46 lines
2.0 KiB
Markdown
46 lines
2.0 KiB
Markdown
|
---
|
||
|
|
license: apache-2.0
|
||
|
|
---
|
||
|
|
|
||
|
|
# LM-cocktail Mistral 7B v1
|
||
|
|
|
||
|
|
|
||
|
|
This is a 50%-50% model of two best Mistral models
|
||
|
|
|
||
|
|
https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2
|
||
|
|
|
||
|
|
https://huggingface.co/xDAN-AI/xDAN-L1-Chat-RL-v1
|
||
|
|
|
||
|
|
both claimed to be better than chatgpt-3.5-turbo in almost all metrics.
|
||
|
|
|
||
|
|
# Alpaca Eval
|
||
|
|
|
||
|
|
I am thrilled to announce that ChatGPT has ranked LMCocktail 7B as the second best model next to GPT4 on AlpcaEval in my local community run, even greater than my previously best [LMCocktail-10.7B-v1](https://huggingface.co/Yhyu13/LMCocktail-10.7B-v1.git) model. You can also check the leaderboard at [./Alpaca_eval/chatgpt_fn_--LMCocktail-Mistral-7B-v1/](./Alpaca_eval/chatgpt_fn_--LMCocktail-Mistral-7B-v1/)
|
||
|
|
|
||
|
|
```
|
||
|
|
win_rate standard_error n_total avg_length
|
||
|
|
gpt4 73.79 1.54 805 1365
|
||
|
|
LMCocktail-7B-v1(new) 73.54 1.55 805 1870
|
||
|
|
LMCocktail-10.7B-v1(new) 73.45 1.56 804 1203
|
||
|
|
claude 70.37 1.60 805 1082
|
||
|
|
chatgpt 66.09 1.66 805 811
|
||
|
|
wizardlm-13b 65.16 1.67 805 985
|
||
|
|
vicuna-13b 64.10 1.69 805 1037
|
||
|
|
guanaco-65b 62.36 1.71 805 1249
|
||
|
|
oasst-rlhf-llama-33b 62.05 1.71 805 1079
|
||
|
|
alpaca-farm-ppo-human 60.25 1.72 805 803
|
||
|
|
falcon-40b-instruct 56.52 1.74 805 662
|
||
|
|
text_davinci_003 50.00 0.00 805 307
|
||
|
|
alpaca-7b 45.22 1.74 805 396
|
||
|
|
text_davinci_001 28.07 1.56 805 296
|
||
|
|
```
|
||
|
|
|
||
|
|
|
||
|
|
# Code
|
||
|
|
|
||
|
|
The LM-cocktail is novel technique for merging multiple models https://arxiv.org/abs/2311.13534
|
||
|
|
|
||
|
|
Code is backed up by this repo https://github.com/FlagOpen/FlagEmbedding.git
|
||
|
|
|
||
|
|
Merging scripts available under the [./scripts](./scripts) folder
|