Add ability to provide label descriptions

[rlmelvin]

This is a feature request.
The ability to optionally provide category/label descriptions could enhance model outputs.
For example, a possible label of "positive" might come with a user-provided description of "a document with an overall positive tone."

[AndreasKarasenko]

You can simply override the prompt_template attribute and provide a custom prompt with descriptions yourself. You just have to ensure that the required fields are still in your custom prompt.

E.g.

from skllm.datasets import get_classification_dataset
from skllm.config import SKLLMConfig
from skllm.models.gpt.classification.zero_shot import ZeroShotGPTClassifier
from skllm.models.gpt.classification.few_shot import FewShotGPTClassifier

# Configure the credentials
SKLLMConfig.set_openai_key("<YOUR_KEY>")
SKLLMConfig.set_openai_org("<YOUR_ORGANIZATION_ID>")

# Define custom prompts
BASIC_SENTIMENT = """
What is the most likely sentiment (using these labels: {labels}) for the following review
### YOUR DESCRIPTION HERE? ###

Review: ```{x}```
"""

FEW_SHOT = """
What is the most likely sentiment (using these labels: {labels}) for the review at the end?
Use these training examples:
Training data:
{training_data}

What is the sentiment for the following review?
Review: ```{x}```
"""

# Load a demo dataset
X, y = get_classification_dataset() # labels: positive, negative, neutral

# Initialize the model and make the predictions
clf = ZeroShotGPTClassifier(model="gpt-4")
clf.prompt_template = BASIC_SENTIMENT

clf.fit(X,y)
clf.predict(X)