RedisDL

A Redis module for serving tensors and executing deep learning graphs. Expect changes in the API and internals.

Cloning

If you want to run examples, make sure you have git-lfs installed when you clone.

Building

This will checkout and build Redis and download the libtensorflow binaries for your platform.

bash get_deps.sh
make install

Running the server

make run

Running tests

Tests are in the works, there's just a check script. Make sure the server is running, then

cd examples/models
# build a graph and write it out
python tf-minimal.py

On the client, load the graph

./deps/redis/src/redis-cli -x DL.GSET foo TF < graph.pb

Then create the input tensors, run the computation graph and get the output tensor (see load_model.sh). Note the signatures:

• DL.TSET tensor_key data_type ndims dim1..dimN [BLOB data | VALUES val1..valN]
• DL.GRUN graph_key ninputs input_key input_name_in_graph ... output_key output_name_in_graph ...

redis-cli
> DL.TSET bar FLOAT 1 2 VALUES 2 3
> DL.TSET baz FLOAT 1 2 VALUES 2 3
> DL.GRUN foo 2 bar a baz b jez c
> DL.TGET jez VALUES
1) (integer) 4
2) (integer) 9

DL.TSET tensor_key data_type dim shape1..shapeN [BLOB data | VALUES val1..valN]

Stores a tensor of defined type (FLOAT, DOUBLE, INT8, INT16, INT32, INT64, UINT8, UINT16) with N dimensions (dim) and shape given by shape1..shapeN

DL.TGET tensor_key [BLOB | VALUES]

DL.TDATATYPE tensor_key

DL.TDIM tensor_key

DL.TSHAPE tensor_key

DL.TBYTESIZE tensor_key

DL.GSET graph_key TF graph_blob prefix

Stores a graph provided as a protobuf blob

DL.GRUN graph_key ninputs input_key input_name_in_graph ... output_key output_name_in_graph ...

License

BSD license https://opensource.org/licenses/BSD-3-Clause

Copyright 2018, Luca Antiga, Orobix Srl (www.orobix.com).