search

david8862 / keras-YOLOv3-model-set

end-to-end YOLOv4/v3/v2 object detection pipeline, implemented on tf.keras with different technologies
MIT License
640 stars 222 forks source link

canot train YOLOv3 with gpu #214

Open s07362022 opened 2 years ago

s07362022 commented 2 years ago

I use : cuda 11.2 cudnn 8

and pp install the requirement.txt from document, I just replace tensorflow-gpu to tensorflow-gpu==2.5.0

then, I cannot run the code.....

Traceback (most recent call last): File "train.py", line 20, in <module> from common.callbacks import EvalCallBack, CheckpointCleanCallBack, DatasetShuffleCallBack File "E:\workspace\DL\keras_YOLOv3_model_set-master\common\callbacks.py", line 15, in <module> from eval import eval_AP File "E:\workspace\DL\keras_YOLOv3_model_set-master\eval.py", line 19, in <module> import MNN File "C:\Users\User\AppData\Local\Programs\Python\Python38\lib\site-packages\MNN\__init__.py", line 7, in <module> from _mnncengine import * Exception: initMNN: init numpy failed

I try to install other tensorflow-gpu, but all fail. I want to know how to fix it.

or

what version is True by RTX3070

david8862 commented 2 years ago

seems to be a MNN env issue. what's your MNN & numpy version?