Upload 3 files

llm_ans.py

@@ -0,0 +1,63 @@
+import os
+import glob
+import textwrap
+import time
+import langchain
+from langchain.document_loaders import PyPDFLoader, DirectoryLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain import PromptTemplate, LLMChain
+from langchain.vectorstores import FAISS
+from langchain.llms import HuggingFacePipeline
+from langchain.embeddings import HuggingFaceInstructEmbeddings
+from langchain.chains import RetrievalQA
+import torch
+import transformers
+from model import qa_chain
+
+def wrap_text_preserve_newlines(text, width=700):
+    # Split the input text into lines based on newline characters
+    lines = text.split('\n')
+
+    # Wrap each line individually
+    wrapped_lines = [textwrap.fill(line, width=width) for line in lines]
+
+    # Join the wrapped lines back together using newline characters
+    wrapped_text = '\n'.join(wrapped_lines)
+
+    return wrapped_text
+
+
+def process_llm_response(llm_response):
+    ans = wrap_text_preserve_newlines(llm_response['result'])
+    
+    sources_used = ' \n'.join(
+        [
+            source.metadata['source'].split('/')[-1][:-4]
+            + ' - page: '
+            + str(source.metadata['page'])
+            for source in llm_response['source_documents']
+        ]
+    )
+    
+    ans = ans + '\n\nSources: \n' + sources_used
+    return ans
+
+def llm_ans(query):
+    start = time.time()
+    
+    llm_response = qa_chain.invoke(query)
+    ans = process_llm_response(llm_response)
+    
+    end = time.time()
+
+    time_elapsed = int(round(end - start, 0))
+    time_elapsed_str = f'\n\nTime elapsed: {time_elapsed} s'
+    ans_loc=ans.find("Answer:")
+    ans_loc+=len("Answer: ")
+    return ans[ans_loc:]
+
+
+# query = "what are computer networks?"
+# result=llm_ans(query)
+# print(result)
+# print(type(result))

model.py

@@ -0,0 +1,119 @@
+import warnings
+warnings.filterwarnings("ignore")
+
+import os
+import glob
+import textwrap
+import time
+import langchain
+from langchain.document_loaders import PyPDFLoader, DirectoryLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain import PromptTemplate, LLMChain
+from langchain.vectorstores import FAISS
+from langchain.llms import HuggingFacePipeline
+from langchain.embeddings import HuggingFaceInstructEmbeddings
+from langchain.chains import RetrievalQA
+import torch
+import transformers
+from transformers import (
+    AutoTokenizer, AutoModelForCausalLM,
+    BitsAndBytesConfig,
+    pipeline
+)
+
+
+class CFG:
+    # LLMs
+    model_name = 'llama2-13b-chat' # wizardlm, llama2-7b-chat, llama2-13b-chat, mistral-7B
+    temperature = 0
+    top_p = 0.95
+    repetition_penalty = 1.15   
+    # splitting
+    split_chunk_size = 800
+    split_overlap = 0    
+    # embeddings
+    embeddings_model_repo = 'sentence-transformers/all-MiniLM-L6-v2'   
+    # similar passages
+    k = 6    
+    # paths
+    Embeddings_path =  'C:/Studies/main project/codes/final/model/cse-vectordb/faiss_index_hp'
+    # Output_folder = './cse-vectordb'
+
+model_repo = 'daryl149/llama-2-7b-chat-hf'     
+tokenizer = AutoTokenizer.from_pretrained(model_repo, use_fast=True)
+        
+bnb_config = BitsAndBytesConfig(
+            load_in_4bit = True,
+            bnb_4bit_quant_type = "nf4",
+            bnb_4bit_compute_dtype = torch.float16,
+            bnb_4bit_use_double_quant = True,
+        )
+        
+model = AutoModelForCausalLM.from_pretrained(
+            model_repo,
+            quantization_config = bnb_config,
+            device_map = 'auto',
+            low_cpu_mem_usage = True,
+            trust_remote_code = True
+        )
+        
+max_len = 2048
+
+### hugging face pipeline
+pipe = pipeline(
+    task = "text-generation",
+    model = model,
+    tokenizer = tokenizer,
+    pad_token_id = tokenizer.eos_token_id,
+#     do_sample = True,
+    max_length = max_len,
+    temperature = CFG.temperature,
+    top_p = CFG.top_p,
+    repetition_penalty = CFG.repetition_penalty
+)
+
+### langchain pipeline
+llm = HuggingFacePipeline(pipeline = pipe)
+
+
+### download embeddings model
+embeddings = HuggingFaceInstructEmbeddings(
+    model_name = CFG.embeddings_model_repo,
+    model_kwargs = {"device": "cuda"}
+)
+
+### load vector DB embeddings
+vectordb = FAISS.load_local(
+    CFG.Embeddings_path, # from input folder
+#     CFG.Output_folder + '/faiss_index_hp', # from output folder
+    embeddings,
+    allow_dangerous_deserialization=True
+)
+
+prompt_template = """
+Don't try to make up an answer, if you don't know just say that you don't know.
+Answer in the same language the question was asked.
+Use only the following pieces of context to answer the question at the end.
+
+{context}
+
+Question: {question}
+Answer:"""
+
+
+PROMPT = PromptTemplate(
+    template = prompt_template, 
+    input_variables = ["context", "question"]
+)
+
+retriever = vectordb.as_retriever(search_kwargs = {"k": CFG.k, "search_type" : "similarity"})
+
+qa_chain = RetrievalQA.from_chain_type(
+    llm = llm,
+    chain_type = "stuff", # map_reduce, map_rerank, stuff, refine
+    retriever = retriever, 
+    chain_type_kwargs = {"prompt": PROMPT},
+    return_source_documents = True,
+    verbose = False
+)
+print("Hello")

