palmer
a better base model
This model is palmer-002-2401 scaled to 32k by merging and fine-tuning with TinyLlama-1.1B-32k-Instruct by Doctor-Shotgun
evaluation 🧪
note that this is a zero-shot setting as opposite to open llm leaderboard's few-shot evals
model ARC-C OBQA HellaSwag PIQA Winogrande Average
tinyllama | 0.3029 | 0.3600 | 0.5935 | 0.7329 | 0.5959 | 0.5170 |
palmer-002-2401 | 0.3294 | 0.3700 | 0.5950 | 0.7399 | 0.5896 | 0.5247 |
palmer-002-32k | 0.3268 | 0.3780 | 0.5785 | 0.7492 | 0.6251 | 0.5315 | (this)
babbage-002 | 0.3285 | 0.3620 | 0.6380 | 0.7606 | 0.6085 | 0.5395 |
This model's performance is close to openai's one while being capable of using 2x the context size.
prompt 📝
no prompt 🚀