File size: 7,626 Bytes
e294914 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 |
import os
import logging
from abc import ABC
from typing import Any
from llm.utils.hf_interface import HFInterface
from llm.utils.config import config
from langchain_community.llms import HuggingFaceEndpoint
logger = logging.getLogger(__name__)
logger.setLevel(logging.ERROR) # because if something went wrong in execution, application can't be work anyway
file_handler = logging.FileHandler(
"logs/chelsea_llm_huggingfacehub.log") # for all modules here template for logs file is "llm/logs/chelsea_{module_name}_{dir_name}.log"
logger.setLevel(logging.INFO) # informed
formatted = logging.Formatter("%(asctime)s - %(levelname)s - %(message)s")
file_handler.setFormatter(formatted)
logger.addHandler(file_handler)
logger.info("Getting information from apimodel module")
_api = os.environ.get("HUGGINGFACEHUB_API_TOKEN")
class HF_Mistaril(HFInterface, ABC):
"""
This class represents an interface for the Mistaril large language model from Hugging Face.
It inherits from `HFInterface` (likely an interface from a Hugging Face library)
and `ABC` (for abstract base class) to enforce specific functionalities.
"""
def __init__(self):
"""
Initializer for the `HF_Mistaril` class.
- Retrieves configuration values for the Mistaril model from a `config` dictionary:
- `repo_id`: The ID of the repository containing the Mistaril model on Hugging Face.
- `max_length`: Maximum length of the generated text.
- `temperature`: Controls randomness in the generation process.
- `top_k`: Restricts the vocabulary used for generation.
- Raises a `ValueError` if the `api` key (presumably stored elsewhere) is missing.
- Creates an instance of `HuggingFaceEndpoint` using the retrieved configuration
and the `api` key.
"""
repo_id = config["HF_Mistrail"]["model"]
max_length = config["HF_Mistrail"]["max_new_tokens"]
temperature = config["HF_Mistrail"]["temperature"]
top_k = config["HF_Mistrail"]["top_k"]
if not _api:
raise ValueError(f"API key not provided {_api}")
self.llm = HuggingFaceEndpoint(
repo_id=repo_id, max_length=max_length, temperature=temperature, top_k=top_k, token=_api
)
def execution(self) -> Any:
"""
This method attempts to return the underlying `llm` (likely a language model object).
It wraps the retrieval in a `try-except` block to catch potential exceptions.
On success, it returns the `llm` object.
On failure, it logs an error message with the exception details using a logger
(assumed to be available elsewhere).
"""
try:
return self.llm # `invoke()`
except Exception as e:
logger.error("Something wrong with API or HuggingFaceEndpoint", exc_info=e)
print(f"Something wrong with API or HuggingFaceEndpoint: {e}")
def model_name(self):
"""
Simple method that returns the Mistaril model name from the configuration.
This can be useful for identifying the specific model being used.
"""
return config["HF_Mistrail"]["model"]
def __str__(self):
"""
Defines the string representation of the `HF_Mistaril` object for human readability.
It combines the class name and the model name retrieved from the `model_name` method
with an underscore separator.
"""
return f"{self.__class__.__name__}_{self.model_name()}"
def __repr__(self):
"""
Defines the representation of the `HF_Mistaril` object for debugging purposes.
It uses `hasattr` to check if the `llm` attribute is set.
- If `llm` exists, it returns a string like `HF_Mistaril(llm=HuggingFaceEndpoint(...))`,
showing the class name and the `llm` object information.
- If `llm` is not yet set (during initialization), it returns
`HF_Mistaril(llm=not initialized)`, indicating the state.
"""
llm_info = f"llm={self.llm}" if hasattr(self, 'llm') else 'llm=not initialized'
return f"{self.__class__.__name__}({llm_info})"
class HF_TinyLlama(HFInterface, ABC):
"""
This class represents an interface for the TinyLlama large language model from Hugging Face.
It inherits from `HFInterface` (likely an interface from a Hugging Face library)
and `ABC` (for abstract base class) to enforce specific functionalities.
"""
def __init__(self):
"""
Initializer for the `HF_TinyLlama` class.
- Retrieves configuration values for the Mistaril model from a `config` dictionary:
- `repo_id`: The ID of the repository containing the TinyLlama model on Hugging Face.
- `max_length`: Maximum length of the generated text.
- `temperature`: Controls randomness in the generation process.
- `top_k`: Restricts the vocabulary used for generation.
- Raises a `ValueError` if the `api` key (presumably stored elsewhere) is missing.
- Creates an instance of `HuggingFaceEndpoint` using the retrieved configuration
and the `api` key.
"""
repo_id = config["HF_TinyLlama"]["model"]
max_length = config["HF_TinyLlama"]["max_new_tokens"]
temperature = config["HF_TinyLlama"]["temperature"]
top_k = config["HF_TinyLlama"]["top_k"]
if not _api:
raise ValueError(f"API key not provided {_api}")
self.llm = HuggingFaceEndpoint(
repo_id=repo_id, max_length=max_length, temperature=temperature, top_k=top_k, token=_api
)
def execution(self) -> Any:
"""
This method attempts to return the underlying `llm` (likely a language model object).
It wraps the retrieval in a `try-except` block to catch potential exceptions.
On success, it returns the `llm` object.
On failure, it logs an error message with the exception details using a logger
(assumed to be available elsewhere).
"""
try:
return self.llm
except Exception as e:
logger.error("Something wrong with API or HuggingFaceEndpoint", exc_info=e)
print(f"Something wrong with API or HuggingFaceEndpoint: {e}")
def model_name(self):
"""
Simple method that returns the TinyLlama model name from the configuration.
This can be useful for identifying the specific model being used.
"""
return config["HF_TinyLlama"]["model"]
def __str__(self):
"""
Defines the string representation of the `HF_TinyLlama` object for human readability.
It combines the class name and the model name retrieved from the `model_name` method
with an underscore separator.
"""
return f"{self.__class__.__name__}_{self.model_name()}"
def __repr__(self):
"""
Defines the representation of the `HF_TinyLlama` object for debugging purposes.
It uses `hasattr` to check if the `llm` attribute is set.
- If `llm` exists, it returns a string like `HF_TinyLlama(llm=HuggingFaceEndpoint(...))`,
showing the class name and the `llm` object information.
- If `llm` is not yet set (during initialization), it returns
`HF_TinyLlama(llm=not initialized)`, indicating the state.
"""
llm_info = f"llm={self.llm}" if hasattr(self, 'llm') else 'llm=not initialized'
return f"{self.__class__.__name__}({llm_info})"
|