-
`import hezar` is slow for some reason. We have to run a profiler on the imports and find out what's causing this and how we can fix it.
-
Right now, in order to use a model to predict on raw data you have to do it like this:
```python
from hezar import Model, Tokenizer
model_path = "hezarai/bert-fa-sentiment-digikala-snappfood"
…
-
-
Provide a github action file to build the docs and deploy a github page from it to host the docs.
-
Add Roberta model from Transformers for sequence labeling. See `BertSequenceLabeling` and `RobertaTextClassification` for references.
-
How can we access the preprocessors in the Trainer properly?
-
https://github.com/hezarai/hezar/blob/c7b0ae3a09eda2aaa2302060e6dc2640ed80d4be/pyproject.toml#L34
For example:
torch = ">=1.10.0, =1.10.0"
-
https://github.com/hezarai/hezar/blob/2f64bae79189d40cc99a4e6e37491d812b91c261/.github/workflows/pypi-release.yml#L6
It is an old habit to add v at the beginning of a tag. Please remove it.
-
I attempted to test the workflow by creating a new tag and it raised an error.
See it here: https://github.com/hezarai/hezar/actions/runs/5399733975/jobs/9807244244#step:3:10
@pooya-mohammadi Ca…
-
![Screenshot from 2023-08-05 14-04-54](https://github.com/hezarai/hezar/assets/55460936/5e80b3c3-df54-484f-9748-3d81104355fd)