Indeterministic results with LongFormer
Offen
- Vorherrschende Sprache
- Python
- Sterne
- 2.2k
- Forks
- 285
- PR-Merge-Kennzahlen
- Keine gemergten PRs in 30 T.
Beschreibung
**Results are not reproducible!!!**
Recently, I noticed that my results of LongFormer varies even after fixing the random seed. Any body can help me solve this question?
The weird thing is that if you drop the max-sequence length from 1024 to 512 you suddenly get deterministic results. Also if I go from 512 to 513 I no longer get deterministic results, This question has puzzled me for a long time.
Hope someone can help me out! Thanks a lot.
Beitragsleitfaden
Für dieses Repository ist kein Beitragsleitfaden indexiert
Bewertung
Dieses Issue wurde noch nicht bewertet.