profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/csrhddlam/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Huiyu Wang csrhddlam Johns Hopkins University Baltimore, MD

joe-siyuan-qiao/WeightStandardization 500

Standardizing weights to accelerate micro-batch training

csrhddlam/axial-deeplab 343

This is a PyTorch re-implementation of Axial-DeepLab (ECCV 2020 Spotlight)

csrhddlam/examples 0

A set of examples around pytorch in Vision, Text, Reinforcement Learning, etc.

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha fccee86db2f5242809ee4def2d90377e2e9a1492

Update index.md

view details

push time in 11 days

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 5d7a0c0dba2469dabc2874e28fd3a030b9e60a26

Update index.md

view details

push time in 11 days

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 41c4cf2ee4935e3bc0d1e24870d128ccf0669208

Update index.md

view details

push time in 12 days

issue commentgoogle-research/deeplab2

Can't run trainer/train.py well

Hi fxf000,

Thank you for your interest in our work and the effort at running the code. There seems to be around two months since the last update. Are you able to load the 'libcusolver.so.11' now?

fxf000

comment created time in a month

issue commentgoogle-research/deeplab2

Execution of trainer/train.py stops in the middle.

Hi,

Thanks for your interest in our work. I think this issue is probably because MaX-DeepLab requires TensorFlow>=2.6.0, while your docker tensorflow:21.08 uses TensorFlow==2.5.0.

For more information about why TensorFlow 2.5 does not work, please refer to issue #24. For more elaborative error messages, you could also try disabling the loop in Orbit as is done in issue #24. The orbit loop suppresses error messages.

Best, Huiyu

shu1106

comment created time in a month

starteddeepmind/alphafold

started time in 2 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha bb2f250cba0d1318fa24f364c96aa79d451ab45e

remove wrong CV.

view details

push time in 2 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 1d6e240e989a0f55773b52a9a51e43612485ce9e

Add files via upload

view details

push time in 2 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 0819e94f8c75bc39bf76da4e7467f119d9c8a963

Add files via upload

view details

push time in 2 months

issue commentgoogle-research/deeplab2

Questions about batch_norm_on_an_extra_axis

Hi,

Thanks for your interest in our work.

  1. The use of "batch_norm_on_an_extra_axis" can avoid careful scaling of the last layer which is the multiplication of an L2-normalized pixel space feature and a batch-normalized transformer kernel. This choice is simple and works well in practice.
  2. Without the batch norm but with a careful choice of the scaling factor and a good learning rate: We didn't try this option, but I think it can probably work as well as the default batch norm choice.
  3. Without the batch norm or a proper scaling factor, the output scale or the multiplication is not controlled and might not work as well as the previous two choices.

The PyTorch implementation looks correct to me. But note that we use synchronized batch norm for all segmentation cases.

Best, Huiyu

BritaryZhou

comment created time in 2 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 29998b898d4d148f619d5f096535417b82f84cc9

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 71d3e7422845e9f7eaf09bc2d267fc9d728c3329

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 936991d591720a57d3f8d9381ede989a87da78d8

Update _base.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha c260e338ae5d0bd9445f8b73aac39cfaf922cbb3

Update _base.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 20f421a21e327adb7d03dd8b191ad008e5ebeb1a

Update custom.html

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha a724b90e3c87632fb813ac84ec6db9e828cc7fee

Update _base.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha d1611f51ed4f457683457e7df967beb7caa2ee93

Update _base.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha aaf0a9a4c4844d32bb76af63bf91a0dbf0eec8cf

Update _base.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 7acdbc458591ae0abedaca2434d98cc7dba1aa69

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 3849023b0a06874626356e28990f6dc18620de34

Update _base.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha d617e42e5d1713776d7db3df9a5a63e124d77cc1

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha f5a70f75cc4f059d4b1650181a1c64af91161307

Update _sidebar.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha ea9fbabb81cada423ca129b382f0273daed6afc9

Update _base.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 0d25fa61fd30bacbce18c08ba97390af5adecafb

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 167556e3cbedf679b084c445a63794de8154d931

Update _variables.scss

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha f8832f9bf2ff3b646ca6828d4b951419d2306f98

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 3082ce80ad5232b869cbb39053aa3b159f25b572

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha e11954eed119bd91df9a85397e951b0ca68eda84

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 05a606dbb35725e72cadbfda5950d3656b3ae4ab

Update index.md

view details

push time in 3 months

push eventcsrhddlam/csrhddlam.github.io

Huiyu Wang

commit sha 5551147b543f627d8d8c43b728f8f7d5592c8ce5

Update index.md

view details

push time in 3 months