Implementação do RAG.

.gitignore

@@ -0,0 +1,3 @@
+.idea
+venv
+.env

README.md

@@ -1,10 +1,10 @@
 ---
 title: Ross Gpt
-emoji: 🏃
-colorFrom: red
-colorTo: pink
+emoji: 🌍
+colorFrom: yellow
+colorTo: gray
 sdk: streamlit
-sdk_version: 1.41.0
+sdk_version: 1.40.2
 app_file: app.py
 pinned: false
 ---

app.py

@@ -0,0 +1,136 @@
+import streamlit as st
+import multiprocessing
+from langchain.docstore.document import Document as LangChainDocument
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_community.vectorstores import FAISS
+from huggingface_hub import login
+from loguru import logger
+import os
+from dotenv import load_dotenv
+
+multiprocessing.freeze_support()
+load_dotenv()
+access_token = os.getenv("ACCESS_TOKEN")
+login(token=access_token)
+logger.info('Login realizado com sucesso.')
+
+logger.info('Carregando arquivo no qual será baseado o RAG.')
+with open('train.txt', 'r') as f:
+    data = f.read()
+
+logger.info('Representando o documento utilizando o LangChainDocument.')
+raw_database = LangChainDocument(page_content=data)
+
+MARKDOWN_SEPARATORS = [
+    "\n#{1,6} ",
+    "```\n",
+    "\n\\*\\*\\*+\n",
+    "\n---+\n",
+    "\n___+\n",
+    "\n\n",
+    "\n",
+    " ",
+    "",
+]
+
+logger.info('Quebrando o documento para a criação dos chunks.')
+splitter = RecursiveCharacterTextSplitter(separators=MARKDOWN_SEPARATORS, chunk_size=1000, chunk_overlap=100)
+process_data = splitter.split_documents([raw_database])
+process_data = process_data[:5]  # TODO: REMOVER DEPOIS
+
+embedding_model_name = "thenlper/gte-small"
+logger.info(f'Definição do modelo de embeddings: {embedding_model_name}.')
+embedding_model = HuggingFaceEmbeddings(
+    model_name=embedding_model_name,
+    multi_process=True,
+    model_kwargs={"device": "cuda"},
+    encode_kwargs={"normalize_embeddings": True},  # Set `True` for cosine similarity
+)
+
+logger.info('Criação da base de dados vetorial (em memória).')
+vectors = FAISS.from_documents(process_data, embedding_model)
+
+from transformers import pipeline
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
+
+# model_name = "meta-llama/Llama-3.2-1B"
+model_name = "HuggingFaceH4/zephyr-7b-beta"
+# model_name = "mistralai/Mistral-7B-Instruct-v0.3"
+# model_name = "meta-llama/Llama-3.2-3B-Instruct"
+logger.info(f'Carregamento do modelo de linguagem principal: {model_name}')
+
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_use_double_quant=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.bfloat16,
+)
+model = AutoModelForCausalLM.from_pretrained(model_name, quantization_config=bnb_config)
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+llm_model = pipeline(
+    model=model,
+    tokenizer=tokenizer,
+    task="text-generation",
+    do_sample=True,
+    temperature=0.4,
+    repetition_penalty=1.1,
+    return_full_text=False,
+    max_new_tokens=500
+)
+logger.info(f'Modelo {model_name} carregado com sucesso.')
+
+prompt = """
+<|system|>
+You are a helpful assistant that answers on medical questions based on the real information provided from different sources and in the context.
+Give the rational and well written response. If you don't have proper info in the context, answer "I don't know"
+Respond only to the question asked.
+
+<|user|>
+Context:
+{}
+---
+Here is the question you need to answer.
+
+Question: {}
+---
+<|assistant|>
+"""
+
+st.title("Echo Bot")
+
+if "messages" not in st.session_state:
+    st.session_state.messages = []
+
+for message in st.session_state.messages:
+    with st.chat_message(message["role"]):
+        st.markdown(message["content"])
+
+question = st.chat_input("How can I help you?")
+if question:
+    with st.chat_message("user"):
+        st.markdown(prompt)
+
+    st.session_state.messages.append({"role": "user", "content": prompt})
+
+    search_results = vectors.similarity_search(question, k=3)
+
+    logger.info('Contexto: ')
+    for i, search_result in enumerate(search_results):
+        logger.info(f"{i + 1}) {search_result.page_content}")
+
+    context = " ".join([search_result.page_content for search_result in search_results])
+    final_prompt = prompt.format(context, question)
+    logger.info(f'\n{final_prompt}\n')
+
+    answer = llm_model(final_prompt)
+    text_answer = answer[0]['generated_text']
+
+    logger.info("AI response: ", text_answer)
+
+    with st.chat_message("assistant"):
+        st.markdown(text_answer)
+
+    st.session_state.messages.append({"role": "assistant", "content": text_answer})

