The code implementation is very different to the paper.

[YangJae96]

In the paper, in the figure architecture, it uses same ODSSBlock in the backbone and PAFPN.

But in the code, the config file uses different block as illustrated below.

ultralytics/cfg/models/mamba-yolo/Mamba-YOLO-T.yaml

nc: 80  # number of classes
scales:   # [depth, width, max_channels]
  T: [0.33, 0.25, 1024]  #Mamba-YOLOv8-T summary: 6.1M parameters,   14.3GFLOPs


# Mamba-YOLO backbone 
backbone:
  # [from, repeats, module, args]
  - [-1, 1, SimpleStem, [128, 3]]   # 0-P2/4
  - [-1, 3, VSSBlock, [128]]               # 1
  - [-1, 1, VisionClueMerge, [256]]      # 2 p3/8
  - [-1, 3, VSSBlock, [256]]              # 3
  - [-1, 1, VisionClueMerge, [512]]      # 4 p4/16
  - [-1, 9, VSSBlock, [512]]              # 5
  - [-1, 1, VisionClueMerge, [1024]]      # 6 p5/32
  - [-1, 3, VSSBlock, [1024]]              # 7
  - [-1, 1, SPPF, [1024, 5]]               # 8

# Mamba-YOLO PAFPN
head:
  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [[-1, 5], 1, Concat, [1]]  # cat backbone P4
  - [-1, 3, XSSBlock, [512]]  # 11

  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [[-1, 3], 1, Concat, [1]]  # cat backbone P3
  - [-1, 3, XSSBlock, [256]]  # 14 (P3/8-small)

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 11], 1, Concat, [1]]  # cat head P4
  - [-1, 3, XSSBlock, [512]]  # 17 (P4/16-medium)

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 8], 1, Concat, [1]]  # cat head P5
  - [-1, 3, XSSBlock, [1024]]  # 20 (P5/32-large)

  - [[14, 17, 20], 1, Detect, [nc]]  # Detect(P3, P4, P5)

What is the difference between VSSBlock and XSSBlock?

Also, you mentioned using [3,6,6,3] number of repetitions of the ODSSBlock in the backbone is the best.
But there seems no match with the config file. I checked all config files in ultralytics/cfg/models/mamba-yolo. Tiny,Base and Large. They all use [3,3,9,3] as shown in above config file.

[zhang123-lf]

请教一下，如果用自己的数据集训练，需要更改代码中的哪些部分呢，现在一直在出错

[zhang123-lf]

大佬，请教一下，你在训练前做了怎样的一个修改，我的模型在训练的时候参数很好，但是在验证集上各项参数为0，这可能是什么问题呢