I’m a newb. Like I’m going to google most of the words you use. You’ve been warned.
Fell into utltralytics hub from some random link. Previously I’ve been using MS Customvision AI, but the random hoops to upload datasets annoyed me out of all proportion. The accuracy(?) is better here too.
I’m pushing up datasets and training it with “bring your own” on a windows 10 computer (3080TI/2080Super) When I start the training there is about 2 hours of this repeating before it ever seems to get to the training.
Transferred 349/349 items from epoch-39.pt
AMP: checks passed
AutoBatch: Computing optimal batch size for --imgsz 640
AutoBatch: CUDA:0 (NVIDIA GeForce RTX 3080 Ti) 12.00G total, 0.16G reserved, 0.01G allocated, 11.82G free
Params GFLOPs GPU_mem (GB) forward (ms) backward (ms) input output
1784212 4.284 0.365 29 13.33 (1, 3, 640, 640) list
1784212 8.569 0.403 18.67 12.33 (2, 3, 640, 640) list
1784212 17.14 0.646 17.67 18 (4, 3, 640, 640) list
1784212 34.28 0.956 83.33 15.33 (8, 3, 640, 640) list
1784212 68.55 1.711 303.3 80.67 (16, 3, 640, 640) list
AutoBatch: Using batch-size 101 for CUDA:0 9.58G/12.00G (80%)
optimizer: SGD(lr=0.01) with parameter groups 57 weight(decay=0.0), 60 weight(decay=0.0007890625), 60 bias
Resuming training from epoch-39.pt from epoch 40 to 100 total epochs
WARNING DP not recommended, use torch.distributed.run for best DDP Multi-GPU results.
See Multi-GPU Tutorial at https://github.com/ultralytics/yolov5/issues/475 to get started.
train: Scanning C:\code\textprepocesstestpython\datasets\WarzoneSummarySymbols (7)\WarzoneSummarySymbols\labels\train.c
train: 7.7GB RAM required, 1.0/63.9GB available, not caching images
Ultralytics:
An attempt has been made to start a new process before the
current process has finished its bootstrapping phase.
This probably means that you are not using fork to start your
child processes and you have forgotten to use the proper idiom
in the main module:
if __name__ == '__main__':
freeze_support()
...
The "freeze_support()" line can be omitted if the program
is not going to be frozen to produce an executable.
Any info on if this is the correct behavior or if there is away to speed it up would be appreciated.