quiz_gen.py

@@ -0,0 +1,83 @@
+import os
+from IPython.display import Markdown, display
+from langchain import PromptTemplate,HuggingFaceHub
+from langchain.chat_models import ChatOpenAI
+from langchain.chains import LLMChain
+
+import re
+import json
+from langchain.chat_models import ChatOpenAI
+from langchain.schema import HumanMessage
+from langchain.prompts import PromptTemplate, ChatPromptTemplate, HumanMessagePromptTemplate
+from langchain.output_parsers import StructuredOutputParser, ResponseSchema
+
+import warnings
+warnings.filterwarnings("ignore")
+import os
+import textwrap
+import langchain
+from langchain.llms import HuggingFacePipeline
+import torch
+import transformers
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import LlamaTokenizer, LlamaForCausalLM, pipeline
+from langchain.vectorstores import Chroma, FAISS
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.chains import RetrievalQA, VectorDBQA
+from langchain.document_loaders import PyPDFLoader
+from langchain.document_loaders import DirectoryLoader
+from InstructorEmbedding import INSTRUCTOR
+from langchain.embeddings import HuggingFaceInstructEmbeddings
+from langchain.llms import CTransformers
+# import random
+import random
+from langchain.chains.question_answering import load_qa_chain
+from model import llm,vectordb as index,embeddings
+ 
+
+# mistral = CTransformers(
+#     model = "mistral-7b-instruct-v0.2.Q4_K_S.gguf",
+#     model_type="mistral",
+#     max_new_tokens = 4096,
+#     temperature = 0,
+#     repetition_penalty= 1.1,
+#     device="cuda" if torch.cuda.is_available() else "cpu")
+
+# llm=HuggingFaceHub(repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1", 
+#                                         model_kwargs={"temperature":0.05, 
+#                                                       "max_length":1024,"top_p":0.95,"repetition_penalty":1.15,"torch_dtype":"torch.float16", "device_map":"auto"})
+
+# topic=["Artificial intelligence", "Algorithm analysis","Computer graphics and image processing", "Computer organization and architecture","Compiler Design",
+#     "Computer networks", "Data Structure", "Database management system", "Distributed computing", "internet of things", "mobile computing", "management of software system",
+#     "Java", "Operating system", "Python programming", "Soft Computing", "Web programming"]
+
+# select=[i for i in range(len(topic)-1)]
+
+response_schemas = [
+    ResponseSchema(name="question", description="Question generated from provided input text data."),
+    ResponseSchema(name="choices", description="Available options for a multiple-choice question in comma separated."),
+    ResponseSchema(name="answer", description="Correct answer for the asked question.")
+]
+output_parser = StructuredOutputParser.from_response_schemas(response_schemas)
+format_instructions = output_parser.get_format_instructions()
+prompt = ChatPromptTemplate(
+    messages=[
+        HumanMessagePromptTemplate.from_template("""Please generate {num_questions} multiple choice questions 
+        from {user_prompt}. 
+        \n{format_instructions}\n{user_prompt}""")  
+    ],
+    input_variables=["user_prompt"],
+    partial_variables={"format_instructions": format_instructions}
+)
+final_query = prompt.format_prompt(user_prompt = "computer networks",num_questions=5)
+
+
+chain = LLMChain(llm=llm,
+                 prompt=prompt)
+
+# sub=topic[random.choice(select)]
+# # chain = LLMChain(prompt=prompt, 
+# #                      llm=llm)
+
+quiz_response = chain.run(user_prompt = "computer networks",num_questions=5)
+print(quiz_response)