SupportedLLMs
Supported LLMs
- OpenAI
- Anthropic
- HuggingFace
- Azure
- Groq
- Meta
- Gemini
How to use LLMs
To use an LLM, you need to specify the llm
parameter in the Pipeline
class.
from continuous_eval.llms import LLMFactory
llm = LLMFactory.get("openai:gpt-4o-mini")
print( llm.run( { "system_prompt": "You are a helpful assistant.", "user_prompt": "What is the capital of France?", } ))
By default only OpenAI is supported. To use other LLMs, you need to install the corresponding package and register the provider.
Suppose we want to use Anthropic
’s Claude 3.5 Sonnet
.
poetry install -E anthropic# orpip install continuous-eval[anthropic]
Then, register the provider.
from continuous_eval.llms import LLMFactoryfrom continuous_eval.llms.anthropic import Anthropic
LLMFactory.register_provider("anthropic", Anthropic)llm = LLMFactory.get("anthropic:claude-3-5-sonnet-20241022")
alternatively, we can also register a specific model, for example with Azure
from continuous_eval.llms import LLMFactoryfrom continuous_eval.llms.azure_openai import AzureOpenAIFactory
LLMFactory.register_provider( "azure_openai", model="gpt-4o-mini", provider_class=AzureOpenAIFactory( api_key=os.getenv("AZURE_OPENAI_API_KEY"), api_version=os.getenv("AZURE_OPENAI_API_VERSION"), endpoint=os.getenv("AZURE_OPENAI_ENDPOINT"), deployment=os.getenv("AZURE_OPENAI_DEPLOYMENT"), ),)llm = LLMFactory.get("azure_openai:gpt-4o-mini")