-
Notifications
You must be signed in to change notification settings - Fork 5.9k
Description
问题描述 Issue Description
Hello.
I am trying to use HPI on paddlex as given in https://www.paddleocr.ai/latest/en/version3.x/deployment/serving.html using the --use_hpi flag.
To setup HPI I have
paddlex --install serving
paddleocr install_hpi_deps gpu
Installed CUDA 12.6 and CUDnn 9.5 as according to my OS specifications (Linux x86_64 Ubuntu 24.04 deb(local)
installed tensorrt via pip install tensorrt
The GPU version paddlepaddle via python3 -m pip install paddlepaddle-gpu==3.2.0 -i https://www.paddlepaddle.org.cn/packages/stable/cu126/
I have a config file as below
SubModules:
TextDetection:
use_hpip=True
box_thresh: 0.6
limit_side_len: 64
limit_type: min
max_side_limit: 4000
model_dir: PP-OCRv5_server_det #or null
model_name: PP-OCRv5_server_det
module_name: text_detection
thresh: 0.3
unclip_ratio: 1.5
hpi_config:
backend: onnxruntime
TextLineOrientation:
batch_size: 6
model_dir: null
model_name: null
module_name: textline_orientation
TextRecognition:
use_hpip=True
batch_size: 6
model_dir: PP-OCRv5_server_rec #or null
model_name: PP-OCRv5_server_rec
module_name: text_recognition
hpi_config:
backend: tensorrt. #or onnxruntime
score_thresh: 0.0
SubPipelines:
DocPreprocessor:
SubModules:
DocOrientationClassify:
model_dir: null
model_name: null
module_name: doc_text_orientation
DocUnwarping:
model_dir: null
model_name: null
module_name: image_unwarping
pipeline_name: doc_preprocessor
use_doc_orientation_classify: false
use_doc_unwarping: false
pipeline_name: OCR
text_type: general
use_doc_preprocessor: false
use_textline_orientation: false
return_word_box: true
The default server paddlex --serve --pipeline ocr_config.yaml works fine but when I enable hpi
paddlex --serve --pipeline ocr_config.yaml --use_hpi I get
RuntimeError: No inference backend and configuration could be suggested. Reason: (3, 2, 0, None) is not a supported Paddle version. The same error comes for paddle paddle version 2.6.0 as well
(Sorry I do not have the full stack trace, it was on a cloud machine. I can run again and give the full trace if necessary)
Can anyone please help me?? @jzhang533
Thank you
版本&环境信息 Version & Environment Information
I do not have all the system info saved. If someone can look into the problem I am happy to setup everything again and send the information !!