Closed chschroeder closed 1 year ago
Seems that I, on the other hand, a was a few argilla versions behind. This is more complicated than just raising the version number and running the notebook once.
When I execute the following part from the first notebook:
import argilla as rg
# Choose a name for the dataset
DATASET_NAME = "trec_with_active_learning"
# Define labeling schema
labels = trec["train"].features["coarse_label"].names
settings = rg.TextClassificationSettings(label_schema=labels)
# Create dataset with a label schema
rg.configure_dataset_settings(name=DATASET_NAME, settings=settings)
Then, rg.configure_dataset_settings()
raises the following error:
BadRequestApiError: Argilla server returned an error with http status: 400
Error details: [{'code': 'argilla.api.errors::MissingInputParamError', 'params': {'message': 'A workspace must be
provided'}}]
Can this be used without setting up an explicit workspace? As far as I have understood, using workspaces would require additional setup steps, which I would like to avoid.
@chschroeder thanks for taking a look at this. A PR is very welcome.
I think the above can be resolved by adding a workspace
parameter rg.configure_dataset_settings()
. Alternatively, you could set rg.set_workspace()
.
Thanks for the feedback, but that a workspace could help is what I had given by the error message. What I was trying to ask is more like: is this really needed? Is this the right way?
Next, I tried it by following this example: https://docs.argilla.io/en/latest/guides/llms/examples/curating-feedback-instructiondataset.html
try:
workspace = rg.Workspace.from_name(username)
except:
workspace = rg.Workspace.create(username)
user = rg.User.from_name(username)
workspace.add_user(user.id)
Here rg.Workspace.create()
does not work with the default admin user:
PermissionError: User with role=admin is not allowed to call `create`. Only users with role=[<UserRole.owner:
'owner'>] are allowed to call this function.
Of course I can add a user now, but adds two extra steps (creating a workspace, creating a user) to the notebook, compared to before.
Is this really necessary? And if so, why does it seem not to be necessary for other tutorials?
Edit:
Same for user rg.User.from_name()
. This is really confusing and not clear by reading the docs.
PermissionError: User with role=admin is not allowed to call `create`. Only users with role=[<UserRole.owner:
'owner'>] are allowed to call this function.
There are two notebooks, which reference small-text versions v1.1.0/v1.1.1.
I will likely create a PR later or tomorrow. This issue is just for the purposes of documentation.
Is your feature request related to a problem? Please describe. I just published the bugfix release v1.3.1, which brings important bugfixes for transformer-based classification. Moreover, the additions from v1.2.0 will be accessible as well.
Describe the solution you'd like Two notebooks need to be updated:
Describe alternatives you've considered
--
Additional context
--