Percival 01 7b Slerp

🚀 Percival_01-7b-slerp

Percival_01-7b-slerp是一个强大的模型，在OPENLLM排行榜上位列7B模型第二！它通过LazyMergekit将以下两个模型进行合并：

• liminerity/M7-7b
• Gille/StrangeMerges_32-7B-slerp

🚀 快速开始

本部分将介绍如何快速使用Percival_01-7b-slerp模型。

安装依赖

!pip install -qU transformers accelerate

代码示例

from transformers import AutoTokenizer
import transformers
import torch

model = "AurelPx/Percival_01-7b-slerp"
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"])

✨ 主要特性

• 模型合并：通过LazyMergekit将liminerity/M7-7b和Gille/StrangeMerges_32-7B-slerp两个模型进行合并。
• 高性能：在OPENLLM排行榜上位列7B模型第二。

📦 安装指南

运行以下命令安装所需的库：

!pip install -qU transformers accelerate

💻 使用示例

基础用法

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "AurelPx/Percival_01-7b-slerp"
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"])

📚 详细文档

🧩 配置信息

以下是模型合并的配置信息：

slices:
  - sources:
      - model: liminerity/M7-7b
        layer_range: [0, 32]
      - model: Gille/StrangeMerges_32-7B-slerp
        layer_range: [0, 32]
merge_method: slerp
base_model: liminerity/M7-7b
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

📄 许可证

本项目采用Apache-2.0许可证。