CUDA 11.6 + CUDNN 8.4 运行OCR 返回的Result为None，另外一台纯CPU模式同一张图是可以的

shaoerkuai commented 8 months ago

请提供下述完整信息以便快速定位问题/Please provide the following information to quickly locate the problem

系统环境/System Environment：
Windows 11
版本号/Version：Paddle： PaddleOCR：问题相关组件/Related components：
运行指令/Command Code：参考的是官方的Python代码

ocr = PaddleOCR(use_angle_cls=True, lang="ch") # need to run only once to download and load model into memory img_path = "D:\img\20231122140525.jpg" t1 = time.time() print("21321") result = ocr.ocr(img_path, det=True, cls=True) # 这里Result就是None了 for idx in range(len(result)): res = result[idx] for line in res:
print(line)

完整报错/Complete Error Message：

[2023/11/22 17:06:25] ppocr DEBUG: Namespace(help='==SUPPRESS==', use_gpu=True, use_xpu=False, use_npu=False, ir_opti
m=True, use_tensorrt=False, min_subgraph_size=15, precision='fp32', gpu_mem=500, gpu_id=0, image_dir=None, page_num=0
, det_algorithm='DB', det_model_dir='C:\\Users\\XXXXX/.paddleocr/whl\\det\\ch\\ch_PP-OCRv4_det_infer', det_limit_
side_len=960, det_limit_type='max', det_box_type='quad', det_db_thresh=0.3, det_db_box_thresh=0.6, det_db_unclip_rati
o=1.5, max_batch_size=10, use_dilation=False, det_db_score_mode='fast', det_east_score_thresh=0.8, det_east_cover_thr
esh=0.1, det_east_nms_thresh=0.2, det_sast_score_thresh=0.5, det_sast_nms_thresh=0.2, det_pse_thresh=0, det_pse_box_t
hresh=0.85, det_pse_min_area=16, det_pse_scale=1, scales=[8, 16, 32], alpha=1.0, beta=1.0, fourier_degree=5, rec_algo
rithm='SVTR_LCNet', rec_model_dir='C:\\Users\\XXX/.paddleocr/whl\\rec\\ch\\ch_PP-OCRv4_rec_infer', rec_image_in
verse=True, rec_image_shape='3, 48, 320', rec_batch_num=6, max_text_length=25, rec_char_dict_path='D:\\Python310\\lib
\\site-packages\\paddleocr\\ppocr\\utils\\ppocr_keys_v1.txt', use_space_char=True, vis_font_path='./doc/fonts/simfang
.ttf', drop_score=0.5, e2e_algorithm='PGNet', e2e_model_dir=None, e2e_limit_side_len=768, e2e_limit_type='max', e2e_p
gnet_score_thresh=0.5, e2e_char_dict_path='./ppocr/utils/ic15_dict.txt', e2e_pgnet_valid_set='totaltext', e2e_pgnet_m
ode='fast', use_angle_cls=True, cls_model_dir='C:\\Users\\XXX/.paddleocr/whl\\cls\\ch_ppocr_mobile_v2.0_cls_inf
er', cls_image_shape='3, 48, 192', label_list=['0', '180'], cls_batch_num=6, cls_thresh=0.9, enable_mkldnn=False, cpu
_threads=10, use_pdserving=False, warmup=False, sr_model_dir=None, sr_image_shape='3, 32, 128', sr_batch_num=1, draw_
img_save_dir='./inference_results', save_crop_res=False, crop_res_save_dir='./output', use_mp=False, total_process_nu
m=1, process_id=0, benchmark=False, save_log_path='./log_output/', show_log=True, use_onnx=False, output='./output',
table_max_len=488, table_algorithm='TableAttn', table_model_dir=None, merge_no_span_structure=True, table_char_dict_p
ath=None, layout_model_dir=None, layout_dict_path=None, layout_score_threshold=0.5, layout_nms_threshold=0.5, kie_alg
orithm='LayoutXLM', ser_model_dir=None, re_model_dir=None, use_visual_backbone=True, ser_dict_path='../train_data/XFU
ND/class_list_xfun.txt', ocr_order_method=None, mode='structure', image_orientation=False, layout=True, table=True, o
cr=True, recovery=False, use_pdf2docx_api=False, invert=False, binarize=False, alphacolor=(255, 255, 255), lang='ch',
det=True, rec=True, type='ocr', ocr_version='PP-OCRv4', structure_version='PP-StructureV2')
21321
[2023/11/22 17:06:34] ppocr DEBUG: dt_boxes num : 0, elapsed : 1.3370857238769531
[2023/11/22 17:06:34] ppocr DEBUG: cls num  : 0, elapsed : 0
[2023/11/22 17:06:34] ppocr DEBUG: rec_res num  : 0, elapsed : 0.0
Traceback (most recent call last):
File "E:\XXX\paddletest\a.py", line 25, in <module>
for line in res:
TypeError: 'NoneType' object is not iterable

