fastDeploy CLI usage explained
fastDeploy - - help
# or
python - m fastDeploy - - help
Start prediction loop on your recipe
fastdeploy - - loop - - recipe ./ recipes / echo
Optional config can be passed with --config
flag
fastdeploy - - loop - - recipe ./ recipes / echo - - config "predictor_name=predictor.py;optimal_batch_size=0"
Config
Description
Default
predictor_name
predictor.py or predictor_N.py, name of the predictor run in the loop
predictor.py
optimal_batch_size
integer max batch size for the predictor
0 (auto determine)
Same config can also be passed as env variables
export PREDICTOR_NAME = predictor .py
export OPTIMAL_BATCH_SIZE = 0
fastdeploy - - loop - - recipe ./ recipes / echo
Start API server on your recipe
fastdeploy - - rest - - recipe ./ recipes / echo
Optional config can be passed with --config
flag
fastdeploy - - rest - - recipe ./ recipes / echo - - config "max_request_batch_size=0;workers=3;timeout=480;host=0.0.0.0;port=8080;only_async=false;allow_pickle=true;keep_alive=60"
Same config can also be passed as env variables
export MAX_REQUEST_BATCH_SIZE = 0
export WORKERS = 3
export TIMEOUT = 480
export HOST = 0.0 .0.0
export PORT = 8080
export ONLY_ASYNC = false
export ALLOW_PICKLE = true
export KEEP_ALIVE = 60
fastdeploy - - rest - - recipe ./ recipes / echo
--config options explained
Config
Description
Default
max_request_batch_size
integer max number of inputs in a batch. useful when exposing to outside directly to limit max number of inputs that can be in a request
0 (None)
workers
number of rest api gunicorn workers. 3 is more than enoough generally
3
timeout
seconds after which request will fail
480
host
host for the REST server
0.0.0.0
port
port for the REST server
8080
only_async
true/false
false
allow_pickle
true/false - use for disallowing pickle protocol when expecting external inputs
true
keep_alive
gunicorn gevent keep alive
60
Build generate docker image for your recipe
fastdeploy - - build - - recipe ./ recipes / echo
also supports optional config via --config
flag
both rest and loop config options can be passed here in the same config string