5176 Commits

Author SHA1 Message Date
Tingquan Gao
078a7a8c44
support to pir (#3194) 2024-07-24 14:47:59 +08:00
wanghuancoder
39adc19904
refine PPLCNet benchmark config (#3191)
* refine PPLCNet benchmark config

* refine
2024-07-17 10:37:37 +08:00
Nyakku Shigure
6aa3047157
Respect is_rec when eval (#3187) 2024-07-10 15:45:37 +08:00
Nyakku Shigure
caa6fdf57d
Fix missing label for classification model in eval mode (#3185) 2024-07-09 17:02:15 +08:00
Nyakku Shigure
d6df377f44
Mark some models as rec model in yaml (#3182) 2024-07-08 17:17:53 +08:00
Tingquan Gao
587698a9df
bugfix (#3179) 2024-07-05 16:17:42 +08:00
Tingquan Gao
91e8eb3632
update to be compatible with V100 (#3178) 2024-07-05 14:52:32 +08:00
Nyakku Shigure
1dcab0a7bc
Fix wrong InputSpec setting for AST Dy2St (#3176) 2024-07-05 14:05:02 +08:00
Tingquan Gao
b1ee8f911b
update to be compatible with V100 (#3177) 2024-07-04 16:56:04 +08:00
Nyakku Shigure
cc24ead0ad
Use inspect.getfullargspec instead of deprecated inspect.getargspec (#3168) 2024-06-20 20:16:36 +08:00
zhuyipin
cea80378ac
adapt AdaptiveAvgPool2D for npu for PPHGNet (#3162) 2024-06-17 21:07:16 +08:00
gmm
be6e0b7b53
【benchmark】add use_fused_attn=True for benchmark (#3160)
* add max_memory_reserved for benchmark

* change unit

* fix

* fix nvidia-dali-cuda110;DLTP-54994

* add use_fused_attn=True  for benchmark
2024-06-06 11:25:51 +08:00
Tingquan Gao
00e018c18c
support FusedLinear (#3159) 2024-06-04 11:15:47 +08:00
gaotingquan
87ad22342e add use_fused_attn field 2024-05-29 11:24:48 +08:00
gaotingquan
46c1a0168c set use_fused_op False by default 2024-05-29 11:24:48 +08:00
gaotingquan
0bfed92cb2 perf:
1. use nn.GELU instead of QuickGELU
2. support FusedLinear
2024-05-29 11:24:48 +08:00
gaotingquan
d8beab064d support fused attn 2024-05-29 11:24:48 +08:00
gaotingquan
40042f89fa dbg: support fused attn 2024-05-24 14:27:29 +08:00
Tingquan Gao
e3aaa3cefb
support fused attn (#3131) 2024-05-16 13:33:46 +08:00
will-jl944
a0ae182547
[Bug Fix] Fix bug in AdaptiveAvgPool2D for NPU (#3147)
* Fix bug in AdaptiveAvgPool2D for NPU

* fix export error
2024-05-16 12:23:36 +08:00
wanghuancoder
0015f98041
refine drop_path (#3132)
* refine drop_path

* refine
2024-05-15 17:51:51 +08:00
zhuyipin
0f915713ec
convert npu roll op into paddle roll (#3138)
* convert npu roll op into paddle roll

* convert npu roll op into paddle roll
2024-05-15 17:11:08 +08:00
will-jl944
d1ae38d30d
AdaptiveAvgPool2D for NPU (#3145) 2024-05-15 16:24:36 +08:00
zhangyubo0722
e23bdb4df5
merge uapi paddleclas (#3135) 2024-05-13 14:27:09 +08:00
Wang Huan
5b06e18b24 refine drop_path 2024-05-07 14:20:10 +08:00
Wang Huan
07de476c51 refine 2024-04-26 14:58:55 +08:00
Wang Huan
0cb74b1741 refine 2024-04-26 14:58:55 +08:00
Wang Huan
be33107d89 refine batch_cost stat 2024-04-26 14:58:55 +08:00
wanghuancoder
80abf9f789
use tensor.shape bug not paddle.shape(tensor) (#3120) 2024-04-12 15:34:44 +08:00
jzhang533
740c0076ff
try using github action to automate release package uploading (#3114) 2024-03-27 10:44:27 +08:00
gmm
3ba7a8d273
【benchmark】update nvidia_dali_cuda110 version (#3115)
* add max_memory_reserved for benchmark

* change unit

* fix

* fix nvidia-dali-cuda110;DLTP-54994
2024-03-26 19:52:20 +08:00
zxcd
967c6a7829
adapter new type promotion rule for Paddle 2.6 v2 (#3111) 2024-03-20 11:39:04 +08:00
zxcd
67c6517b02
adapter new type promotion rule for Paddle 2.6 (#3108) 2024-03-12 20:32:51 +08:00
changdazhou
a39351537d
Ctrl mem log (#3107)
* support ctrl print memory info

* support ctrl print memory info

* fix print log bug

* support default print_mem_info=True for tipc
2024-03-12 20:32:27 +08:00
changdazhou
d189384ab8
Control print memory info (#3079)
* support ctrl print memory info

* support ctrl print memory info

* fix print log bug
2024-03-05 11:28:23 +08:00
sky
54767fdda4
fix adamwdl bug (#3097) 2024-03-04 14:18:29 +08:00
sky
1ded6d1cbd
Update svtrnet.py (#3037)
* Update svtrnet.py

* Update svtrnet.py

remove test part

* isolated model

* Update svtrnet.py

* support static graph

* try to fix acc problem by remove torch format

* fix sercret bug for losing accuracy

* Revert "fix sercret bug for losing accuracy"

This reverts commit c8226e739005c7ebd0c0ca3772c568bccd958c4f.

