Abdelrhman Rabea
Totally disagree with @6. Without knowing who you're talking to, how can you know *how* to talk? It's fundamental.
عرض النقاش
F
Freya Rossi
@freyarossi9428
Forum member
0منشور
0متابِع
0يتابع
7تعليق
0إعجاب
0مشاهدة
This is great! I've been struggling with vague outputs from ChatGPT, and I think 'C' (Context) and 'Y' (Yield) are where I fall short most often. Specifically, defining the desired output format is something I always forget until the AI gives me something I can't use.
عرض النقاش
What about the inference costs post-fine-tuning? Are you running it on your own server or a managed service? That's where recurring costs can sneak up.
عرض النقاش
Beginner here: When you say 'deterministic responses', does that mean it will always give the exact same output for the same prompt if the temperature is 0.1, or just very similar ones?
عرض النقاش
Totally agree! For creative stuff, though, I sometimes bump it up to 1.2 or even 1.5 if the model allows it, paired with top_p around 0.9. It's chaos but sometimes pure gold comes out!
عرض النقاش
Look into the 'lift and shift' vs 'cloud native' approach. For a startup, cloud native is almost always better long-term, despite the initial re-engineering.
عرض النقاش
To @comment13: It's usually a small pencil icon next to your post, or sometimes under a '...' menu. Took me a sec too!
عرض النقاش
