README
¶
What is tfcgo?
tfcgo was born out of necessity. Tensorflow go bindings are slow to progress. Relying on protobufs to generate wrapper code is dependant on the growth of the c/c++ public api - and the underlying framework has much more to offer! This library will eventually do the hard work bridging the gap between go and the Tensorflow c++ framework. Training a TF machine learning algorithm is possible in go. Please note contributions in this repo will be submitted to the main tensorflow repo.
Docker Installation
Build your own image from source
git clone this repo:
git clone https://github.com/ctava/tfcgo
Run the following commands:
cd tfcgo
docker build -t ctava/tfcgo .
To run with shared volume, you can add /go to shared paths from Docker -> Preferences... -> File Sharing.
docker run -v /go/src/github.com/ctava/tfcgo:/go/src/github.com/ctava/tfcgo -it -p 8888:8888 ctava/tfcgo:latest
and your in. You now have tensorflow + golang + tfcgo available.
Running from a prebuilt/publish docker image
Run the following commands:
docker pull ctava/tfcgo:latest
docker run -it --security-opt=seccomp:unconfined -p 8888:8888 ctava/tfcgo:latest
--security-opt for the purposes of using delve
and your in. You now have tensorflow + golang + tfcgo available.
Confirm Golang, Tensorflow and tfcgo installation
Run the following commands:
go run versioncheck.go
You should see the versions of golang, tensorflow and tfcgo as specified in the docker file.
Now for the fun stuff
go run listops.go
You should see a list of all of the ops available in tensorflow
Training
go run linear.go
Multi-Class Classification
go run iris.go (coming soon)
Additional resources
Documentation
¶
Index ¶
- func Log(s *op.Scope, value interface{}) (*tf.Tensor, tf.Output, error)
- func MakeConst(s *op.Scope, value interface{}) (tf.Output, error)
- func MakeTensorAndOutput(s *op.Scope, value interface{}) (*tf.Tensor, tf.Output, error)
- func RandomSample(t *tf.Tensor) (*tf.Tensor, error)
- func ReduceMean(t *tf.Tensor) (*tf.Tensor, error)
- func RegisteredOps() (*pb.OpList, error)
- func SaySomething(saying string) string
- func Square(s *op.Scope, value interface{}) (*tf.Tensor, tf.Output, error)
- func SquareT(t *tf.Tensor) (*tf.Tensor, error)
- func SumOfSquares(s *op.Scope, value interface{}) (*tf.Tensor, tf.Output, error)
- func Variable(scope *op.Scope, initialValue tf.Output) (init *tf.Operation, handle, value tf.Output)
- func VariableAndTensor(scope *op.Scope, initialValue tf.Output, v interface{}) (t *tf.Tensor, init *tf.Operation, handle, value tf.Output)
- func Version() float64
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func MakeTensorAndOutput ¶
func ReduceMean ¶
ReduceMean - mean accross elements
func RegisteredOps ¶
RegisteredOps returns all of the supported TF c/c++ api
func SumOfSquares ¶
SumOfSquares - accepts a value (expecting [][]float64) and returns sum of squares
func Variable ¶
func Variable(scope *op.Scope, initialValue tf.Output) (init *tf.Operation, handle, value tf.Output)
Variable - this function creats a variable for training purposes.
Returns: the init operation - very important - this needs to be run in a session to be initialized a handle to the variable - for use in assignment operations an Output that produces the current value of the variable
func VariableAndTensor ¶
func VariableAndTensor(scope *op.Scope, initialValue tf.Output, v interface{}) (t *tf.Tensor, init *tf.Operation, handle, value tf.Output)
Variable - this function creats a variable for training purposes.
Returns: the init operation - very important - this needs to be run in a session to be initialized a handle to the variable - for use in assignment operations an Output that produces the current value of the variable
Types ¶
This section is empty.
Source Files
Directories