Decoder Specifications

This page details the inputs and outputs for the decoder's init and step functions.

Init Function

Signature: init

Inputs

Name	Tensor Name	Type	Key Info
`args_0`	`init_args_0:0`	`float32`	Shape: `[1, 196, 256]` ID: `0`
`args_1`	`init_args_1:0`	`int64`	Shape: `[1, 1]` ID: `1`

Outputs

Name	Tensor Name	Type	Key Info
`output_0`	`StatefulPartitionedCall:0`	`float32`	Shape: `[1, 9415]` ID: `711`
`output_1`	`StatefulPartitionedCall:1`	`float32`	Shape: `[4, 1, 4, 1, 64]` ID: `713`
`output_2`	`StatefulPartitionedCall:2`	`float32`	Shape: `[4, 1, 4, 1, 64]` ID: `715`
`output_3`	`StatefulPartitionedCall:3`	`float32`	Shape: `[4, 1, 4, 196, 64]` ID: `307`
`output_4`	`StatefulPartitionedCall:4`	`float32`	Shape: `[4, 1, 4, 196, 64]` ID: `309`

Step Function

Signature: step

Inputs

Name	Tensor Name	Type	Key Info
`args_0`	`step_args_0:0`	`float32`	Shape: `[1, 196, 256]` ID: `3`
`args_1`	`step_args_1:0`	`int64`	Shape: `[1, 1]` ID: `6`
`args_2`	`step_args_2:0`	`int64`	Shape: `[1, 1]` ID: `1`
`args_3`	`step_args_3:0`	`float32`	Shape: `[4, 1, 4, 256, 64]` ID: `2`
`args_4`	`step_args_4:0`	`float32`	Shape: `[4, 1, 4, 256, 64]` ID: `5`
`args_5`	`step_args_5:0`	`float32`	Shape: `[4, 1, 4, 196, 64]` ID: `0`
`args_6`	`step_args_6:0`	`float32`	Shape: `[4, 1, 4, 196, 64]` ID: `4`

Outputs

Name	Tensor Name	Type	Key Info
`output_0`	`StatefulPartitionedCall_1:0`	`float32`	Shape: `[1, 9415]` ID: `679`
`output_1`	`StatefulPartitionedCall_1:1`	`float32`	Shape: `[4, 1, 4, 1, 64]` ID: `681`
`output_2`	`StatefulPartitionedCall_1:2`	`float32`	Shape: `[4, 1, 4, 1, 64]` ID: `683`