-
Notifications
You must be signed in to change notification settings - Fork 32
CLI
Huaizheng edited this page Apr 22, 2021
·
4 revisions
MLModelCI is an easy-to-use toolkit to build high-performing ML products from newly trained research models.
APIs described in this are classified by different levels
-
modelci service init -
modelci service stop -
modelci service clean
-
modelci service monitor
modelci modelhub publish
Publish a model to the ModelCI. You can choose either use a YAML file (using the -f option), or input all parameters using other options/arguments.
Arguments:
[FILE_OR_DIR] Model weight files
Options:
-n, --name TEXT Architecture
-fw, --framework [TensorFlow|PyTorch]
Framework
-e, --engine [NONE|TFS|TORCHSCRIPT|ONNX|TRT|TVM|CUSTOMIZED|PYTORCH]
Engine
-v, --version INTEGER RANGE Version number
-t, --task [Image_Classification|Object_Detection|Segmentation]
Task
-d, --dataset TEXT Dataset name
--metric DICT Metrics in the form of mapping JSON string.
The map type is
`Dict[types.models.mlmodel.Metric, float]`.
An example is '{"acc": 0.76}.' [default:
{}]
-i, --input PYDANTIC MODEL JSON
List of shape definitions for input tensors.
An example of one shape definition is
'{"name": "input", "shape": [-1, 3, 224,
224], "dtype": "TYPE_FP32", "format":
"FORMAT_NCHW"}' [default: ]
-o, --output PYDANTIC MODEL JSON
List of shape definitions for output
tensors. An example of one shape definition
is '{"name": "output", "shape": [-1, 1000],
"dtype": "TYPE_FP32"}' [default: ]
-c, --convert Convert the model to other possible format.
[default: True]
-p, --profile Profile the published model(s). [default:
False]
-f, --yaml-file PATH Path to configuration YAML file. You should
either set the `yaml_file` field or fields
(`FILE_OR_DIR`, `--name`, `--framework`,
`--engine`, `--version`, `--task`,
`--dataset`,`--metric`, `--input`,
`--output`).
--help Show this message and exit.
modelci modelhub ls
Show a table that lists all models published in MLModelCI.
Options:
-n, --name TEXT Model architecture name
-fw, --framework [TensorFlow|PyTorch]
Framework
-e, --engine [NONE|TFS|TORCHSCRIPT|ONNX|TRT|TVM|CUSTOMIZED|PYTORCH]
Serving engine
-v, --version INTEGER Version
-a, --all Display queried models. otherwise, only
partial result will be shown. [default:
False]
--help Show this message and exit.
modelci modelhub get
download a model weight file from an url
modelci modelhub export
For quickly download a model from Pytorch/Tensorflow Hub to our system. Not officially supported API
modelci modelhub detail
Show a single model in detail.
Arguments:
MODEL_ID Model ID [required]
Options:
--help Show this message and exit.
modelci modelhub update
Update a single model info
Arguments:
MODEL_ID Model ID [required]
Options:
-n, --name TEXT Architecture
-fw, --framework [TensorFlow|PyTorch]
Framework
-e, --engine [NONE|TFS|TORCHSCRIPT|ONNX|TRT|TVM|CUSTOMIZED|PYTORCH]
Engine
-v, --version INTEGER RANGE Version number
-t, --task [Image_Classification|Object_Detection|Segmentation]
Task
-d, --dataset TEXT Dataset name
--metric DICT Metrics in the form of mapping JSON string.
The map type is
`Dict[types.models.mlmodel.Metric, float]`.
An example is '{"acc": 0.76}.'
-i, --input PYDANTIC MODEL JSON
List of shape definitions for input tensors.
An example of one shape definition is
'{"name": "input", "shape": [-1, 3, 224,
224], "dtype": "TYPE_FP32", "format":
"FORMAT_NCHW"}' [default: ]
-o, --output PYDANTIC MODEL JSON
List of shape definitions for output
tensors. An example of one shape definition
is '{"name": "output", "shape": [-1, 1000],
"dtype": "TYPE_FP32"}' [default: ]
--help Show this message and exit.
modelci modelhub delete
Delete a single model by its id
Arguments:
MODEL_ID Model ID [required]
Options:
--help Show this message and exit.
modelci modelhub convert
convert a single model to all possible alternatives by its id or yaml file.
Arguments:
You should use one and only option from -i and -f as an argument to
assign the model to be converted.
Options:
-i, --id model id in the modelhub
-f, --yaml-file Path to configuration YAML file. You should
either set the `yaml_file` field or fields
(`FILE_OR_DIR`, `--name`, `--framework`,
`--engine`, `--version`, `--task`,
`--dataset`,`--metric`, `--input`,
`--output`).
-r, --register whether to register the converted models to modelhub,
default false
--help Show this message and exit.
-
modelci modelhub profile -
modelci modelhub dispatch
-
modelci modelps visulize
pop up a webpage to display a model's structure
-
modelci modelps edit
pop up a webpage to guide users edit a model step by step
-
modelci modelps tune
fine-tune a model on a specific dataset