Experiment artifacts
In addition to scoring artifacts, experiments can have other types of artifacts attached to them. This page describes how to interact with those artifacts.
- Connect to H2O MLOps.
import h2o_mlops
mlops = h2o_mlops.Client()
- Get an existing experiment to work with by creating a new one.
project = mlops.projects.create(name="demo")
experiment = project.experiments.create(
data="/Users/jgranados/Downloads/GBM_model_python_1649367037255_1.zip",
name="experiment-from-client"
)
Add artifacts
json_artifact = experiment.artifacts.add("./artifact.json")
Note that it is possible to specify the MIME type if the client cannot automatically detect it.
docx_artifact = experiment.artifacts.add(
data="./artifact.docx",
mime_type="application/vnd.openxmlformats-officedocument.wordprocessingml.document"
)
List artifacts
You can view all the artifacts included in the experiment. Note the mime_type
, which indicates the media type of the data.
experiment.artifacts.list()
Output:
| name | mime_type | uid
----+---------------+---------------------------+--------------------------------------
0 | h2o3/mojo | application/zip | dacea816-1499-4848-b4eb-e151ec4152c6
1 | artifact.json | application/json | d6a0723c-8545-4821-bcff-fa8c42ef7d23
2 | artifact.docx | application/vnd.openxmlfo | 100a8f1d-1bf1-46fd-990a-cf54dd461eaf
Retrieve artifacts
json_artifact = experiment.artifacts.list(name="artifact.json")[0]
Note that it is also possible to use experiment.artifacts.get(uid)
to directly retrieve an artifact.
Consume artifacts
The following are several examples of things that can be done with an artifact.
Convert JSON artifacts to a Python dictionary or string
JSON artifacts can be converted to a Python dictionary or string. Text artifacts can only be converted to string.
Input:
json_artifact.to_dictionary()["key"]
Output:
'value'
Input:
json_artifact.to_string()
Output:
'{\n "key": "value"\n}'
Download artifacts
All artifacts can be downloaded.
Input:
docx_artifact.download(overwrite=True)
Output:
'artifact.docx'
Feedback
- Submit and view feedback for this page
- Send feedback about H2O MLOps to cloud-feedback@h2o.ai