tfsum

Tensorflow provides golang api for model training and inference, however currently tensorboard is only supported when using python.

This repository enables using tensorboard with tensorflow golang api. tfsum.Writer writes file that tensorboard could understand.

Before using the following piece of code, you have to build tensorflow golang api: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/go/README.md

Example

package main

import (
        "fmt"

        "github.com/helinwang/tfsum"
        tf "github.com/tensorflow/tensorflow/tensorflow/go"
)

func main() {
        step := 0
        w := &tfsum.Writer{Dir: "./tf-log", Name: "train"}
        var s *tf.Session
        var g *tf.Graph
        var input *tf.Tensor
        sum, err := s.Run(
                map[tf.Output]*tf.Tensor{
                        g.Operation("input").Output(0): input,
                },
                []tf.Output{
                        g.Operation("MergeSummary/MergeSummary").Output(0),
                },
                []*tf.Operation{
                        g.Operation("train_step"),
                })
        if err != nil {
                fmt.Println(err)
        }
        err = w.AddEvent(sum[0].Value().(string), int64(step))
        if err != nil {
                fmt.Println(err)
        }
}

Then run tensorboard normally

tensorboard --logdir=./tf-log