5e814882e3
change python3 -> python, otherwise < python3.6 is used.
2022-11-15 16:31:48 +08:00
efa7d3035f
make loss deterministic while running on gpu
2022-11-15 10:58:46 +08:00
5ad2817470
add statuscheck and follow the tipc_standard
2022-11-14 15:25:58 +08:00
c9f8b87ef4
fix by code review
2022-11-14 15:25:58 +08:00
0f6d71f021
[TIPC-Dy2St] add correctness tipc script
2022-11-14 15:25:58 +08:00
b033661bcf
tipc: VAN_tiny -> VAN_B0
2022-10-20 18:49:44 +08:00
cde853bfa7
Polish benchmark_train.sh for dynamicTostatic
2022-10-19 14:55:32 +08:00
d7014129f1
tipc: set DataLoader.Train.sampler.shuffle=False
...
benchmark needs the field to set Global.print_batch_step=1. Check file "./test_tipc/benchmark_train.sh#L76" for details.
2022-10-17 11:24:57 +08:00
4438acfda8
Merge pull request #2361 from TingquanGao/dev/tipc_fix_log_path
...
turn off the mkldnn and trt in ConvNeXt infer
2022-09-29 14:07:24 +08:00
9f64ff6c1c
tipc: turn off mkldnn and trt
...
it will time out when turn on trt or mkldnn
2022-09-29 05:22:44 +00:00
be434dc66c
tipc: fix the inference model name
2022-09-29 11:16:22 +08:00
b55791f696
tipc: styling code
2022-09-28 13:33:19 +00:00
091cb607f6
fix tipc log
2022-09-23 18:53:35 +08:00
8897aba3d0
fix tipc ptq log
2022-09-23 18:53:35 +08:00
69b53bc95f
tipc: modify the statistical range of log
...
the ips is unstable in the first 50 iter when training lite modle.
2022-09-21 13:58:00 +08:00
983d512a45
tipc: fix the keyword of ShiTu model
...
In benchmark_train mode, the config name is benchmark.txt. So the keyword of ShiTu should be model_name instead of the config name.
2022-09-19 14:30:54 +08:00
42758f5463
fix benchmark train's config for GeneralRecognitionV2, and polish TIPC docs
2022-09-14 08:54:22 +00:00
dafcb5cddf
tipc: fix log path for GeneralRecognition
2022-09-14 10:44:29 +08:00
b1c2350a26
[TIPC] add more scripts change, test=develop
2022-09-13 15:03:14 +08:00
ec21470a35
Merge pull request #2258 from TingquanGao/dev/tipc_fix_log_path
...
tipc: fix log path
2022-09-10 00:06:54 +08:00
7e239952a8
tipc: set num_workers to 4
...
The benchmark requires threads num to be set 4 when DALI is used for best performance.
And the param num_workers is not worked before commit 80c10bb995
2022-09-09 10:10:18 +00:00
0d9ddc0812
[TIPC] add scripts for npu and xpu, test=develop
2022-09-09 02:39:48 +08:00
0cea550455
tipc: fix log path
2022-09-08 10:57:03 +00:00
2700beea82
tipc: fix log path
2022-09-08 08:15:33 +00:00
fb669a05b8
tipc: fix log path
2022-09-06 13:25:09 +00:00
b3d26a987d
tipc: fix log path
2022-09-05 15:57:49 +00:00
44b55fb67c
tipc: paddle2onnx --enable_dev_version=False
2022-09-05 19:04:20 +08:00
4ee9dd49e0
fix small bug in prepare.sh
2022-09-04 14:22:44 +00:00
c449ba3420
add fleet chain and update inference_rec.yaml and corresponding output for feature_extraction.md
2022-09-04 14:20:07 +00:00
4620cfb9bc
fix typo in config files
2022-09-04 13:52:58 +00:00
94b48ac568
add cpp serving & cpp infer chain, update TIPC docs
2022-09-04 13:42:53 +00:00
38e78b0c81
add paddle2onnx_infer chain
2022-09-01 03:12:11 +00:00
a8278e84c4
add pact chain
2022-09-01 03:01:51 +00:00
6adee2720f
add ptq infer
2022-09-01 02:52:24 +00:00
d2ffc5e408
add lite_train_lite_infer chain
2022-08-31 17:28:48 +00:00
2f5c0c71da
fix: replace ResNet50.yaml with ResNet50_amp_O2_ultra.yaml when AMPO2
...
It is needed that use dali, NHWC and 4 channels when training ResNet50 with AMPO2
2022-08-26 10:59:16 +08:00
ea66adbec1
fix: need install paddleclas whl
2022-08-26 10:59:16 +08:00
5810695672
tipc: set num_wokers=12
2022-08-05 14:34:14 +08:00
b77a431451
tipc: set num workers=12 iff fp16
2022-08-05 14:34:14 +08:00
81b42f627b
tipc: rm useless bs
2022-08-05 14:34:14 +08:00
430fbd5df9
tipc: support fp32/fp16 training benchmark of PP-Shitu
2022-08-05 14:34:14 +08:00
a7f124042b
mv
2022-08-05 14:34:14 +08:00
483c0383e5
tipc: set bs=256 to avoid oom
2022-08-05 14:34:14 +08:00
b9de09e547
tipc: set num_wokers=12
2022-08-05 14:34:14 +08:00
7990fe5767
tipc: set num_wokers=8
2022-08-05 14:34:14 +08:00
3d1fb324a3
tipc: support AMP O2
2022-08-05 14:34:14 +08:00
595a6c5bcf
format code style
2022-08-05 14:34:14 +08:00
ac1c206d4a
tipc: support benchmark with fp32/fp16 training
2022-08-05 14:34:14 +08:00
b1f0d6d083
Merge pull request #1908 from flytocc/PeleeNet_PR
...
add PeleeNet
2022-07-14 17:50:54 +08:00
c5b5b34306
Merge pull request #2151 from RainFrost1/develop
...
修复benchmark中bs过大时导致log过少的问题
2022-07-14 14:21:40 +08:00
xiongkun
gaotingquan
0x45f
cuicheng01
zhengya01
HydrogenSulfate
qili93
root
Walter