mirror of
https://github.com/clearml/clearml
synced 2025-01-31 00:56:57 +00:00
Fix style and grammar mistakes
This commit is contained in:
parent
ac1ae7f447
commit
05c048d0bd
@ -2,13 +2,13 @@
|
||||
|
||||
## Rationale
|
||||
|
||||
Due to an issue with ClearML SDK versions 1.11.x, URLs of objects uploaded to the Google Cloud Storage were stored in the ClearML backend as a quoted string. This causes issues accessing these objects directly from the ClearML SDK. This is relevant for URLs of models, datasets, artifacts, and media files/debug samples. In case you have such objects uploaded with the affected ClearML SDK versions and wish to be able to access them programmatically using the ClearML SDK using the version 1.12 and above (note that access from the ClearML UI is still possible), you should perform one of the actions in the section below.
|
||||
Due to an issue with ClearML SDK versions 1.11.x, URLs of objects uploaded to the Google Cloud Storage were stored in the ClearML backend as a quoted string. This behavior causes issues accessing registered objects using the ClearML SDK. The issue affects the URLs of models, datasets, artifacts, and media files/debug samples. In case you have such objects uploaded with the affected ClearML SDK versions and wish to be able to access them programmatically using the ClearML SDK using version 1.12 and above (note that access from the ClearML UI is still possible), you should perform the actions listed in the section below.
|
||||
|
||||
## Recommended Steps
|
||||
|
||||
The code snippets below should serve as an example rather than an actual conversion script.
|
||||
|
||||
The general flow is that you will first need to download these files by a custom access method, then upload them with the fixed SDK version. Depending on what object you're trying to fix, you should pick the respective lines of code from step 1 and 2.
|
||||
The general flow is that you will first need to download these files by a custom access method, then upload them with the fixed SDK version. Depending on what object you're trying to fix, you should pick the respective lines of code from steps 1 and 2.
|
||||
|
||||
|
||||
|
||||
@ -67,7 +67,7 @@ for sample in samples:
|
||||
|
||||
## Alternative methods
|
||||
|
||||
These methods are more advanced (read "more likely to mess up"). If you're unsure whether to use them or not, better don't. Both methods described below will alter (i.e. modify **in-place**) the existing objects. Note that you still need to run the code from step 1 to have access to all required metadata.
|
||||
The methods described next are more advanced (read "more likely to mess up"). If you're unsure whether to use them or not, better don't. Both methods described below will alter (i.e., modify **in-place**) the existing objects. Note that you still need to run the code from step 1 to have access to all required metadata.
|
||||
|
||||
**Method 1**: You can try to alter the existing unpublished experiments/models using the lower-level `APIClient`
|
||||
```python
|
||||
@ -84,4 +84,4 @@ client.tasks.add_or_update_artifacts(task=task.id, force=True, artifacts=[{"uri"
|
||||
client.models.edit(model=model.id, force=True, uri=url) # For any unpublished Model
|
||||
```
|
||||
|
||||
**Method 2**: There's an option available only to those who self-host their ClearML server. It is possible to manually update the values registered in MongoDB, but beware - this is an advanced procedure that should be handled with extreme care, as it can lead to an inconsistent state if mishandled.
|
||||
**Method 2**: There's an option available only to those who self-host their ClearML server. It is possible to manually update the values registered in MongoDB, but beware - this advanced procedure should be performed with extreme care, as it can lead to an inconsistent state if mishandled.
|
||||
|
Loading…
Reference in New Issue
Block a user