Fresh Alpasta, done Al Dente!
It's da logical choice! Now with a similar personality emulation quality to GPT4-X-Alpasta-30b!
Model Info:
ChanSung's Alpaca-LoRA-30B-elina merged with Open Assistant's second Finetune
Benchmarks:
Wikitext2: 4.662261962890625
PTB: 24.547462463378906
C4: 7.05504846572876
4bit:
Wikitext2: 5.016242980957031
PTB: 25.576189041137695
C4: 7.332120418548584
~ Thanks to askmyteapot for performing these benchmarks!
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 57.05 |
| ARC (25-shot) | 60.58 |
| HellaSwag (10-shot) | 81.81 |
| MMLU (5-shot) | 56.63 |
| TruthfulQA (0-shot) | 48.38 |
| Winogrande (5-shot) | 78.14 |
| GSM8K (5-shot) | 26.76 |
| DROP (3-shot) | 47.06 |