the function calls for reshape.
This PR updates the optimization to turn reshape into nop:
- Tag a fused function as reshape only if it only contains reshape.
- Upda
AlexanderSerov commented on pull request #7942:
URL: https://github.com/apache/tvm/pull/7942#issuecomment-829274051
Update PR but it not solve the problem, checks locally. Can't understand
what's wrong
--
This is an automated message from the Apache Git Service.
To respond to the message
tqchen opened a new pull request #7944:
URL: https://github.com/apache/tvm/pull/7944
Dear community:
Please join us to welcome @Meteorix as a new reviewer.
He contributed a few important improvements to TensorCore's enablement and
brings
helpful discussions to performance opti
tqchen commented on pull request #7942:
URL: https://github.com/apache/tvm/pull/7942#issuecomment-829165009
@AlexanderSerov please resolve the lint error
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above t
Tantalus13A98B5F commented on pull request #7250:
URL: https://github.com/apache/tvm/pull/7250#issuecomment-829153018
Hi there. The removed `RPCRunner.ref_input` interface can be used to supply
input for certain operators, e.g. a sparse one where random indices are
unacceptable. Maybe we s
giuseros commented on pull request #7785:
URL: https://github.com/apache/tvm/pull/7785#issuecomment-829117394
Hi @areusch ,
Should we set for a middle ground solution? I moved back the python
interface of `tvm.relay.build`, which now retrieves the executor from the
target. The inner fun
huochaitiantang commented on pull request #7937:
URL: https://github.com/apache/tvm/pull/7937#issuecomment-829092342
@mbrookhart Thanks for your comment!
I have checked the supported input types of operators in ORT:
https://github.com/microsoft/onnxruntime/blob/master/docs/OperatorKern
hgt312 opened a new pull request #7943:
URL: https://github.com/apache/tvm/pull/7943
With this change, I can use `te.gradient` on `roi_align`. Otherwise, there
will be:
```
TVMError:
---
An internal invariant was vi
AlexanderSerov commented on pull request #7942:
URL: https://github.com/apache/tvm/pull/7942#issuecomment-829049394
@tqchen
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
AlexanderSerov opened a new pull request #7942:
URL: https://github.com/apache/tvm/pull/7942
sort.cc added to wasn runtime for nms support
Solves [this issue](https://github.com/apache/tvm/issues/7849)
--
This is an automated message from the Apache Git Service.
To respond to the messa
63 matches
Mail list logo