mindyolo
mindyolo copied to clipboard
mindspore-lab
MindSpore YOLO series toolbox and benchmark
我用的命令：python test.py --config ./configs/yolov5/yolov5n.yaml --weight ./yolov5n.ckpt \ --device_target Ascend 定位在这一行： File "/home/ma-user/work/mindyolo/mindyolo/data/dataset.py", line 1183, in test_collate_fn hw_scale = [sample.pop('hw_scale') for sample in batch_samples] 日志如下： Traceback (most recent call last): File...
[2024-01-04 09:05:14 +0800] [46] [INFO] Starting gunicorn 20.1.0 [2024-01-04 09:05:14 +0800] [46] [INFO] Listening at: http://172.16.0.78:8443 (46) [2024-01-04 09:05:14 +0800] [46] [INFO] Using worker: gthread [2024-01-04 09:05:14 +0800] [49] [INFO]...
采用yolov-tiny训练BDD100K数据集可以正常训练，但是由于模型太小，最后验证集精度比较低，IoU=0.50:0.95 在0.29-0.30附近，想用大点的模型重新训练。 但是在数据集相同的条件下，换用yolov7l直接损失为nan；换用yolov8l损失一直不下降。 请问我该采用什么策略，才能使用yolov7l或者yolov8l正常训练。 训练环境为GPU mindspore==2.0.0 如果需要调节学习率，请帮忙解释一下下面的学习率参数的意义，以及该如何调节？（以yolov7l为例） optimizer: optimizer: momentum lr_init: 0.01 # initial learning rate (SGD=1E-2, Adam=1E-3) momentum: 0.937 # SGD momentum/Adam beta1 nesterov: True # update gradients with...
yolov5
运行python train.py --config ./configs/yolov5/yolov5n.yaml显示package not installed, albumentations load failed，但是已经下载了这个包，这样要怎么办
训练模型报错，运行python train.py --config ./configs/yolov7/yolov7.yaml ``` Exception: Error loading data from ./coco/train2017.txt: coco/train2017.txt does not exist ```
I am trying to train my model on my computer with RTX3060. However, the training process was killed frequently. Would you mind help me with that issue  ?
您好，麻烦帮看下这个问题  RuntimeError: The device address type is wrong: type name in address:CPU, type name in context:Ascend CANN：7.0.RC1 MindSpore2.1.0 Ascend310
loading annotations into memory... Done (t=2.84s) creating index... index created! 2023-11-22 21:21:50,302 [INFO] Dataset Cache file hash/version check success. 2023-11-22 21:21:50,302 [INFO] Load dataset cache from [small_coco\train2017.cache.npy] success. Scanning 'small_coco\train2017.cache.npy'...
 在win64环境下推理yolo5报错ImportError: cannot import name 'process_mask_upsample' from 'mindyolo.utils.metrics' 环境： mindspore：2.1 python：3.7.10 操作系统：win11 执行的命令： `python demo/predict.py --config ./configs/yolov5/yolov5s.yaml --weight=/path_to_ckpt/yolov5s_300e_mAP376-860bcf3b.ckpt --image_path /path_to_image/dog.jpg --device_target=CPU`
