Spaces:
Runtime error
Is it the full ERNIE or a smaller model of ERNIE?
Hello! Is it the full ERNIE or a smaller model of ERNIE?
I'm testing/ documenting AIs on complex/ novel/ lengthy/ hard/ interesting/ etc tests (and realize ERNIE can't do as long prompts as DALL-E 2 can) and on the prompts that fit into ERNIE I noticed ERNIE isn't keeping up as good as DALL-E 2 on most tests - just enough to tell. ERNIE also has lower quality/resolution I see from my tests. I mostly am using the deepl translator. Sometimes ERNIE misses an extra thing or does it worse - so the translator I assume is as good as it gets (besides using a real human translator). Or could that be enough to cause problems?
I also wonder if ERNIE is as big a model as DALL-E 2 / same compute / or data trained on? Just wondering why is it seems to not be doing as good as DALL-E 2. Maybe ex. it got a better score say if the benchmark was using smaller dataset, but it is smaller model benchmark?