
Trelis Research
@trelisresearch
Tutorials for AI Model Fine-tuning and Inference
ID: 1667096163902685187
https://Trelis.com/About 09-06-2023 09:08:58
668 Tweet
593 Takipçi
193 Takip Edilen

Moved inference over to Fireworks AI because the API is more reliable. Getting about 162 toks on concurrent requests with llama 3.1 8B model, plus this allows for 100k+ context. Structured generation is also supported which makes the citation verification approach more