rag_test.py

@@ -0,0 +1,118 @@
+import multiprocessing
+from langchain.docstore.document import Document as LangChainDocument
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain_community.vectorstores import FAISS
+from huggingface_hub import login
+from loguru import logger
+from transformers import pipeline
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
+import os
+from dotenv import load_dotenv
+
+
+def main():
+    load_dotenv()
+    logger.info('Carregando arquivo no qual será baseado o RAG.')
+    with open('train.txt', 'r') as f:
+        data = f.read()
+
+    logger.info('Representando o documento utilizando o LangChainDocument.')
+    raw_database = LangChainDocument(page_content=data)
+
+    MARKDOWN_SEPARATORS = [
+        "\n#{1,6} ",
+        "```\n",
+        "\n\\*\\*\\*+\n",
+        "\n---+\n",
+        "\n___+\n",
+        "\n\n",
+        "\n",
+        " ",
+        "",
+    ]
+
+    logger.info('Quebrando o documento para a criação dos chunks.')
+    splitter = RecursiveCharacterTextSplitter(separators=MARKDOWN_SEPARATORS, chunk_size=1000, chunk_overlap=100)
+    process_data = splitter.split_documents([raw_database])
+    process_data = process_data[:5]  # TODO: REMOVER DEPOIS
+
+    embedding_model_name = "thenlper/gte-small"
+    logger.info(f'Definição do modelo de embeddings: {embedding_model_name}.')
+    embedding_model = HuggingFaceEmbeddings(
+        model_name=embedding_model_name,
+        multi_process=True,
+        model_kwargs={"device": "cuda"},
+        encode_kwargs={"normalize_embeddings": True},  # Set `True` for cosine similarity
+    )
+
+    logger.info('Criação da base de dados vetorial (em memória).')
+    vectors = FAISS.from_documents(process_data, embedding_model)
+
+    # model_name = "meta-llama/Llama-3.2-1B"
+    model_name = "HuggingFaceH4/zephyr-7b-beta"
+    # model_name = "mistralai/Mistral-7B-Instruct-v0.3"
+    # model_name = "meta-llama/Llama-3.2-3B-Instruct"
+    logger.info(f'Carregamento do modelo de linguagem principal: {model_name}')
+
+    bnb_config = BitsAndBytesConfig(
+        load_in_4bit=True,
+        bnb_4bit_use_double_quant=True,
+        bnb_4bit_quant_type="nf4",
+        bnb_4bit_compute_dtype=torch.bfloat16,
+    )
+    model = AutoModelForCausalLM.from_pretrained(model_name, quantization_config=bnb_config)
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+    llm_model = pipeline(
+        model=model,
+        tokenizer=tokenizer,
+        task="text-generation",
+        do_sample=True,
+        temperature=0.4,
+        repetition_penalty=1.1,
+        return_full_text=False,
+        max_new_tokens=500
+    )
+    logger.info(f'Modelo {model_name} carregado com sucesso.')
+
+    prompt = """
+    <|system|>
+    You are a helpful assistant that answers on medical questions based on the real information provided from different sources and in the context.
+    Give the rational and well written response. If you don't have proper info in the context, answer "I don't know"
+    Respond only to the question asked.
+
+    <|user|>
+    Context:
+    {}
+    ---
+    Here is the question you need to answer.
+
+    Question: {}
+    ---
+    <|assistant|>
+    """
+
+    question = "What is Cardiogenic shock?"
+    search_results = vectors.similarity_search(question, k=3)
+
+    logger.info('Contexto: ')
+    for i, search_result in enumerate(search_results):
+        logger.info(f"{i + 1}) {search_result.page_content}")
+
+    context = " ".join([search_result.page_content for search_result in search_results])
+    final_prompt = prompt.format(context, question)
+    logger.info(f'\n{final_prompt}\n')
+
+    answer = llm_model(final_prompt)
+
+    logger.info("AI response: ", answer[0]['generated_text'])
+
+
+if __name__ == '__main__':
+    multiprocessing.freeze_support()
+    access_token = os.getenv("ACCESS_TOKEN")
+    login(token=access_token)
+    logger.info('Login realizado com sucesso.')
+    main()

