Where's the RAGColbertReranker usage in llama-cpp-agent?

[svjack]

Thanks this bravo project, that can give easy format constrain of llama-cpp.
And where's the RAGColbertReranker usage ?
That I can try a usage example on agent ability with the help of llama-cpp-agent. 😊

[Maximilian-Winter]

I'm gonna add an example today, I have to add some comments first.

[Maximilian-Winter]

@svjack Added example in readme and in the examples folder.

[svjack]

@svjack Added example in readme and in the examples folder.

I try the demo in readme Structured Output
by replace main_model by qwen1.5 14b

import llama_cpp
import llama_cpp.llama_tokenizer

main_model = llama_cpp.Llama.from_pretrained(
    repo_id="Qwen/Qwen1.5-14B-Chat-GGUF",
    filename="*q4_0.gguf",
    tokenizer=llama_cpp.llama_tokenizer.LlamaHFTokenizer.from_pretrained("Qwen/Qwen1.5-14B"),
    verbose=False,
    n_gpu_layers = -1,
    n_ctx = 3060
)

This yield error of

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[22], line 4
      1 structured_output_agent = StructuredOutputAgent(llama, debug_output=True)
      3 text = """The Feynman Lectures on Physics is a physics textbook based on some lectures by Richard Feynman, a Nobel laureate who has sometimes been called "The Great Explainer". The lectures were presented before undergraduate students at the California Institute of Technology (Caltech), during 1961–1963. The book's co-authors are Feynman, Robert B. Leighton, and Matthew Sands."""
----> 4 print(structured_output_agent.create_object(Book, text))

File /environment/miniconda3/lib/python3.10/site-packages/llama_cpp_agent/structured_output_agent.py:215, in StructuredOutputAgent.create_object(self, model, data)
    204 """
    205 Creates an object of the given model from the given data.
    206 
   (...)
    212     object: The created object.
    213 """
    214 if model not in self.grammar_cache:
--> 215     grammar, documentation = generate_gbnf_grammar_and_documentation(
    216         [model],
    217         model_prefix="Response Model",
    218         fields_prefix="Response Model Field",
    219     )
    221     self.grammar_cache[model] = grammar, documentation
    222 else:

File /environment/miniconda3/lib/python3.10/site-packages/llama_cpp_agent/gbnf_grammar_generator/gbnf_grammar_from_pydantic_models.py:1451, in generate_gbnf_grammar_and_documentation(pydantic_model_list, outer_object_name, outer_object_content, model_prefix, fields_prefix, list_of_outputs, documentation_with_field_description, add_inner_thoughts, allow_only_inner_thoughts, inner_thoughts_field_name, add_request_heartbeat, request_heartbeat_field_name, request_heartbeat_models)
   1415 def generate_gbnf_grammar_and_documentation(
   1416     pydantic_model_list,
   1417     outer_object_name: str | None = None,
   (...)
   1428     request_heartbeat_models: List[str] = None,
   1429 ):
   1430     """
   1431     Generate GBNF grammar and documentation for a list of Pydantic models.
   1432 
   (...)
   1449         tuple: GBNF grammar string, documentation string.
   1450     """
-> 1451     documentation = generate_text_documentation(
   1452         copy(pydantic_model_list),
   1453         model_prefix,
   1454         fields_prefix,
   1455         documentation_with_field_description=documentation_with_field_description,
   1456     )
   1457     grammar = generate_gbnf_grammar_from_pydantic_models(
   1458         pydantic_model_list,
   1459         outer_object_name,
   (...)
   1467         request_heartbeat_models,
   1468     )
   1469     grammar = remove_empty_lines(grammar + get_primitive_grammar(grammar))

File /environment/miniconda3/lib/python3.10/site-packages/llama_cpp_agent/gbnf_grammar_generator/gbnf_grammar_from_pydantic_models.py:1116, in generate_text_documentation(pydantic_models, model_prefix, fields_prefix, documentation_with_field_description)
   1112             if isclass(element_type) and issubclass(
   1113                 element_type, BaseModel
   1114             ):
   1115                 pyd_models.append((element_type, False))
-> 1116 if isclass(field_type) and issubclass(field_type, BaseModel):
   1117     pyd_models.append((field_type, False))
   1118 documentation += generate_field_text(
   1119     name,
   1120     field_type,
   1121     model,
   1122     documentation_with_field_description=documentation_with_field_description,
   1123 )

File /environment/miniconda3/lib/python3.10/abc.py:123, in ABCMeta.__subclasscheck__(cls, subclass)
    121 def __subclasscheck__(cls, subclass):
    122     """Override for issubclass(subclass, cls)."""
--> 123     return _abc_subclasscheck(cls, subclass)

TypeError: issubclass() arg 1 must be a class

But when I use case in
https://github.com/Maximilian-Winter/llama-cpp-agent/blob/master/examples/02_Structured_Output/book_dataset_creation.py

I get favorable output. Does this mean, you should update your readme ?

