TFaaS

TensorFlow as a Service¶

TensorFlow as a Service (TFaas) was developed as a general purpose service which can be deployed on any infrastruction from personal laptop, VM, to cloud infrastructure, inculding kubernetes/docker based ones. The main repository contains all details about the service, including install, end-to-end example, and demo.

For CERN users we already deploy TFaaS on the following URL: https://cms-tfaas.cern.ch

It can be used by CMS members using any HTTP based client. For example, here is a basic access from curl client:

curl -k https://cms-tfaas.cern.ch/models
[
  {
    "name": "luca",
    "model": "prova.pb",
    "labels": "labels.csv",
    "options": null,
    "inputNode": "dense_1_input",
    "outputNode": "output_node0",
    "description": "",
    "timestamp": "2021-10-22 14:04:52.890554036 +0000 UTC m=+600537.976386186"
  },
  {
    "name": "test_luca_1024",
    "model": "saved_model.pb",
    "labels": "labels.txt",
    "options": null,
    "inputNode": "dense_input_1:0",
    "outputNode": "dense_3/Sigmoid:0",
    "description": "",
    "timestamp": "2021-10-22 14:04:52.890776518 +0000 UTC m=+600537.976608672"
  },
  {
    "name": "vk",
    "model": "model.pb",
    "labels": "labels.txt",
    "options": null,
    "inputNode": "dense_1_input",
    "outputNode": "output_node0",
    "description": "",
    "timestamp": "2021-10-22 14:04:52.890903234 +0000 UTC m=+600537.976735378"
  }
]

The following APIs are available: - /upload to push your favorite TF model to TFaaS server either for Form or as tar-ball bundle, see examples below - /delete to delete your TF model from TFaaS server - /models to view existing TF models on TFaaS server - /predict/json to serve TF model predictions in JSON data-format - /predict/proto to serve TF model predictions in ProtoBuffer data-format - /predict/image to serve TF model predictions forimages in JPG/PNG formats

➀ look-up your favorite model¶

You may easily look-up your ML model from TFaaS server, e.g.

curl https://cms-tfaas.cern.ch/models
# possible output may looks like this
[
  {
    "name": "luca",
    "model": "prova.pb",
    "labels": "labels.csv",
    "options": null,
    "inputNode": "dense_1_input",
    "outputNode": "output_node0",
    "description": "",
    "timestamp": "2021-11-08 20:07:18.397487027 +0000 UTC m=+2091094.457327022"
  }
  ...
]

The provided /models API will list the name of the model, its file name, labels file, possible options, input and output nodes, description and proper timestamp when it was added to TFaaS repository

➁ upload your TF model to TFaaS server¶

If your model is not in TFaaS server you may easily add it as following:

# example of image based model upload
curl -X POST https://cms-tfaas.cern.ch/upload
-F 'name=ImageModel' -F 'params=@/path/params.json'
-F 'model=@/path/tf_model.pb' -F 'labels=@/path/labels.txt'

# example of TF pb file upload
curl -s -X POST https://cms-tfaas.cern.ch/upload \
    -F 'name=vk' -F 'params=@/path/params.json' \
    -F 'model=@/path/model.pb' -F 'labels=@/path/labels.txt'

# example of bundle upload produce with Keras TF
# here is our saved model area
ls model
assets         saved_model.pb variables
# we can create tarball and upload it to TFaaS via bundle end-point
tar cfz model.tar.gz model
curl -X POST -H "Content-Encoding: gzip" \
             -H "content-type: application/octet-stream" \
             --data-binary @/path/models.tar.gz https://cms-tfaas.cern.ch/upload

➂ get your predictions¶

Finally, you may obtain predictions from your favorite model by using proper API, e.g.

# obtain predictions from your ImageModel
curl https://cms-tfaas.cern.ch/image -F 'image=@/path/file.png' -F 'model=ImageModel'

# obtain predictions from your TF based model
cat input.json
{"keys": [...], "values": [...], "model":"model"}

# call to get predictions from /json end-point using input.json
curl -s -X POST -H "Content-type: application/json" \
    -d@/path/input.json https://cms-tfaas.cern.ch/json

Fore more information please visit curl client page.

TFaaS interface¶

Clients communicate with TFaaS via HTTP protocol. See examples for Curl, Python and C++ clients.

TFaaS benchmarks¶

Benchmark results on CentOS, 24 cores, 32GB of RAM serving DL NN with 42x128x128x128x64x64x1x1 architecture (JSON and ProtoBuffer formats show similar performance): - 400 req/sec for 100 concurrent clients, 1000 requests in total - 480 req/sec for 200 concurrent clients, 5000 requests in total

For more information please visit bencmarks page.

Last update: January 8, 2024