PaddleX/latest/pipeline_usage/instructions/pipeline_CLI_usage

[giscus[bot]]

PaddleX/latest/pipeline_usage/instructions/pipeline_CLI_usage

https://paddlepaddle.github.io/PaddleX/latest/pipeline_usage/instructions/pipeline_CLI_usage.html

[songmingjun3]

按照文档使用自定义的配置文件进行模型产线推理测试报错
(PaddleX) PS D:\data\PycharmWorkspace\PaddleX> paddlex --pipeline configs/image_classification.yaml --input https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_image_classification_001.jpg --save_path ./output/
Unsupported pipeline: configs/image_classification.yaml, CLI predict only supports these pipelines: OCR, object_detection, image_classification, image_multilabel_classification, pedestrian_attribute_recognition, vehicl
e_attribute_recognition, human_keypoint_detection, table_recognition, table_recognition_v2, seal_recognition, layout_parsing, PP-StructureV3, ts_forecast, ts_anomaly_detection, ts_classification, formula_recognition, i
nstance_segmentation, semantic_segmentation, small_object_detection, anomaly_detection, video_classification, video_detection, doc_preprocessor, rotated_object_detection, open_vocabulary_detection, open_vocabulary_segmentation, 3d_bev_detection, multilingual_speech_recognition

[songmingjun3]

解决了，是configs目录不存在导致直接保存在当前目录，使用配置的文件不对导致的，不过报错看起来不是很友好

[sfeng1996]

请问能否自定义产品线，该产品线包含多个不同类别的功能模块，包含 ocr、图像分类、音频、视频

[sfeng1996]

我们这边有这样的需求，需要部署一个服务能够提供 http api，该服务能够包含 paddle 多个模型，例如包含 ocr 相关的、视频、音频功能，请问这种能实现吗？