put a new part in the rtdetr-l model

[sichengjiang28-ui]

TypeError: conv2d() received an invalid combination of arguments - got (dict, Parameter, NoneType, tuple, tuple, tuple, int), but expected one of:

• (Tensor input, Tensor weight, Tensor bias, tuple of ints stride, tuple of ints padding, tuple of ints dilation, int groups)
  didn't match because some of the arguments have invalid types: (!dict!, !Parameter!, !NoneType!, !tuple of (int, int)!, !tuple of (int, int)!, !tuple of (int, int)!, int)
• (Tensor input, Tensor weight, Tensor bias, tuple of ints stride, str padding, tuple of ints dilation, int groups)
  didn't match because some of the arguments have invalid types: (!dict!, !Parameter!, !NoneType!, !tuple of (int, int)!, !tuple of (int, int)!, !tuple of (int, int)!, int)

[UltralyticsAssistant]

👋 Hello @sichengjiang28-ui, thank you for your interest in Ultralytics 🚀! This is an automated response 🤖, and an Ultralytics engineer will also assist you soon. In the meantime, we recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

It looks like you’re modifying the RT-DETR-L model and hitting a conv2d() type error where the input is a dict instead of a Tensor. To help us pinpoint exactly where that dict is coming from (e.g., a custom module forward output, a dataloader batch, or a model head interface mismatch), we’ll need a bit more detail 👀

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it. Ideally include:
-Your exact model/config change (what “new part” you added and where)
-A small runnable snippet that triggers the error
-The full traceback and your ultralytics + PyTorch versions

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the Ultralytics community where it suits you best. For real-time chat, head to Discord 🎧. Prefer in-depth discussions? Check out Discourse. Or dive into threads on our Subreddit to share knowledge with the community.

Upgrade

Upgrade to the latest ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8 to verify your issue is not already resolved in the latest version:

pip install -U ultralytics

Environments

YOLO may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLO Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

That conv2d() error means a nn.Conv2d layer is being fed a dict instead of a torch.Tensor—this usually happens when a newly inserted module returns a dict (or when the training batch dict is accidentally forwarded into the model), and in RT-DETR only the decoder takes batch as an optional arg while conv blocks require tensors (see RTDETRDecoder.forward signature and Conv.forward); please paste the full traceback (with the exact file/line where conv2d is called) plus yolo checks, and as a quick isolate add this right before the failing conv to find where the dict first appears:

assert not isinstance(x, dict), type(x)