BatchedMatMul
Input
- x1:
- Required: yes
- Type: fp16
- Description: tensor of 3–8 dimensions. If adj_x1 = False, x1 has shape [batch, ..., M, K]. If adj_x1 = True, x1 has shape [batch, ..., K, M]. batch, ... must be the same as those in x2.
- Restrictions: none
- x2:
- Required: yes
- Type: fp16
- Description: tensor of 3–8 dimensions. If adj_x2 = False, x2 has shape [batch, ..., M, K]. If adj_x2 = True, x2 has shape [batch, ..., K, M]. batch, ... must be the same as those in x1.
- Restrictions: none
Attribute
- adj_x1:
- Required: no
- Type: bool
- Description: whether to transpose x1 (default = False)
- Restrictions: none
- adj_x2:
- Required: no
- Type: bool
- Description: whether to transpose x2 (default = False)
- Restrictions: none
Output
y:
- Required: yes
- Type: fp16
- Description: output y (x1 * x2), with shape [batch, ..., M, N]
- Restrictions: none
Parent topic: Supported Caffe Operators