236 lines
7.7 KiB
Markdown
236 lines
7.7 KiB
Markdown
---
|
|
language:
|
|
- en
|
|
license: apache-2.0
|
|
library_name: transformers
|
|
tags:
|
|
- merge
|
|
- mergekit
|
|
- lazymergekit
|
|
inference: false
|
|
base_model:
|
|
- senseable/Westlake-7B
|
|
- Guilherme34/Samantha-v2
|
|
- uukuguy/speechless-mistral-six-in-one-7b
|
|
pipeline_tag: text-generation
|
|
model-index:
|
|
- name: sethuiyer/Nandine-7b
|
|
results:
|
|
- task:
|
|
type: text-generation
|
|
name: Text Generation
|
|
dataset:
|
|
name: AI2 Reasoning Challenge (25-Shot)
|
|
type: ai2_arc
|
|
config: ARC-Challenge
|
|
split: test
|
|
args:
|
|
num_few_shot: 25
|
|
metrics:
|
|
- type: acc_norm
|
|
value: 69.28
|
|
name: normalized accuracy
|
|
source:
|
|
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Nandine-7b
|
|
name: Open LLM Leaderboard
|
|
- task:
|
|
type: text-generation
|
|
name: Text Generation
|
|
dataset:
|
|
name: HellaSwag (10-Shot)
|
|
type: hellaswag
|
|
split: validation
|
|
args:
|
|
num_few_shot: 10
|
|
metrics:
|
|
- type: acc_norm
|
|
value: 87.01
|
|
name: normalized accuracy
|
|
source:
|
|
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Nandine-7b
|
|
name: Open LLM Leaderboard
|
|
- task:
|
|
type: text-generation
|
|
name: Text Generation
|
|
dataset:
|
|
name: MMLU (5-Shot)
|
|
type: cais/mmlu
|
|
config: all
|
|
split: test
|
|
args:
|
|
num_few_shot: 5
|
|
metrics:
|
|
- type: acc
|
|
value: 64.83
|
|
name: accuracy
|
|
source:
|
|
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Nandine-7b
|
|
name: Open LLM Leaderboard
|
|
- task:
|
|
type: text-generation
|
|
name: Text Generation
|
|
dataset:
|
|
name: TruthfulQA (0-shot)
|
|
type: truthful_qa
|
|
config: multiple_choice
|
|
split: validation
|
|
args:
|
|
num_few_shot: 0
|
|
metrics:
|
|
- type: mc2
|
|
value: 62.1
|
|
source:
|
|
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Nandine-7b
|
|
name: Open LLM Leaderboard
|
|
- task:
|
|
type: text-generation
|
|
name: Text Generation
|
|
dataset:
|
|
name: Winogrande (5-shot)
|
|
type: winogrande
|
|
config: winogrande_xl
|
|
split: validation
|
|
args:
|
|
num_few_shot: 5
|
|
metrics:
|
|
- type: acc
|
|
value: 83.19
|
|
name: accuracy
|
|
source:
|
|
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Nandine-7b
|
|
name: Open LLM Leaderboard
|
|
- task:
|
|
type: text-generation
|
|
name: Text Generation
|
|
dataset:
|
|
name: GSM8k (5-shot)
|
|
type: gsm8k
|
|
config: main
|
|
split: test
|
|
args:
|
|
num_few_shot: 5
|
|
metrics:
|
|
- type: acc
|
|
value: 62.4
|
|
name: accuracy
|
|
source:
|
|
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Nandine-7b
|
|
name: Open LLM Leaderboard
|
|
---
|
|
|
|
# Nandine-7b
|
|
|
|
<p align="center">
|
|
<img src="https://huggingface.co/sethuiyer/Nandine-7b/resolve/main/nandine.webp" height="128px" alt="Nandine">
|
|
</p>
|
|
|
|
This is Nandine-7b, rated **87.47/100** by GPT-4 on a collection of 30 synthetic prompts generated by GPT-4.
|
|
|
|
Nandine-7b is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
|
|
* [senseable/Westlake-7B](https://huggingface.co/senseable/Westlake-7B)
|
|
* [Guilherme34/Samantha-v2](https://huggingface.co/Guilherme34/Samantha-v2)
|
|
* [uukuguy/speechless-mistral-six-in-one-7b](https://huggingface.co/uukuguy/speechless-mistral-six-in-one-7b)
|
|
|
|
Nandine-7b represents a harmonious amalgamation of narrative skill, empathetic interaction, intellectual depth, and eloquent communication.
