yolov7
yolov7 copied to clipboard
WongKinYiu
Error While Training with Hyperparameter Evolution
Hello,
I was trying to train a model with hyperparameter evolution, but the error occured. Did I do any thing wrong?
Below was the error I met and the command I used in terminal.
I just sent a pull request to fix this error. The problem is that the 'anchor' parameter is missing from your hyp.yaml file you can add it manually : anchors:3 #3 is a good default value Also you'll run to another error because the train.py file is also missing the 'paste_in' and 'copy_paste' values from the metadata variable called 'meta' line 614
@Dhiaeddine-Oussayed thanks for your helping. I added anchor in the hyp.yaml and the problem solved. However, I met a new error. The error is at below. How can I solve it?
@Galafala did you add the 'paste_in' and 'copy_paste' to the train.py ?
@Galafala if you didn't you can just replace this in your train.py at line 614:
meta = {'lr0': (1, 1e-5, 1e-1), # initial learning rate (SGD=1E-2, Adam=1E-3)
'lrf': (1, 0.01, 1.0), # final OneCycleLR learning rate (lr0 * lrf)
'momentum': (0.3, 0.6, 0.98), # SGD momentum/Adam beta1
'weight_decay': (1, 0.0, 0.001), # optimizer weight decay
'warmup_epochs': (1, 0.0, 5.0), # warmup epochs (fractions ok)
'warmup_momentum': (1, 0.0, 0.95), # warmup initial momentum
'warmup_bias_lr': (1, 0.0, 0.2), # warmup initial bias lr
'box': (1, 0.02, 0.2), # box loss gain
'cls': (1, 0.2, 4.0), # cls loss gain
'cls_pw': (1, 0.5, 2.0), # cls BCELoss positive_weight
'obj': (1, 0.2, 4.0), # obj loss gain (scale with pixels)
'obj_pw': (1, 0.5, 2.0), # obj BCELoss positive_weight
'iou_t': (0, 0.1, 0.7), # IoU training threshold
'anchor_t': (1, 2.0, 8.0), # anchor-multiple threshold
'anchors': (2, 2.0, 10.0), # anchors per output grid (0 to ignore)
'fl_gamma': (0, 0.0, 2.0), # focal loss gamma (efficientDet default gamma=1.5)
'hsv_h': (1, 0.0, 0.1), # image HSV-Hue augmentation (fraction)
'hsv_s': (1, 0.0, 0.9), # image HSV-Saturation augmentation (fraction)
'hsv_v': (1, 0.0, 0.9), # image HSV-Value augmentation (fraction)
'degrees': (1, 0.0, 45.0), # image rotation (+/- deg)
'translate': (1, 0.0, 0.9), # image translation (+/- fraction)
'scale': (1, 0.0, 0.9), # image scale (+/- gain)
'shear': (1, 0.0, 10.0), # image shear (+/- deg)
'perspective': (0, 0.0, 0.001), # image perspective (+/- fraction), range 0-0.001
'flipud': (1, 0.0, 1.0), # image flip up-down (probability)
'fliplr': (0, 0.0, 1.0), # image flip left-right (probability)
'mosaic': (1, 0.0, 1.0), # image mixup (probability)
'mixup': (1, 0.0, 1.0), # image mixup (probability)
'copy_paste': (1, 0.0, 1.0), # segment copy-paste (probability)
'paste_in': (1, 0.0, 1.0)} # segment copy-paste (probability)
@Dhiaeddine-Oussayed thanks for your response. The promblem still exist. I noticed that the problem raised after the anchors=3 was added to the hyp.yaml. And the problem happened at both without evolution training and evolution training.
@Dhiaeddine-Oussayed This is the zip file of my train_azu.py. Should I past the code or the zip file is fine?
@Galafala My pull request have been merged try pulling master and test again
@Dhiaeddine-Oussayed , I have pull the latest file and trained a mdoel with evolution. The train.py is working, but train_aux.py still raised errors. Below are the errors.
did you chnage the "nc" : number of classes in your config file ?
the len of parameters in train.py-->mata should equal to the len of parameters in hpy.scatch.p5.yaml
@Dhiaeddine-Oussayed Yes, I did.
Hi, I met exactly the same problem. Have you solved this problem? Thanks!!
