gptmed 0.5.2__py3-none-any.whl → 0.5.5__py3-none-any.whl

gptmed/framework/cli/__init__.py

@@ -3,6 +3,22 @@ from .startproject import startproject
 
 def main():
     if len(sys.argv) < 3 or sys.argv[1] != "startproject":
-        print("Usage: gptmed startproject <projectname>")
+        print("Usage: gptmed startproject <projectname> [--qna|--conversational]")
         sys.exit(1)
-    startproject(sys.argv[2])
+    
+    project_name = sys.argv[2]
+    project_type = None
+    
+    # Check for flags
+    if len(sys.argv) > 3:
+        flag = sys.argv[3]
+        if flag == "--qna":
+            project_type = "qna"
+        elif flag == "--conversational":
+            project_type = "conversational"
+        else:
+            print(f"Invalid flag: {flag}")
+            print("Usage: gptmed startproject <projectname> [--qna|--conversational]")
+            sys.exit(1)
+    
+    startproject(project_name, project_type)

gptmed/framework/cli/__main__.py

@@ -0,0 +1,7 @@
+"""
+CLI entry point for gptmed framework
+"""
+from . import main
+
+if __name__ == "__main__":
+    main()

gptmed/framework/cli/startproject.py

@@ -1,17 +1,855 @@
 import os
 import sys
 
-def startproject(project_name):
-    if not project_name.isidentifier():
-        print("Invalid project name.")
-        sys.exit(1)
-    if os.path.exists(project_name):
-        print(f"Directory '{project_name}' already exists.")
+def create_qna_templates(project_name):
+    """Create boilerplate for QNA model training architecture"""
+    
+    # Create directory structure
+    os.makedirs(os.path.join(project_name, "configs"))
+    os.makedirs(os.path.join(project_name, "data", "raw"))
+    os.makedirs(os.path.join(project_name, "data", "processed"))
+    os.makedirs(os.path.join(project_name, "models", "checkpoints"))
+    os.makedirs(os.path.join(project_name, "tokenizer"))
+    os.makedirs(os.path.join(project_name, "logs"))
+    os.makedirs(os.path.join(project_name, "inference"))
+    
+    # Create main.py
+    with open(os.path.join(project_name, "main.py"), "w") as f:
+        f.write("""\"\"\"
+Main entry point for QNA Model Training
+\"\"\"
+import gptmed
+from pathlib import Path
+
+def main():
+    # Step 1: Create configuration
+    config_path = 'configs/training_config.yaml'
+    if not Path(config_path).exists():
+        gptmed.create_config(config_path)
+        print(f"Configuration file created at {config_path}")
+        print("Please edit the configuration file and run again.")
+        return
+    
+    # Step 2: Train the model
+    print("Starting QNA model training...")
+    results = gptmed.train_from_config(config_path, device='auto')
+    
+    print(f"\\nTraining completed!")
+    print(f"Best checkpoint: {results['best_checkpoint']}")
+    print(f"Final validation loss: {results['final_val_loss']}")
+
+if __name__ == "__main__":
+    main()
+""")
+    
+    # Create preprocess.py
+    with open(os.path.join(project_name, "preprocess.py"), "w") as f:
+        f.write("""\"\"\"
+Data preprocessing for QNA dataset
+\"\"\"
+import json
+from pathlib import Path
+
+def preprocess_qna_data(input_file, output_file):
+    \"\"\"
+    Preprocess QNA data from raw format to training format.
+    
+    Expected input format (JSONL):
+    {"question": "What is X?", "answer": "X is..."}
+    
+    Output format:
+    {"text": "Q: What is X?\\nA: X is..."}
+    \"\"\"
+    processed_data = []
+    
+    with open(input_file, 'r', encoding='utf-8') as f:
+        for line in f:
+            item = json.loads(line)
+            question = item.get('question', '').strip()
+            answer = item.get('answer', '').strip()
+            
+            if question and answer:
+                formatted_text = f"Q: {question}\\nA: {answer}"
+                processed_data.append({"text": formatted_text})
+    
+    # Save processed data
+    output_path = Path(output_file)
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    
+    with open(output_file, 'w', encoding='utf-8') as f:
+        for item in processed_data:
+            f.write(json.dumps(item) + '\\n')
+    
+    print(f"Processed {len(processed_data)} QNA pairs")
+    print(f"Saved to {output_file}")
+
+if __name__ == "__main__":
+    # Example usage
+    preprocess_qna_data(
+        'data/raw/qna_data.jsonl',
+        'data/processed/train.jsonl'
+    )
+""")
+    
+    # Create inference script
+    with open(os.path.join(project_name, "inference", "generate_answer.py"), "w") as f:
+        f.write("""\"\"\"
+Generate answers using trained QNA model
+\"\"\"
+import gptmed
+from pathlib import Path
+
+def generate_answer(question, checkpoint_path, tokenizer_path, max_length=200):
+    \"\"\"Generate answer for a given question\"\"\"
+    
+    # Format question
+    prompt = f"Q: {question}\\nA:"
+    
+    # Generate answer
+    answer = gptmed.generate(
+        checkpoint=checkpoint_path,
+        prompt=prompt,
+        tokenizer=tokenizer_path,
+        max_length=max_length,
+        temperature=0.7,
+        top_k=50
+    )
+    
+    # Extract answer (remove the prompt)
+    answer = answer.replace(prompt, '').strip()
+    
+    return answer
+
+if __name__ == "__main__":
+    # Example usage
+    checkpoint = "models/checkpoints/best_model.pt"
+    tokenizer = "tokenizer/qna_tokenizer.model"
+    
+    question = "What is machine learning?"
