ReMM-v2.2-L2-13B / README.md
Undi95's picture
Adding Evaluation Results (#1)
899ae09
metadata
license: cc-by-nc-4.0

Re:MythoMax v2.2 (ReMM v2.2) is a recreation trial of the original MythoMax-L2-B13 with updated models.

This merge use SLERP merging method to merge ReML v2.2 and Huginn v1.2.

Explaination :

- ReML-v2.2: (Chronos-Beluga v2/Hermes/Airboros 2.2)
=> Keeping The-Face-Of-Goonery/Chronos-Beluga-v2-13bfp16
=> Replacing jondurbin/airoboros-l2-13b-2.2 by jondurbin/airoboros-l2-13b-2.2.1 (last version)
=> Keeping NousResearch/Nous-Hermes-Llama2-13b

With that :
- ReMM-v2.2: (ReML/Huginn v1.2)
=> Replacing ReMM by the one above (ReML v2.1)
=> Keeping The-Face-Of-Goonery/Huginn-13b-v1.2 (hottest)

Description

This repo contains fp16 files of ReMM v2.1, a recreation of the original MythoMax, but updated and merged with SLERP.

Models used

  • The-Face-Of-Goonery/Chronos-Beluga-v2-13bfp16
  • jondurbin/airoboros-l2-13b-2.2.1
  • NousResearch/Nous-Hermes-Llama2-13b
  • The-Face-Of-Goonery/Huginn-13b-v1.2
  • ReML-v2.1-L2-13B (Private recreation trial of an updated Mythologic-L2-13B)

Prompt template: Alpaca

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{prompt}

### Response:

Special thanks to Sushi.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 50.45
ARC (25-shot) 61.26
HellaSwag (10-shot) 84.16
MMLU (5-shot) 56.22
TruthfulQA (0-shot) 51.35
Winogrande (5-shot) 75.61
GSM8K (5-shot) 14.03
DROP (3-shot) 10.56