Training metrics explaination #2411
-
|
Hi, |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 1 reply
-
|
The metrics are computed for each batch and aggregated for all items in the dataset (completed during a single epoch). That's why the dashboard reports the metric evaluated for the current batch and the metric evaluated the whole dataset (seen during the last complete epoch). |
Beta Was this translation helpful? Give feedback.
-
|
The following is another snapshot. It seems that the training (red color) and validation (blue color) were performed by turns. Why weren't them performed synchronously ?
|
Beta Was this translation helpful? Give feedback.
-
|
The training and validation are indeed perform in turns! At the end of each training epoch, the trained model is evaluated on the validation set. When the validation is completed, the next training epoch starts. |
Beta Was this translation helpful? Give feedback.
The metrics are computed for each batch and aggregated for all items in the dataset (completed during a single epoch).
That's why the dashboard reports the metric evaluated for the current batch and the metric evaluated the whole dataset (seen during the last complete epoch).