|
|
|
|
## OpenLLM Benchmark
|
|
|
|
| Model | Average ⬆️ | ARC | HellaSwag | MMLU | TruthfulQA | Winogrande | GSM8K |
|
|
|--------------------------------|------------|-------|-----------|-------|------------|------------|-------|
|
|
| sethuiyer/Nandine-7b 📑 | 71.47 | 69.28 | 87.01 | 64.83 | 62.1 | 83.19 | 62.4 |
|
|
|
|
|
|
## Nous Benchmark
|
|
| Model |AGIEval|GPT4All|TruthfulQA|Bigbench|Average|
|
|
|---------------------------------------------------------|------:|------:|---------:|-------:|------:|
|
|
|[Nandine-7b](https://huggingface.co/sethuiyer/Nandine-7b)| 43.54| 76.41| 61.73| 45.27| 56.74|
|
|
|
|
For more details, refer [here](https://huggingface.co/sethuiyer/Nandine-7b/blob/main/EVAL.md)
|
|
|
|
|
|
**Pros:**
|
|
1. **Strong Narrative Skills:** Excels in storytelling, creating engaging and imaginative narratives.
|
|
2. **Accurate Information Delivery:** Provides factual and detailed information across various topics.
|
|
3. **Comprehensive Analysis:** Capable of well-rounded discussions on complex and ethical topics.
|
|
4. **Emotional Intelligence:** Shows empathy and understanding in responses requiring emotional sensitivity.
|
|
5. **Clarity and Structure:** Maintains clear and well-structured communication.
|
|
|
|
**Cons:**
|
|
1. **Language Translation Limitations:** Challenges in providing fluent and natural translations.
|
|
2. **Incomplete Problem Solving:** Some logical or mathematical problems are not solved accurately.
|
|
3. **Lack of Depth in Certain Areas:** Needs deeper exploration in some responses for a more comprehensive understanding.
|
|
4. **Occasional Imbalance in Historical Context:** Some historical explanations could be more balanced.
|
|
5. **Room for Enhanced Creativity:** While creative storytelling is strong, there's potential for more varied responses in hypothetical scenarios.
|
|
|
|
**Intended Use:**
|
|
Ideal for users seeking a versatile AI companion for creative writing, thoughtful discussions, and general assistance.
|
|
|
|
## 🧩 Configuration
|
|
|
|
```yaml
|
|
models:
|
|
- model: senseable/Westlake-7B
|
|
parameters:
|
|
weight: 0.55
|
|
density: 0.6
|
|
- model: Guilherme34/Samantha-v2
|
|
parameters:
|
|
weight: 0.10
|
|
density: 0.3
|
|
- model: uukuguy/speechless-mistral-six-in-one-7b
|
|
parameters:
|
|
weight: 0.35
|
|
density: 0.6
|
|
merge_method: dare_ties
|
|
base_model: mistralai/Mistral-7B-v0.1
|
|
parameters:
|
|
int8_mask: true
|
|
dtype: bfloat16
|
|
```
|
|
|
|
## 💻 Usage
|
|
|
|
```python
|
|
!pip install -qU transformers accelerate
|
|
|
|
from transformers import AutoTokenizer
|
|
import transformers
|
|
import torch
|
|
|
|
model = "sethuiyer/Nandine-7b"
|
|
messages = [{"role": "user", "content": "What is a large language model?"}]
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model)
|
|
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
|
pipeline = transformers.pipeline(
|
|
"text-generation",
|
|
model=model,
|
|
torch_dtype=torch.float16,
|
|
device_map="auto",
|
|
)
|
|
|
|
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
|
print(outputs[0]["generated_text"])
|
|
```
|
|
|
|
## GGUF
|
|
GGUF files are available at [Nandine-7b-GGUF](https://huggingface.co/sethuiyer/Nandine-7b-GGUF/tree/main)
|
|
|
|
## Ollama
|
|
Nandine is now available on Ollama. You can use it by running the command ```ollama run stuehieyr/nandine``` in your
|
|
terminal. If you have limited computing resources, check out this [video](https://www.youtube.com/watch?v=Qa1h7ygwQq8) to learn how to run it on
|
|
a Google Colab backend.
|
|
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
|
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_sethuiyer__Nandine-7b)
|
|
|
|
| Metric |Value|
|
|
|---------------------------------|----:|
|
|
|Avg. |71.47|
|
|
|AI2 Reasoning Challenge (25-Shot)|69.28|
|
|
|HellaSwag (10-Shot) |87.01|
|
|
|MMLU (5-Shot) |64.83|
|
|
|TruthfulQA (0-shot) |62.10|
|
|
|Winogrande (5-shot) |83.19|
|
|
|GSM8k (5-shot) |62.40|
|
|
|