Merge pull request #4 from loyal812/feat/chatting-class

feat: chatting class

requirements.txt

@@ -17,4 +17,6 @@ llama_index==0.9.15
 pypdf
 spacy
 ragas
-datasets
+datasets
+docx2txt
+cryptography

src/chatting/ChattingClass.py

@@ -0,0 +1,66 @@
+import openai
+import os
+import argparse
+from dotenv import load_dotenv
+from llama_index import ServiceContext, SimpleDirectoryReader, VectorStoreIndex
+from llama_index.llms import OpenAI
+
+class ChattingClass:
+    def __init__(self, model_id, data_path, api_key="", temperature=0.3):
+        self.model_id = model_id
+        self.data_path = data_path
+        self.temperature = temperature
+        self.set_api_key(api_key)
+        self.set_document(data_path)
+
+    def set_api_key(self, api_key):
+        if api_key:
+            self.api_key = api_key
+        else:
+            load_dotenv()
+            self.api_key = os.getenv("OPENAI_API_KEY")
+
+        if self.api_key is not None:
+            os.environ["OPENAI_API_KEY"] = self.api_key
+            openai.api_key = self.api_key
+            return True
+        else:
+            # Handle the absence of the environment variable
+            # You might want to log an error, raise an exception, or provide a default value
+            # For example, setting a default value
+            os.environ["OPENAI_API_KEY"] = "your_default_api_key"
+            openai.api_key = "openai_api_key"
+            return False
+
+    def set_document(self, data_path):
+        self.documents = SimpleDirectoryReader(
+            data_path
+        ).load_data()
+
+    def ask_question(self, question):
+        ft_context = ServiceContext.from_defaults(
+            llm=OpenAI(model=self.model_id, temperature=self.temperature),
+            context_window=2048
+        )
+
+        index = VectorStoreIndex.from_documents(self.documents, service_context=ft_context)
+        query_engine = index.as_query_engine(service_context=ft_context)
+
+        response = query_engine.query(question)
+        return response
+
+# Using argparse to get the question input from the user
+parser = argparse.ArgumentParser(description='Chatting Engine')
+# parser.add_argument('api_key', type=str, help='Openai API Key')
+parser.add_argument('data_path', type=str, help='Data Path')
+parser.add_argument('model_id', type=str, help='model')
+parser.add_argument('question', type=str, help='question')
+# parser.add_argument('temperature', type=str, help='temperature')
+args = parser.parse_args()
+
+# model_id = 'ft:gpt-3.5-turbo-0613:personal::8XaasBXv'
+# data_path = "./src/test/regression/regression_test003"
+chatbot = ChattingClass(model_id=args.model_id, data_path=args.data_path)
+
+response = chatbot.ask_question(args.question)
+print(response)

src/finetune/FineTuningClass.py

@@ -26,6 +26,7 @@ def __init__(self, data_path, api_key='', model='gpt-3.5-turbo', temperature=0.3
         self.retry_delay = 60
         self.set_api_key(api_key)
         self.set_document(data_path)
+        self.generate_subfolder(data_path)
 
     def set_api_key(self, api_key):
         if api_key:
@@ -52,6 +53,10 @@ def set_document(self, data_path):
             data_path
         ).load_data()
 
+    def generate_subfolder(self, data_path):
+        subfolder_name = "generated_data"
+        subfolder_path = os.path.join(data_path, subfolder_name)
+        os.makedirs(subfolder_path, exist_ok=True)
 
     def train_generation(self):
         for attempt in range(1, self.max_retries + 1):
@@ -84,8 +89,8 @@ def generate_and_save_questions(documents, output_file):
                         for question in questions:
                             f.write(question + "\n")
 
-                generate_and_save_questions(self.documents[:half_point], f'{self.data_path}/train_questions.txt')
-                generate_and_save_questions(self.documents[half_point:], f'{self.data_path}/eval_questions.txt')
+                generate_and_save_questions(self.documents[:half_point], f'{self.data_path}/generated_data/train_questions.txt')
+                generate_and_save_questions(self.documents[half_point:], f'{self.data_path}/generated_data/eval_questions.txt')
 
                 break
             except Exception as e:
@@ -140,7 +145,7 @@ def jsonl_generation(self):
         )
 
         questions = []
-        with open(f'{self.data_path}/train_questions.txt', "r") as f:
+        with open(f'{self.data_path}/generated_data/train_questions.txt', "r") as f:
             for line in f:
                 questions.append(line.strip())
 
@@ -155,12 +160,11 @@ def jsonl_generation(self):
             # Handle the exception here, you might want to log the error or take appropriate action
             print(f"An error occurred: {e}")
         finally:
-            if 'finetuning_handler' in locals() and 'data_path' in locals():
-                finetuning_handler.save_finetuning_events(f'{self.data_path}/finetuning_events.jsonl')
+            finetuning_handler.save_finetuning_events(f'{self.data_path}/generated_data/finetuning_events.jsonl')
 
 
     def finetune(self):
-        file_upload = openai.files.create(file=open(f'{self.data_path}/finetuning_events.jsonl', "rb"), purpose="fine-tune")
+        file_upload = openai.files.create(file=open(f'{self.data_path}/generated_data/finetuning_events.jsonl', "rb"), purpose="fine-tune")
         print("Uploaded file id", file_upload.id)
 
         while True:
@@ -182,7 +186,7 @@ def finetune(self):
                     print("Fine-tuned model info", job_handle)
                     print("Model id", job_handle.fine_tuned_model)
 