* fix secret bug for losing accuracy metric

* Delete launch.json

* Update svtrnet.py

* fix 动转静

* Update svtrnet.py
2024-02-06 21:50:38 +08:00
sky
276e90d9a1
Bigmodel (#3032)
* fix the resolution problem for clip-vision transformer part and swim transformer

fix the resolution problem for clip-vision transformer part and swim transformer

* Revert "Revert "fix resolution problem for swin transformer and clip vit  (#3021)""

This reverts commit 174db431a82fb168c01b0be03fbb1d822314bbb1.

Update foundation_vit.py

Update foundation_vit.py

Revert "fix resolution problem for swin transformer and clip vit  (#3021)"

This reverts commit 61f748de6767daa2ee8c243146b2204689ecd1ba.

* add backbone function

* fix static graph problem

* remove text encoder framework and add classifier header directly

* fix bug in clip when using classifier header

* updated

* support embeding

Note, support embed is only for check the model since there is no related text encoder

* compatible with param transfer

* update setting
2024-02-06 21:46:25 +08:00
Ran chongzhi
5ac8074190
[Feature]Complete the PaddleClas models' auto_compression (#3082)
* add paddleslim act examples

* fulfill_class_act

* update

* Modify config file

* add test shell scripts

* Modify result and docs

* update model urls

---------

Co-authored-by: vax521 <13263397018@163.com>
Co-authored-by: shiyutang <1574572981@qq.com>
2024-02-04 15:12:10 +08:00
gaotingquan
879cb57fe6 fix error dtype 2023-12-08 14:28:29 +08:00
gmm
034236d0f5
【benchmark】fix gpu_mem unit (#3052)
* add max_memory_reserved for benchmark

* change unit

* fix
2023-12-05 15:04:25 +08:00
cuicheng01
d17336da37
Develop (#3047)
* add PPLCNet_x1_0_ssld.yaml
2023-11-24 21:43:43 +08:00
RuohengMa
178eede1db
Fix bug when running on XPU (#3046)
* Fix bug when running on XPU

* Fix bug when running on device other than GPU
2023-11-24 17:14:56 +08:00
RuohengMa
656c149d6b
Fix bug when running on XPU (#3045) 2023-11-24 15:08:26 +08:00
gmm
e4abcc08d0
add max_memory_reserved for benchmark (#3042) 2023-11-22 17:31:30 +08:00
Tingquan Gao
ab087065e9
support to specify rank to log when using Fleet API (#3039)
* support to specify rank to log when using Fleet API

* log max mem reserved

* log_ranks support str type

example: -o Global.log_ranks="0,1"

* log max mem allocated

* support to specify rank to log in static mode

* log max mem reserved and max mem allocated in static mode
2023-11-16 11:32:29 +08:00
Zhenming Lin
cc0ac63fb8
包含所有样本,不管有无目标类别的标注。 (#2963) 2023-11-07 22:12:09 +08:00
zhangyubo0722
adedfe603c
[WIP] add svtr (#3035)
* [WIP] add svtr

* add svtr_base and svtr_tiny
2023-11-06 19:15:55 +08:00
sky
61f748de67
fix resolution problem for swin transformer and clip vit (#3021)
* Update foundation_vit.py

Update .gitignore

fix time cost problem

Update swin_transformer.py

fix the speed and memory problem

reduce the unnecessary calculation when patch matches resolution

fix conflict

remove check resolution function

Revert "fix conflict"

This reverts commit d7a7dade71ea5f9a6ac178e113dbcbd65d5a64b3.

fix conflict

remove the conflict checkpoint function

【Hackathon 5th No.69】 分类大模型--人体视觉任务SOLIDER (#2995)

* add_solider

* add_solider

* add_solider

* add_solider

* add_solider

* add_solider

* add_solider

* add_solider

* add_solider

* add_solider

* add_solider

update doc about PPHGNetV2 (#3002)

fix clip patch embedding resolution problem

support non 224 resolution

integrate the pading function to one

adjust function name

fix the resolution problem for clip-vision transformer part and swim transformer

fix the resolution problem for clip-vision transformer part and swim transformer

* fix cache problem

using the huggingface plan and drop the cache

* Revert "fix cache problem"

This reverts commit 8f7ab55c75d8c16443506a975c21967de8c038f2.

* fix resolution problem

* update big model backbone

* Revert "update big model backbone"

This reverts commit 04a39f701b62120bbfa20f6526a0589c33c21e47.
2023-10-31 10:11:46 +08:00
cuicheng01
60478c382a
remove PP-HGNetV2_B7 config (#3025) 2023-10-30 20:21:41 +08:00