uranus Yu

It is equivalent. Remember to replace code `model_out = tf.add(y_pred, epsilon)` with `model_out = y_pred` to get consistent result. test code ``` y_t=[[0.0,1.0,0.0],[1.0,0.0,0.0],[0.0,0.0,1.0],[1.0,0.0,0.0]] y_p=[[0.9,0.1,0.0],[0.5,0.25,0.25],[0.15,0.05,0.8],[0.5,0.4,0.1]] def focal_loss_fixed(y_true, y_pred, gamma=0.0, alpha=1.0): epsilon...

No, it's good. just rename .tar to .pth.

> pip install pydantic==1.10.11 It works like a charm !! no more hang when edit the image

Here is the conda version in the Google Colab. ## install condacolab package ```python !pip install -q condacolab import condacolab condacolab.install() import sys print(sys.executable) ``` ## clone the source code...