script

application/api/answer/routes.py

@@ -32,20 +32,6 @@ if settings.LLM_NAME == "gpt4":
 else:
     gpt_model = 'gpt-3.5-turbo'
 
-if settings.SELF_HOSTED_MODEL:
-    from langchain.llms import HuggingFacePipeline
-    from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
-
-    model_id = settings.LLM_NAME  # hf model id (Arc53/docsgpt-7b-falcon, Arc53/docsgpt-14b)
-    tokenizer = AutoTokenizer.from_pretrained(model_id)
-    model = AutoModelForCausalLM.from_pretrained(model_id)
-    pipe = pipeline(
-        "text-generation", model=model,
-        tokenizer=tokenizer, max_new_tokens=2000,
-        device_map="auto", eos_token_id=tokenizer.eos_token_id
-    )
-    hf = HuggingFacePipeline(pipeline=pipe)
-
 # load the prompts
 current_dir = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 with open(os.path.join(current_dir, "prompts", "combine_prompt.txt"), "r") as f:

application/core/settings.py

@@ -1,6 +1,8 @@
 from pathlib import Path
+import os
 
 from pydantic import BaseSettings
+current_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
 
 
 class Settings(BaseSettings):
@@ -9,9 +11,8 @@ class Settings(BaseSettings):
     CELERY_BROKER_URL: str = "redis://localhost:6379/0"
     CELERY_RESULT_BACKEND: str = "redis://localhost:6379/1"
     MONGO_URI: str = "mongodb://localhost:27017/docsgpt"
-    MODEL_PATH: str = "./models/gpt4all-model.bin"
+    MODEL_PATH: str = os.path.join(current_dir, "models/docsgpt-7b-f16.gguf")
     TOKENS_MAX_HISTORY: int = 150
-    SELF_HOSTED_MODEL: bool = False
     UPLOAD_FOLDER: str = "inputs"
 
     API_URL: str = "http://localhost:7091"  # backend url for celery worker

application/llm/llama_cpp.py

@@ -4,7 +4,10 @@ class LlamaCpp(BaseLLM):
 
     def __init__(self, api_key, llm_name='/Users/pavel/Desktop/docsgpt/application/models/orca-test.bin'):
         global llama
-        from llama_cpp import Llama
+        try:
+            from llama_cpp import Llama
+        except ImportError:
+            raise ImportError("Please install llama_cpp using pip install llama-cpp-python")
 
         llama = Llama(model_path=llm_name)
 

application/vectorstore/base.py

@@ -2,7 +2,7 @@ from abc import ABC, abstractmethod
 import os
 from langchain.embeddings import (
     OpenAIEmbeddings,
-    HuggingFaceHubEmbeddings,
+    HuggingFaceEmbeddings,
     CohereEmbeddings,
     HuggingFaceInstructEmbeddings,
 )
@@ -22,7 +22,7 @@ class BaseVectorStore(ABC):
     def _get_docsearch(self, embeddings_name, embeddings_key=None):
         embeddings_factory = {
             "openai_text-embedding-ada-002": OpenAIEmbeddings,
-            "huggingface_sentence-transformers/all-mpnet-base-v2": HuggingFaceHubEmbeddings,
+            "huggingface_sentence-transformers/all-mpnet-base-v2": HuggingFaceEmbeddings,
             "huggingface_hkunlp/instructor-large": HuggingFaceInstructEmbeddings,
             "cohere_medium": CohereEmbeddings
         }