stormtrooper

突击队员

Scikit-Learn管道的零/少数射击学习组件，具有大型模型和变压器。

文档

新的1.0.0

Trooper

全新的Trooper接口使您无需指定要使用的型号类型。 Stormtrooper将自动从指定名称检测模型类型。

 from stormtrooper import Trooper

# This loads a setfit model
model = Trooper ( "all-MiniLM-L6-v2" )

# This loads an OpenAI model
model = Trooper ( "gpt-4" )

# This loads a Text2Text model
model = Trooper ( "google/flan-t5-base" )

统一的零和几乎没有分类

初始化时，您不再需要指定模型是少数还是零弹分的分类器。如果您不通过任何培训示例，将自动假定该模型应为零。

 # This is a zero-shot model
model . fit ( None , [ "dog" , "cat" ])

# This is a few-shot model
model . fit ([ "he was a good boy" , "just lay down on my laptop" ], [ "dog" , "cat" ])

型号类型

您可以在Stormtroper中使用各种变压器模型进行少量和零射击分类。

1. 指令微调生成模型，例如Trooper("HuggingFaceH4/zephyr-7b-beta")
2. 带有SetFit的编码器模型，例如Trooper("all-MiniLM-L6-v2")
3. Text2Text模型例如Trooper("google/flan-t5-base")
4. Openai型号，例如Trooper("gpt-4")
5. NLI模型例如Trooper("facebook/bart-large-mnli")

示例用法

在我们的文档中找到更多。

pip install stormtrooper

 from stormtrooper import Trooper

class_labels = [ "atheism/christianity" , "astronomy/space" ]
example_texts = [
    "God came down to earth to save us." ,
    "A new nebula was recently discovered in the proximity of the Oort cloud."
]
new_texts = [ "God bless the reailway workers" , "The frigate is ready to launch from the spaceport" ]

# Zero-shot classification
model = Trooper ( "google/flan-t5-base" )
model . fit ( None , class_labels )
model . predict ( new_texts )
# ["atheism/christianity", "astronomy/space"]

# Few-shot classification
model = Trooper ( "google/flan-t5-base" )
model . fit ( example_texts , class_labels )
model . predict ( new_texts )
# ["atheism/christianity", "astronomy/space"]

模糊匹配

默认情况下，生成和text2Text模型将模糊的结果与最接近的类标签匹配，您可以通过指定fuzzy_match=False禁用此行为。

如果您想要模糊的匹配加速，则应安装python-Levenshtein 。

推断GPU

从0.2.2版本，您可以在GPU上运行型号。您可以在初始化模型时指定设备：

 classifier = Trooper ( "all-MiniLM-L6-v2" , device = "cuda:0" )

对多个GPU的推断

您可以使用device_map参数在多个设备优先级GPU -> CPU + Ram -> Disk以及多个设备上运行模型。请注意，这仅适用于Text2Text和生成模型。

 model = Trooper("HuggingFaceH4/zephyr-7b-beta", device_map="auto")