-                    with open(f'{self.data_path}/model.txt', "w") as f:
+                    with open(f'{self.data_path}/generated_data/model.txt', "w") as f:
                         f.write(job_handle.fine_tuned_model + "\n")
                     break
                 time.sleep(3)

src/finetune/finetune.py

@@ -18,9 +18,9 @@
     os.environ["OPENAI_API_KEY"] = "your_default_api_key"
     openai.api_key = "openai_api_key"
 
-data_path = "./src/test/regression/regression_test004"
+data_path = "./src/test/regression/regression_test003"
 
-file_upload = openai.files.create(file=open(f'{data_path}/finetuning_events.jsonl', "rb"), purpose="fine-tune")
+file_upload = openai.files.create(file=open(f'{data_path}/generated_data/finetuning_events.jsonl', "rb"), purpose="fine-tune")
 print("Uploaded file id", file_upload.id)
 
 while True:

src/finetune/initital_eval.py

@@ -15,14 +15,14 @@
     # For example, setting a default value
     os.environ["OPENAI_API_KEY"] = "your_default_api_key"
 
-data_path = "./src/test/regression/regression_test004"
+data_path = "./src/test/regression/regression_test003"
 
 documents = SimpleDirectoryReader(
     data_path
 ).load_data()
 
 questions = []
-with open(f'{data_path}/eval_questions.txt', "r") as f:
+with open(f'{data_path}/generated_data/eval_questions.txt', "r") as f:
     for line in f:
         questions.append(line.strip())
 

src/finetune/jsonl_generation.py

@@ -16,7 +16,7 @@
     # For example, setting a default value
     os.environ["OPENAI_API_KEY"] = "your_default_api_key"
 
-data_path = "./src/test/regression/regression_test004"
+data_path = "./src/test/regression/regression_test003"
 
 documents = SimpleDirectoryReader(
     data_path
@@ -33,7 +33,7 @@
 )
 
 questions = []
-with open(f'{data_path}/train_questions.txt', "r") as f:
+with open(f'{data_path}/generated_data/train_questions.txt', "r") as f:
     for line in f:
         questions.append(line.strip())
 
@@ -50,5 +50,4 @@
     # Handle the exception here, you might want to log the error or take appropriate action
     print(f"An error occurred: {e}")
 finally:
-    if 'finetuning_handler' in locals() and 'data_path' in locals():
-        finetuning_handler.save_finetuning_events(f'{data_path}/finetuning_events.jsonl')
+    finetuning_handler.save_finetuning_events(f'{data_path}/generated_data/finetuning_events.jsonl')

src/finetune/run-model.py

@@ -16,7 +16,7 @@
     os.environ["OPENAI_API_KEY"] = "your_default_api_key"
     openai.api_key = "openai_api_key"
 
-data_path = "./src/test/regression/regression_test004"
+data_path = "./src/test/regression/regression_test003"
 
 model_id = "ft:gpt-3.5-turbo-0613:personal::8XCvxg1X"
 

src/finetune/train_generation.py

@@ -15,7 +15,7 @@
 else:
     os.environ["OPENAI_API_KEY"] = "your_default_api_key"
 
-data_path = "./src/test/regression/regression_test004"
+data_path = "./src/test/regression/regression_test003"
 
 max_retries = 5
 retry_delay = 60  # in seconds
@@ -54,8 +54,8 @@ def generate_and_save_questions(documents, output_file):
                 for question in questions:
                     f.write(question + "\n")
 
-        generate_and_save_questions(documents[:half_point], f'{data_path}/train_questions.txt')
-        generate_and_save_questions(documents[half_point:], f'{data_path}/eval_questions.txt')
+        generate_and_save_questions(documents[:half_point], f'{data_path}/generated_data/train_questions.txt')
+        generate_and_save_questions(documents[half_point:], f'{data_path}/generated_data/eval_questions.txt')
 
         break
     except Exception as e:

src/test/regression/regression_test003/generated_data/eval_questions.txt

@@ -0,0 +1,11 @@
+What is the correlation between the distance in the latent space and the MPJPE in the AI Golf swing analysis tool?
+What is the title of the paper mentioned in the context?
+What is the title of the paper mentioned in the context?
+Question: What is the title of the paper referenced in citation [28]?
+Question: Who received the B.S. degree from the School of Computer Science and Engineering, Korea University of Technology and Education, Cheonan, South Korea, in 2017?
+Question: What method was used to remove the background pixels and isolate the human body in the video dataset?
+What evaluation metric was used to assess the alignment precision of the network in the golf swing analysis tool?
+What is the main challenge addressed in the proposed golf swing analysis tool using neural networks?
+What recent advancements in technology have allowed for the development of systems to help users understand the movements of professionals in sports?
+What are the main contributions of the study on the golf swing analysis tool for self-training?
+What is one of the main contributions of the proposed golf swing analysis method discussed in the study?