Timeline for What does from_logits=True do in SparseCategoricalcrossEntropy loss function?
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 10, 2023 at 14:45 | comment | added | Peter Leopold |
This seems like an API bug. The output layer either normalizes or it doesn't. The function signature should include the option from_normalized or input_is_normalized rather than the rather odd, confusing, off-point and far-from-universal sub-case from_logits.
|
|
| May 19, 2021 at 21:57 | comment | added | Crispen Gari | perfect explanation. | |
| Jul 25, 2020 at 13:03 | history | edited | today | CC BY-SA 4.0 |
fix typo
|
| Apr 30, 2020 at 12:23 | vote | accept | Nagendra Prasad | ||
| May 1 at 9:16 | |||||
| Apr 30, 2020 at 3:11 | comment | added | today | @NagendraPrasad That's right. | |
| Apr 29, 2020 at 20:07 | vote | accept | Nagendra Prasad | ||
| Apr 29, 2020 at 20:09 | |||||
| Apr 29, 2020 at 20:05 | comment | added | Nagendra Prasad | so basically what it means is if softmax layer is not being added at the last layer then we need to have the from_logits=True to indicate the probabilities are not normalized , | |
| Apr 29, 2020 at 20:03 | vote | accept | Nagendra Prasad | ||
| Apr 29, 2020 at 20:03 | |||||
| Apr 28, 2020 at 20:32 | history | edited | today | CC BY-SA 4.0 |
add clarification
|
| Apr 28, 2020 at 20:00 | history | edited | today | CC BY-SA 4.0 |
added note and clarified things a bit more
|
| Apr 28, 2020 at 19:51 | history | answered | today | CC BY-SA 4.0 |