Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- openbmb/UltraFeedback
|
| 5 |
+
- openbmb/UltraInteract_pair
|
| 6 |
+
tags:
|
| 7 |
+
- reasoning
|
| 8 |
+
- preference_learning
|
| 9 |
+
- kto
|
| 10 |
+
pipeline_tag: text-generation
|
| 11 |
+
base_model: openbmb/Eurus-7b-kto
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# openbmb/Eurus-7b-kto-GGUF
|
| 15 |
+
|
| 16 |
+
- This is quantized version of [openbmb/Eurus-7b-kto](https://huggingface.co/openbmb/Eurus-7b-kto)
|
| 17 |
+
|
| 18 |
+
# Model Description
|
| 19 |
+
|
| 20 |
+
Eurus-7B-KTO is [KTO](https://arxiv.org/abs/2402.01306) fine-tuned from [Eurus-7B-SFT](https://huggingface.co/openbmb/Eurus-7b-sft) on all multi-turn trajectory pairs in [UltraInteract](https://huggingface.co/openbmb/UltraInteract) and all pairs in [UltraFeedback](https://huggingface.co/openbmb/UltraFeedback).
|
| 21 |
+
|
| 22 |
+
It achieves the best overall performance among open-source models of similar sizes and even outperforms specialized models in corresponding domains in many cases. Notably, Eurus-7B-KTO outperforms baselines that are 5× larger.
|
| 23 |
+
|
| 24 |
+
## Usage
|
| 25 |
+
|
| 26 |
+
We apply tailored prompts for coding and math, consistent with UltraInteract data formats:
|
| 27 |
+
|
| 28 |
+
**Coding**
|
| 29 |
+
|
| 30 |
+
```
|
| 31 |
+
[INST] Write Python code to solve the task:
|
| 32 |
+
{Instruction} [/INST]
|
| 33 |
+
```
|
| 34 |
+
**Math-CoT**
|
| 35 |
+
|
| 36 |
+
```
|
| 37 |
+
[INST] Solve the following math problem step-by-step.
|
| 38 |
+
Simplify your answer as much as possible. Present your final answer as \\boxed{Your Answer}.
|
| 39 |
+
{Instruction} [/INST]
|
| 40 |
+
```
|
| 41 |
+
|
| 42 |
+
**Math-PoT**
|
| 43 |
+
|
| 44 |
+
```
|
| 45 |
+
[INST] Tool available:
|
| 46 |
+
[1] Python interpreter
|
| 47 |
+
When you send a message containing Python code to python, it will be executed in a stateful Jupyter notebook environment.
|
| 48 |
+
Solve the following math problem step-by-step.
|
| 49 |
+
Simplify your answer as much as possible.
|
| 50 |
+
{Instruction} [/INST]
|
| 51 |
+
```
|
| 52 |
+
|
| 53 |
+
## Evaluation
|
| 54 |
+
- Eurus, both the 7B and 70B variants, achieve the best overall performance among open-source models of similar sizes. Eurus even outperforms specialized models in corresponding domains in many cases. Notably, Eurus-7B outperforms baselines that are 5× larger, and Eurus-70B achieves better performance than GPT-3.5 Turbo.
|
| 55 |
+
- Preference learning with UltraInteract can further improve performance, especially in math and the multi-turn ability.
|
| 56 |
+
<img src="figures_main_exp.png" alt="stats" style="zoom: 40%;" />
|
| 57 |
+
|