requirements.txt

@@ -0,0 +1,114 @@
+accelerate==1.2.0
+aiohappyeyeballs==2.4.4
+aiohttp==3.11.10
+aiosignal==1.3.1
+altair==5.5.0
+annotated-types==0.7.0
+anyio==4.7.0
+attrs==24.2.0
+bitsandbytes==0.45.0
+blinker==1.9.0
+cachetools==5.5.0
+certifi==2024.8.30
+charset-normalizer==3.4.0
+click==8.1.7
+dataclasses-json==0.6.7
+datasets==3.2.0
+dill==0.3.8
+filelock==3.16.1
+frozenlist==1.5.0
+fsspec==2024.9.0
+gitdb==4.0.11
+GitPython==3.1.43
+greenlet==3.1.1
+h11==0.14.0
+httpcore==1.0.7
+httpx==0.28.1
+httpx-sse==0.4.0
+huggingface-hub==0.26.5
+idna==3.10
+Jinja2==3.1.4
+joblib==1.4.2
+jsonpatch==1.33
+jsonpointer==3.0.0
+jsonschema==4.23.0
+jsonschema-specifications==2024.10.1
+langchain==0.3.11
+langchain-community==0.3.11
+langchain-core==0.3.24
+langchain-huggingface==0.1.2
+langchain-text-splitters==0.3.2
+langsmith==0.2.2
+loguru==0.7.3
+markdown-it-py==3.0.0
+MarkupSafe==3.0.2
+marshmallow==3.23.1
+mdurl==0.1.2
+mpmath==1.3.0
+multidict==6.1.0
+multiprocess==0.70.16
+mypy-extensions==1.0.0
+narwhals==1.17.0
+networkx==3.4.2
+numpy==1.26.4
+nvidia-cublas-cu12==12.4.5.8
+nvidia-cuda-cupti-cu12==12.4.127
+nvidia-cuda-nvrtc-cu12==12.4.127
+nvidia-cuda-runtime-cu12==12.4.127
+nvidia-cudnn-cu12==9.1.0.70
+nvidia-cufft-cu12==11.2.1.3
+nvidia-curand-cu12==10.3.5.147
+nvidia-cusolver-cu12==11.6.1.9
+nvidia-cusparse-cu12==12.3.1.170
+nvidia-nccl-cu12==2.21.5
+nvidia-nvjitlink-cu12==12.4.127
+nvidia-nvtx-cu12==12.4.127
+orjson==3.10.12
+packaging==24.2
+pandas==2.2.3
+pillow==11.0.0
+propcache==0.2.1
+protobuf==5.29.1
+psutil==6.1.0
+pyarrow==18.1.0
+pydantic==2.10.3
+pydantic-settings==2.6.1
+pydantic_core==2.27.1
+pydeck==0.9.1
+Pygments==2.18.0
+python-dateutil==2.9.0.post0
+python-dotenv==1.0.1
+pytz==2024.2
+PyYAML==6.0.2
+referencing==0.35.1
+regex==2024.11.6
+requests==2.32.3
+requests-toolbelt==1.0.0
+rich==13.9.4
+rpds-py==0.22.3
+safetensors==0.4.5
+scikit-learn==1.6.0
+scipy==1.14.1
+sentence-transformers==3.3.1
+six==1.17.0
+smmap==5.0.1
+sniffio==1.3.1
+SQLAlchemy==2.0.36
+streamlit==1.41.0
+sympy==1.13.1
+tenacity==9.0.0
+threadpoolctl==3.5.0
+tokenizers==0.21.0
+toml==0.10.2
+torch==2.5.1
+tornado==6.4.2
+tqdm==4.67.1
+transformers==4.47.0
+triton==3.1.0
+typing-inspect==0.9.0
+typing_extensions==4.12.2
+tzdata==2024.2
+urllib3==2.2.3
+watchdog==6.0.0
+xxhash==3.5.0
+yarl==1.18.3