llmware

1

2
from llmware.prompts import Prompt
3
from llmware.models import ModelCatalog
4

5
#   Registered by default
6
#   --dragon models:  dragon-mistral-7b-gguf | dragon-yi-6b-gguf | dragon-llama-7b-gguf
7
#   --The Bloke leading 7b chat models:  llama2-chat | openhermes | zephyr | starling
8

9

10
#   example 1 - how to use a default gguf model in llmware
11
def use_default_gguf_model():
12

13
    selected_gguf_model = "llmware/dragon-mistral-7b-gguf"
14
    prompter = Prompt().load_model(selected_gguf_model)
15

16
    response = prompter.prompt_main("How old am I?", context="I am 36 years old.")
17

18
    print("response: ", response)
19

20
    return response
21

22

23
response = use_default_gguf_model()
24

25

26
#   example 2 - how to use any GGUF model from The Bloke on HuggingFace
27
def register_gguf_model():
28

29
    prompter = Prompt()
30

31
    your_model_name = "my_gguf_model_1"
32
    hf_repo_name = "TheBloke/model_name"
33
    model_file = "abc.gguf"
34

35
    prompter.model_catalog.register_gguf_model(your_model_name,hf_repo_name, model_file, prompt_wrapper="open_chat")
36
    prompter.load_model(your_model_name)
37

38
    return 0
39

40

41
#   example 3 - how to use build-from-source custom/optimized llama.cpp
42
def build_your_own_llama_cpp_lib():
43

44
    import os
45
    os.environ["GGUF_CUSTOM_LIB_PATH"] = "/path/to/your/custom/lib"
46

47
    return 0
48

49