Usable Models
Collection
5 items
•
Updated
•
2
Five different instruction-tuned models (which I'm sure are intuitively obvious from the name) merged using the methodology described in Resolving Interference When Merging Models.
In theory this should retain more of the capabilites of the constituent models than a straight linear merge would. In my testing, it feels quite capable.
Base model used for the merge: TheBloke/Llama-2-13B-fp16
Models merged in:
- OpenOrca-Platypus2-13B
- limarp-13b-merged
- Nous-Hermes-Llama2-13b
- chronos-13b-v2
- airoboros-l2-13b-gpt4-1.4.1
Works quite well with Alpaca-style prompts:
### Instruction:
...
### Response:
The script I used to perform the merge is available here.
The command that produced this model:
python ties_merge.py TheBloke/Llama-2-13B-fp16 ./Chronorctypus-Limarobormes-13b --merge elinas/chronos-13b-v2 --merge Open-Orca/OpenOrca-Platypus2-13B --merge Oniichat/limarp-13b-merged --merge jondurbin/airoboros-l2-13b-gpt4-1.4.1 --merge NousResearch/Nous-Hermes-Llama2-13b --cuda
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 49.88 |
| ARC (25-shot) | 59.9 |
| HellaSwag (10-shot) | 82.75 |
| MMLU (5-shot) | 58.45 |
| TruthfulQA (0-shot) | 51.9 |
| Winogrande (5-shot) | 74.43 |
| GSM8K (5-shot) | 3.87 |
| DROP (3-shot) | 17.89 |
- Downloads last month
- 1,167
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.