A C++ library for TensorBoard. The library provides APIs to write TensorBoard events.
The current release is tested on CentOS, Ubuntu and WSL, with TensorBoard 2.13.0.
- Support
scalar,histogram,histogram_raw,image,images,audio,textandprojectorsummaries.
You can write events by AddXXX function calls like in Python, e.g.,
#include "record/recorder.h"
int main(int argc, char* argv[]) {
// Initialize TensorBoard recoder
nlptk::Recorder recorder();
// add text
recorder.AddText("text", "text message", 0);
// add scalar
for (int i = 0; i < 1000; ++i) {
recorder.AddScalar("scalar", 0.01 * i, i);
}
// ...
return 0;
}
Current, TensorBoard.cpp supports Bazel only for compiling the project from source. Other build systems like CMake may be supported later.
To use TensorBoard.cpp within a project which used the Bazel build tool, add the following lines to you WORKSPACE file:
load("@bazel_tools//tools/build_defs/repo:git.bzl", "new_git_repository")
load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
git_repository(
name = "tensorboard_cpp",
remote = "https://github.com/Mininglamp-Technology/tensorboard.cpp.git",
)
http_archive(
name = "glog",
sha256 = "8a83bf982f37bb70825df71a9709fa90ea9f4447fb3c099e1d720a439d88bad6",
strip_prefix = "glog-0.6.0",
urls = [
"https://github.com/google/glog/archive/v0.6.0.tar.gz"
],
)
new_git_repository(
name = "stb",
remote = "https://github.com/nothings/stb.git",
commit = "5736b15f7ea0ffb08dd38af21067c314d6a3aae9",
shallow_since = "1675018027 -0800",
build_file = "//third_party/stb:BUILD.bazel",
)
http_archive(
name = "com_google_protobuf",
sha256 = "c29d8b4b79389463c546f98b15aa4391d4ed7ec459340c47bffe15db63eb9126",
strip_prefix = "protobuf-3.21.3",
urls = [
"https://github.com/protocolbuffers/protobuf/archive/v3.21.3.tar.gz"
],
)
load("@com_google_protobuf//:protobuf_deps.bzl", "protobuf_deps")
protobuf_deps()You can the add @tensorboard_cpp//record:record to the deps section of a cc_binary or cc_library rule, and #include <record/recorder.h> to include it in your source code. Here's a simple example:
cc_binary(
name = "mainrun",
srcs = [
"main.cc",
],
deps = [
"@tensorboard_cpp//record:record",
],
)Examples could be found in examples. The example1 use default file writer to record. The example2 use sync file writer to record.
- Clone TensorBoard.cpp project
$ git clone https://github.com/Mininglamp-Technology/tensorboard.cpp.git- Compile the example1 with
Bazel, which could be installed follow instructions here.
$ cd tensorboard.cpp
$ bazel build examples:example1- Run example and rendering:
$ bazel-bin/examples/example1
$ tensorboard --logdir runs- Time Series
- Scalars
- Images
- Audio
- Distributions
- Histograms
- Text
- Projector
