-
When using a horizon > 1
```
X,y = SlidingWindowPanel(window_len=72,
unique_id_cols=['cohort'],
stride=1,
start=0,…
-
When running the TSBERT notebook I tried adding a lr_find. lr_find ran and the output looked reasonable, but the subsequent fit_one_cycle gave the error below.
```
learn = ts_learner(udls100, Ince…
-
**Before creating an issue**
Please confirm that you are on the latest version of pytube by installing from the source.
You can do this by running `python -m pip install git+https://github.com/pyt…
-
First of all, thank you for this amazing framework!!
Using last version of fastai, when setting ds_idx=0 and training DataLoader is using `drop_last=True` raise IndexError.
My intuition is that …
-
Feature requests should first be proposed on [the forum](https://forums.fast.ai).
**Link to forum discussion.**
https://forums.fast.ai/t/tabulardataloaders-test-dl-cant-be-created-in-place/7817…
-
Hello,
I'm playing the package with the provided datasets. With sample in shape of batch*var*time_points, I noticed that when the
time_points is too big, the learner will throw error about dimensio…
ymlea updated
3 years ago
-
I'm admittedly fairly new to fastai2, but I did search in the forums for the `from_dsets` keyword and found no useful results.
Please confirm you have the latest versions of fastai, fastcore, and …
-
Since I've upgraded from Pillow v8.2.0 to the latest Pillow version 8.3.0, the predict method from Learner break with this error:
`RuntimeError: Could not infer dtype of PILImage`
-
Hi @ohmeow ,
Is it possible to use `blurr` to do achieve this? I'm trying to do transfer learning on a pretrained NER task model with 39 labels (instead of a LM model like in your example) to a sma…
-
Feature requests should first be proposed on [the forum](https://forums.fast.ai).
**Link to forum discussion.**
https://forums.fast.ai/t/how-to-do-batch-inference/39201
This refers to the ol…