[Maximilian-Winter]

@svjack I can't reproduce the error. But could you send me your complete code that causes the error?

[svjack]

@svjack I can't reproduce the error. But could you send me your complete code that causes the error?

In Python 3.10.12

Install by

pip install llama-cpp-agent
pip install transformers
CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python --force-reinstall --upgrade --no-cache-dir

source code

import llama_cpp
import llama_cpp.llama_tokenizer

llama = llama_cpp.Llama.from_pretrained(
    repo_id="Qwen/Qwen1.5-14B-Chat-GGUF",
    filename="*q4_0.gguf",
    tokenizer=llama_cpp.llama_tokenizer.LlamaHFTokenizer.from_pretrained("Qwen/Qwen1.5-14B"),
    verbose=False,
    n_gpu_layers = -1,
    n_ctx = 3060
)


from enum import Enum

from llama_cpp import Llama
from pydantic import BaseModel, Field

from llama_cpp_agent.structured_output_agent import StructuredOutputAgent


# Example enum for our output model
class Category(Enum):
    Fiction = "Fiction"
    NonFiction = "Non-Fiction"


# Example output model
class Book(BaseModel):
    """
    Represents an entry about a book.
    """
    title: str = Field(..., description="Title of the book.")
    author: str = Field(..., description="Author of the book.")
    published_year: int = Field(..., description="Publishing year of the book.")
    keywords: list[str] = Field(..., description="A list of keywords.")
    category: Category = Field(..., description="Category of the book.")
    summary: str = Field(..., description="Summary of the book.")


structured_output_agent = StructuredOutputAgent(llama, debug_output=True)

text = """The Feynman Lectures on Physics is a physics textbook based on some lectures by Richard Feynman, a Nobel laureate who has sometimes been called "The Great Explainer". The lectures were presented before undergraduate students at the California Institute of Technology (Caltech), during 1961–1963. The book's co-authors are Feynman, Robert B. Leighton, and Matthew Sands."""
print(structured_output_agent.create_object(Book, text))

structured_output_agent.create_object yield the above error. 🤔

[Maximilian-Winter]

@svjack I tried your code and it worked for me, but it lead to an error in llama-cpp-python after creating the object. I have to investigate that.

[GTJoey]

When I run the structured output demo ,the same problem happened!!
my anaconda env: python ==3.10.12

Package Version

---

aiohttp 3.9.5
aiosignal 1.3.1
annotated-types 0.6.0
async-timeout 4.0.3
attrs 23.2.0
build 1.2.1
CacheControl 0.14.0
certifi 2024.2.2
charset-normalizer 3.3.2
cleo 2.1.0
colorama 0.4.6
crashtest 0.4.1
diskcache 5.6.3
distlib 0.3.8
docstring_parser 0.16
dulwich 0.21.7
fastjsonschema 2.19.1
filelock 3.13.4
frozenlist 1.4.1
idna 3.7
importlib_metadata 7.1.0
installer 0.7.0
jaraco.classes 3.4.0
Jinja2 3.1.4
keyring 24.3.1
llama-cpp-agent 0.1.4
llama_cpp_python 0.2.70
MarkupSafe 2.1.5
more-itertools 10.2.0
msgpack 1.0.8
multidict 6.0.5
numpy 1.26.4
packaging 24.0
pexpect 4.9.0
pip 23.3.1
pkginfo 1.10.0
platformdirs 4.2.1
poetry 1.8.2
poetry-core 1.9.0
poetry-plugin-export 1.7.1
ptyprocess 0.7.0
pydantic 2.7.1
pydantic_core 2.18.2
pyproject_hooks 1.0.0
pywin32-ctypes 0.2.2
rapidfuzz 3.8.1
requests 2.31.0
requests-toolbelt 1.0.0
setuptools 68.2.2
shellingham 1.5.4
tomli 2.0.1
tomlkit 0.12.4
trove-classifiers 2024.4.10
typing_extensions 4.11.0
urllib3 2.2.1
virtualenv 20.26.0
wheel 0.41.2
yarl 1.9.4
zipp 3.18.1