[人工智能] mmaction工程解读-timesformer模型架构

开发: C++知识库 Java知识库 JavaScript Python PHP知识库人工智能区块链大数据移动开发嵌入式开发工具数据结构与算法开发测试游戏开发网络协议系统运维
教程: HTML教程 CSS教程 JavaScript教程 Go语言教程 JQuery教程 VUE教程 VUE3教程 Bootstrap教程 SQL数据库教程 C语言教程 C++教程 Java教程 Python教程 Python3教程 C#教程
数码: 电脑笔记本显卡显示器固态硬盘硬盘耳机手机 iphone vivo oppo 小米华为单反装机图拉丁
-> 人工智能 -> mmaction工程解读-timesformer模型架构 -> 正文阅读
[人工智能]mmaction工程解读-timesformer模型架构

Recognizer3D(
  (backbone): TimeSformer(
    (patch_embed): PatchEmbed(
      (projection): Conv2d(3, 768, kernel_size=(16, 16), stride=(16, 16))
    )
    (drop_after_pos): Dropout(p=0.0, inplace=False)
    (drop_after_time): Dropout(p=0.0, inplace=False)
    (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
    (transformer_layers): TransformerLayerSequence(
      (layers): ModuleList(
        (0): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (1): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (2): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (3): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (4): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (5): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (6): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (7): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (8): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (9): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (10): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
        (11): BaseTransformerLayer(
          (attentions): ModuleList(
            (0): DividedTemporalAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
              (temporal_fc): Linear(in_features=768, out_features=768, bias=True)
            )
            (1): DividedSpatialAttentionWithNorm(
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
              (attn): MultiheadAttention(
                (out_proj): _LinearWithBias(in_features=768, out_features=768, bias=True)
              )
              (proj_drop): Dropout(p=0.0, inplace=False)
              (dropout_layer): DropPath()
            )
          )
          (ffns): ModuleList(
            (0): FFNWithNorm(
              (activate): GELU()
              (layers): Sequential(
                (0): Sequential(
                  (0): Linear(in_features=768, out_features=3072, bias=True)
                  (1): GELU()
                  (2): Dropout(p=0.0, inplace=False)
                )
                (1): Linear(in_features=3072, out_features=768, bias=True)
                (2): Dropout(p=0.0, inplace=False)
              )
              (dropout_layer): DropPath()
              (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
            )
          )
          (norms): ModuleList()
        )
      )
    )
  )
  (cls_head): TimeSformerHead(
    (loss_cls): CrossEntropyLoss()
    (fc_cls): Linear(in_features=768, out_features=400, bias=True)
  )
)