 Lin Manhui
|
3685fd2843
[Fix] Let HPIP not automatically choose pptrt when it is not available (#3782)
|
hai 7 meses |
|
Lin Manhui
|
f4a679ff9f
[Feat] Support lazy imports (#3764)
|
hai 7 meses |
 a31413510
|
12ec01f6b8
upload npu ultra_infer tutorials (#3763)
|
hai 7 meses |
|
Lin Manhui
|
bc54f1bb90
[Infra] Prune and sort imports (#3694)
|
hai 7 meses |
|
Lin Manhui
|
a65d269b21
[Feat] Support multi-backend inference (#2860)
|
hai 7 meses |
 zhink
|
80cb6da064
fix Paddle-TRT some model within FP16
|
hai 8 meses |
 gaotingquan
|
c5c819d47b
support to disable trt half ops
|
hai 8 meses |
|
gaotingquan
|
696089a110
call enable_use_gpu() first to use TensorRT engine
|
hai 8 meses |
|
gaotingquan
|
976eeb533f
when pir trt, it's need that enabling use gpu in convert_trt
|
hai 8 meses |
 EnflameGCU
|
0599a222ed
[GCU] Update gcu config (#3605)
|
hai 8 meses |
|
gaotingquan
|
19863acc7f
bugfix
|
hai 8 meses |
|
gaotingquan
|
db1bbfe948
support to set more trt config by model
|
hai 8 meses |
|
gaotingquan
|
d3dc02cb7a
support to set trt config by model
|
hai 8 meses |
|
gaotingquan
|
774737c06b
support paddle_fp32 and paddle_fp16 mode
|
hai 8 meses |
 zhang-prog
|
eb3e60685b
add cache func (#3509)
|
hai 8 meses |
|
gaotingquan
|
5abc0bfdca
set trt config for PP-YOLOE_seg-S
|
hai 8 meses |
|
gaotingquan
|
bf87804450
support trt_dynamic_shape_input_data for pir trt
|
hai 8 meses |
|
gaotingquan
|
0742a15ef8
Manually delete the predictor to trigger its destructor, ensuring that the shape_range_info file would be saved
|
hai 8 meses |
|
Lin Manhui
|
bea5ad68f9
[Fix] Allow bypassing model white list checks when performing internal tests (#3499)
|
hai 8 meses |
|
gaotingquan
|
ded6b3ab2b
bugfix: call enable_use_gpu() before calling enable_tensorrt_engine()
|
hai 8 meses |
|
Lin Manhui
|
8412aa82cb
Circumvent paddle inference bug (#3490)
|
hai 8 meses |
|
zhang-prog
|
3c9400e6dd
Enhance benchmark (#3475)
|
hai 8 meses |
|
zhang-prog
|
5158085f23
[Feat] upgrade benchmark (#3416)
|
hai 8 meses |
|
gaotingquan
|
9a4a934d47
support set trt config by model
|
hai 8 meses |
|
a31413510
|
7b5288bd6f
predict
|
hai 8 meses |
|
gaotingquan
|
68f3ff5b6f
fix redundantly create predictor
|
hai 9 meses |
|
gaotingquan
|
f48f8e4fdf
call enable_gpu() on dcu device
|
hai 9 meses |
|
gaotingquan
|
a5b997876d
use dcu device type
|
hai 9 meses |
|
gaotingquan
|
fcbf6b11d2
dcu also use gpu as the flag
|
hai 9 meses |
|
gaotingquan
|
8e9b8ddca2
change workspace_size to 30 in trt infer to be compatible with more models
|
hai 10 meses |