我们提供了AceIssueSolver来帮助你解答问题，你是否想要它来解答(请填写yes/no)?/We provide AceIssueSolver to solve issues, do you want it? (Please write yes/no):

no

请尽量不要包含图片在问题中/Please try to not include the image in the issue.

shaoerkuai commented 8 months ago

当前版本： paddleocr 2.7.0.3 paddlepaddle-gpu 2.5.2.post116 cudnn 8.4.0.27_cuda11.6 cuda_11.6.1_511.65 驱动支持的CUDA 11.6.127 设置gpu = false 也是可以正常识别的，开启GPU后就不支持。

shaoerkuai commented 8 months ago

试下paddle 的检查： The GPU architecture in your current machine is Maxwell, which is not compatible with Paddle installation with arch: 70 75 80 86 , it is recommended to install the corresponding wheel package according to the installation information on the official Paddle website.

应该是Maxwell架构不支持，更换了老版本后，出现如下情况： [[[984.0, 738.0], [1006.0, 738.0], [1006.0, 752.0], [984.0, 752.0]], ('茂团瞬潜阅予彭任~授仲z充鸟驱俗J’潜俗劣士号画乡傻翼√机谁筋′灭驾j妥剩_腰显', 0.997814953327179)] [[[1132.0, 736.0], [1154.0, 736.0], [1154.0, 752.0], [1132.0, 752.0]], ('乱丫傅丫管富管丫字富管叫喘ｌ锦胃丫μ胃隐', 0 .803952693939209)] [[[1234.0, 736.0], [1274.0, 736.0], [1274.0, 750.0], [1234.0, 750.0]], ('婚亏丑痧授充妇东於团逾仁钜搁佛圳>遵劵滚旅坠識．皆?临話蜂错仁圣措充μ遒丶应租', 0.9999554753303528)] [[[1866.0, 738.0], [1876.0, 738.0], [1876.0, 750.0], [1866.0, 750.0]], ("衍j—奢录承剩吉满陆w吉呈‘谁仁磊号仁归响歡滚虽允喷越贷幽'惟贷遵眼损展°团饿", 0.9942718148231506)] [[[1877.0, 733.0], [1879.0, 753.0], [1888.0, 744.0], [1879.0, 753.0]], ('诺丫傅丫管富管丫字富管叫喘ｌ锦胃丫μ胃隐', 0 .8039463758468628)] [[[32.0, 761.0], [32.0, 781.0], [42.0, 771.0], [32.0, 781.0]], ('方含洗埔²净才晟冶何什豫²介何爷丫胖丫什丫董丫币爷還■ 繁', 0.8569181561470032)] [[[40.0, 763.0], [74.0, 763.0], [74.0, 777.0], [40.0, 777.0]], ('茂团瞬潜阅予彭任~授仲z充鸟驱俗J’潜俗劣士号画乡傻翼 √机谁筋′灭驾j妥剩_腰显', 0.997814953327179)] [[[44.0, 757.0], [54.0, 748.0], [64.0, 757.0], [54.0, 767.0]], ('婚亏丑痧授充妇东於团逾仁钜搁佛圳>遵劵滚旅坠識．皆?临話蜂错仁圣措充μ遒丶应租', 0.9999554753303528)]

全是乱识别的乱码...

shaoerkuai commented 8 months ago

解决方案： GTX 9系列使用以下CUDA版本正常，希望能够解决，当前可以通过降低CUDA版本解决，不过后续升级WIN11，CUDA 10.2就用不了了。当前WIN10还能正常使用。驱动不需要降级，只需要卸载重新安装对应版本的CUDA即可。 CUDA v10.2 CUDNN v7.6.5.32

PaddlePaddle / PaddleOCR

CUDA 11.6 + CUDNN 8.4 运行OCR 返回的Result为None，另外一台纯CPU模式同一张图是可以的 #11289