You can chat with an instruction-tuned model via cnlp_llm chat.
You can choose which model to chat with by passing the --model or -m flag to cnlp_llm chat with a valid model identifier.
If the --model flag is not set, it will default to the model identifier given by the environment variable CNLP_CHAT_MODEL.
| Provider | Example | Notes |
|---|---|---|
| Huggingface Hub | hf/TinyLlama/TinyLlama-1.1B-Chat-v1.0 |
|
| Huggingface Local Model | hf/local |
Specify local model path with -M model_path=./path/to/model |
| OpenAI | openai/gpt-3.5-turbo |
Set API key with OPENAI_API_KEY environment variable |
| Ollama | ollama/llama3.1 |
Set Ollama host address with OLLAMA_BASE_URL environment variable; e.g. http://localhost:11434/v1 |