1 op { 2 graph_op_name: "AudioSummary" 3 in_arg { 4 name: "tag" 5 description: <<END 6 Scalar. Used to build the `tag` attribute of the summary values. 7 END 8 } 9 in_arg { 10 name: "tensor" 11 description: <<END 12 2-D of shape `[batch_size, frames]`. 13 END 14 } 15 out_arg { 16 name: "summary" 17 description: <<END 18 Scalar. Serialized `Summary` protocol buffer. 19 END 20 } 21 attr { 22 name: "sample_rate" 23 description: <<END 24 The sample rate of the signal in hertz. 25 END 26 } 27 attr { 28 name: "max_outputs" 29 description: <<END 30 Max number of batch elements to generate audio for. 31 END 32 } 33 summary: "Outputs a `Summary` protocol buffer with audio." 34 description: <<END 35 The summary has up to `max_outputs` summary values containing audio. The 36 audio is built from `tensor` which must be 3-D with shape `[batch_size, 37 frames, channels]` or 2-D with shape `[batch_size, frames]`. The values are 38 assumed to be in the range of `[-1.0, 1.0]` with a sample rate of `sample_rate`. 39 40 The `tag` argument is a scalar `Tensor` of type `string`. It is used to 41 build the `tag` of the summary values: 42 43 * If `max_outputs` is 1, the summary value tag is '*tag*/audio'. 44 * If `max_outputs` is greater than 1, the summary value tags are 45 generated sequentially as '*tag*/audio/0', '*tag*/audio/1', etc. 46 END 47 } 48
