File size: 2,286 Bytes
75251ab
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
31a9f7b
75251ab
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
---
license: apache-2.0
datasets:
- EleutherAI/muInstruct
- camel-ai/math
language:
- en
tags:
- math
---

`llemma_7b_muinstruct_camelmath` is an instruction-following finetune of [Llemma 7B](https://huggingface.co/EleutherAI/llemma_7b), trained on the [μInstruct](https://huggingface.co/datasets/EleutherAI/muInstruct) and [camel-ai/math](https://huggingface.co/datasets/camel-ai/math) datasets.

## Input Formatting
Format input queries as follows:
```
input_text = f"Input:{input}\n\nResponse:"
```

Note that due to an error during training, this model's end-of-sequence token ID is `0` instead of the `2` which is standard for Llama-2 based models. Inference APIs should handle this automatically by reading this repo's `config.json`, but be aware of this difference if you are doing token surgery.

## Evals
`
llemma_7b_muinstruct_camelmath` compares favorably to other 7B parameter models on the [Hungarian Math Exam](https://huggingface.co/datasets/keirp/hungarian_national_hs_finals_exam/blob/main/README.md). It surpasses the few-shot performance of Llemma 7B whilst being the strongest Llama-2 7B based model. 

| Model                                                                          | Exam Score |
| ------------------------------------------------------------------------------ | ---------- |
| [Code Llama 7B](https://huggingface.co/codellama/CodeLlama-7b-hf) (few-shot)   | 8\%        |
| [MetaMath 7B](https://huggingface.co/meta-math/MetaMath-7B-V1.0)               | 20\%       |
| [MAmmoTH 7B](https://huggingface.co/TIGER-Lab/MAmmoTH-7B)                      | 17\%       | 
| [MAmmoTH Coder 7B](https://huggingface.co/TIGER-Lab/MAmmoTH-Coder-7B)          | 11\%       | 
| [Llemma 7B](https://huggingface.co/EleutherAI/llemma_7b) (few-shot)            | 23\%       | 
| Llemma_7B_muinstruct_camelmath                                                 | 25\%       | 
| -                                                                              | -          | 
| [Mistral 7B](https://huggingface.co/mistralai/Mistral-7B-v0.1) (few-shot)      | 22\%       | 
| [MetaMath Mistral 7B](https://huggingface.co/meta-math/MetaMath-Mistral-7B)    | 29\%       | 
| [OpenChat 3.5](https://huggingface.co/openchat/openchat_3.5)                   | 37\%       |