Add nn.SiLU inplace in attempt_load()

[xlorne]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancement of activation function compatibility in model loading.

📊 Key Changes

• 🚀 Added nn.SiLU to the list of activation functions setting inplace to True during model loading.

🎯 Purpose & Impact

• 💡 Ensures compatibility with PyTorch 1.7.0, allowing models with SiLU activation functions to be loaded without issues.
• 🏃‍♂️ Potentially improves in-place operation performance, reducing memory overhead.
• 👥 Users employing the SiLU activation will experience smoother integrations and upgrades.

[github-actions]

👋 Hello @1991wangliang, thank you for submitting a 🚀 PR! To allow your work to be integrated as seamlessly as possible, we advise you to:

• ✅ Verify your PR is up-to-date with origin/master. If your PR is behind origin/master update by running the following, replacing 'feature' with the name of your local branch:

git remote add upstream https://github.com/ultralytics/yolov5.git
git fetch upstream
git checkout feature  # <----- replace 'feature' with local branch name
git rebase upstream/master
git push -u origin -f

• ✅ Verify all Continuous Integration (CI) checks are passing.
• ✅ Reduce changes to the absolute minimum required for your bug fix or feature addition. "It is not daily increase but daily decrease, hack away the unessential. The closer to the source, the less wastage there is." -Bruce Lee

[glenn-jocher]

@1991wangliang hi there. Could you explain the purpose behind the sleep call?

[xlorne]

perhaps the CPU performance is too poor, or it may be blocked under multi-threading. It will get stuck when loading data.

    # DDP mode
    if cuda and rank != -1:
        model = DDP(model, device_ids=[opt.local_rank], output_device=opt.local_rank)

    print("sleep 3 sec to load data .")
    time.sleep(3)

    # Trainloader
    dataloader, dataset = create_dataloader(train_path, imgsz, batch_size, gs, opt,
                                            hyp=hyp, augment=True, cache=opt.cache_images, rect=opt.rect, rank=rank,
                                            world_size=opt.world_size, workers=opt.workers,
                                            image_weights=opt.image_weights, quad=opt.quad, prefix=colorstr('train: '))
    mlc = np.concatenate(dataset.labels, 0)[:, 0].max()  # max label class

before create_dataloader ,sleep some sec can be running.

[glenn-jocher]

@1991wangliang I'm not sure if the sleep term brings any benefits. Your system may be low on resources. Perhaps you could try training with less --workers if you are running out of threads.

I've removed the sleep term and added a SiLU inplace change to this PR in it's place.

[xlorne]

thanks, i known why can be locked . when train.py execute once ,that can be created labels.cache file on train data path . that train.py will be lock in execute load data. i remove that can be run.