Custom Weights Uploading without Uploading the Images

[lester1027]

I have already trained a yolov7-seg model with a custom dataset. The dataset is too big for the free Roboflow account. Is it possible to just upload the custom model weights without attaching them to the whole dataset? Actually, I do not understand the reason for attaching a set of weights to a dataset. Even without any existing dataset, the model inference can still be done. Thank you.

[Lenny]

Yes, you can upload your custom model weights to Roboflow. However, the weights need to be associated with a dataset version in the Roboflow platform. A “version” is a snapshot of your data, frozen in time, that you can reference at any point. Your weights will be connected to that dataset version, with APIs available for all versions of your dataset that you generate [source].

The reason for attaching a set of weights to a dataset is to provide a reference point for the model. This allows you to track the performance of your model over time and across different versions of your dataset. It also enables you to use Roboflow’s infrastructure and SDKs to deploy your model, reducing the coding, configuration, infrastructure, and developer operations work associated with deploying your model [source].

To upload your custom weights, you can use the model.deploy() method in the Python SDK. Here is an example of how to do it:

import roboflow
# Assuming you have already set up your project and version
version = project.version(1)
version.deploy("model-type", "path/to/training/results/")

The model types supported for upload include yolov7-seg [source].

Please note that model weights uploads are only available for dataset versions without a trained model (i.e not yet trained with Roboflow Train, and no YOLOv5, YOLOv7 or YOLOv8 model weights uploaded) [source]. If weights upload is attempted for dataset versions that already have a trained model, you will receive a RuntimeError [source].

---

Hi! I’m Roboflow’s AI powered bot. This is an automated response powered using AI with knowledge from our docs, blog, and forum. If this solves your issue, please mark it as the solution using the checkbox icon. If this doesn’t solve your issue, please follow up with what you’ve tried/why, and the community will continue to chime in as usual.