+    answer = generate_answer(question, checkpoint, tokenizer)
+    
+    print(f"Question: {question}")
+    print(f"Answer: {answer}")
+""")
+    
+    # Create README.md
+    with open(os.path.join(project_name, "README.md"), "w") as f:
+        f.write(f"""# {project_name} - QNA Model
+
+Question and Answer generation model training architecture.
+
+## Directory Structure
+
+```
+{project_name}/
+├── configs/              # Training configurations
+├── data/
+│   ├── raw/             # Raw QNA data
+│   └── processed/       # Preprocessed data
+├── models/
+│   └── checkpoints/     # Model checkpoints
+├── tokenizer/           # Tokenizer files
+├── logs/                # Training logs
+├── inference/           # Inference scripts
+├── main.py              # Main training script
+├── preprocess.py        # Data preprocessing
+└── README.md
+```
+
+## Getting Started
+
+### 1. Prepare Data
+
+Place your QNA data in JSONL format in `data/raw/qna_data.jsonl`:
+
+```json
+{{"question": "What is X?", "answer": "X is..."}}
+{{"question": "How does Y work?", "answer": "Y works by..."}}
+```
+
+### 2. Preprocess Data
+
+```bash
+python preprocess.py
+```
+
+### 3. Configure Training
+
+```bash
+python main.py  # This will create a default config file
+```
+
+Edit `configs/training_config.yaml` with your settings.
+
+### 4. Train Model
+
+```bash
+python main.py
+```
+
+### 5. Generate Answers
+
+```bash
+python inference/generate_answer.py
+```
+
+## Training Configuration
+
+Edit `configs/training_config.yaml` to customize:
+- Model size (tiny, small, medium)
+- Training parameters (epochs, batch size, learning rate)
+- Data paths
+- Device selection (CPU/GPU)
+
+## Inference
+
+Use the trained model to generate answers:
+
+```python
+import gptmed
+
+answer = gptmed.generate(
+    checkpoint='models/checkpoints/best_model.pt',
+    prompt='Q: Your question?\\nA:',
+    tokenizer='tokenizer/qna_tokenizer.model'
+)
+```
+""")
+
+def create_conversational_templates(project_name):
+    """Create boilerplate for conversational model training architecture"""
+    
+    # Create directory structure
+    os.makedirs(os.path.join(project_name, "configs"))
+    os.makedirs(os.path.join(project_name, "data", "raw"))
+    os.makedirs(os.path.join(project_name, "data", "processed"))
+    os.makedirs(os.path.join(project_name, "models", "checkpoints"))
+    os.makedirs(os.path.join(project_name, "tokenizer"))
+    os.makedirs(os.path.join(project_name, "logs"))
+    os.makedirs(os.path.join(project_name, "inference"))
+    os.makedirs(os.path.join(project_name, "utils"))
+    
+    # Create main.py
+    with open(os.path.join(project_name, "main.py"), "w") as f:
+        f.write("""\"\"\"
+Main entry point for Conversational Model Training
+\"\"\"
+import gptmed
+from pathlib import Path
+
+def main():
+    # Step 1: Create configuration
+    config_path = 'configs/training_config.yaml'
+    if not Path(config_path).exists():
+        gptmed.create_config(config_path)
+        print(f"Configuration file created at {config_path}")
+        print("Please edit the configuration file and run again.")
+        return
+    
+    # Step 2: Train the model
+    print("Starting conversational model training...")
+    results = gptmed.train_from_config(config_path, device='auto')
+    
+    print(f"\\nTraining completed!")
+    print(f"Best checkpoint: {results['best_checkpoint']}")
+    print(f"Final validation loss: {results['final_val_loss']}")
+
+if __name__ == "__main__":
+    main()
+""")
+    
+    # Create preprocess.py for conversational data
+    with open(os.path.join(project_name, "preprocess.py"), "w") as f:
+        f.write("""\"\"\"
+Data preprocessing for conversational dataset
+\"\"\"
+import json
+from pathlib import Path
+from typing import List, Dict
+
+def format_conversation(messages: List[Dict[str, str]], 
+                       user_token="<|user|>", 
+                       assistant_token="<|assistant|>",
+                       end_token="<|endoftext|>") -> str:
+    \"\"\"
+    Format a conversation into training text.
+    
+    Args:
+        messages: List of message dicts with 'role' and 'content'
+        user_token: Token to mark user messages
+        assistant_token: Token to mark assistant messages
+        end_token: Token to mark end of conversation
+    
+    Returns:
+        Formatted conversation string
+    \"\"\"
+    conversation = []
+    
+    for msg in messages:
+        role = msg.get('role', '')
+        content = msg.get('content', '').strip()
+        
+        if role == 'user':
+            conversation.append(f"{user_token} {content}")
+        elif role == 'assistant':
+            conversation.append(f"{assistant_token} {content}")
+    
+    return "\\n".join(conversation) + f"\\n{end_token}"
+
+def preprocess_conversational_data(input_file, output_file):
+    \"\"\"
+    Preprocess conversational data from raw format to training format.
+    
+    Expected input format (JSONL):
+    {
+        "conversation": [
+            {"role": "user", "content": "Hello!"},
+            {"role": "assistant", "content": "Hi! How can I help?"},
+            {"role": "user", "content": "Tell me about AI"},
+            {"role": "assistant", "content": "AI is..."}
+        ]
+    }
+    
+    Output format:
+    {"text": "<|user|> Hello!\\n<|assistant|> Hi! How can I help?\\n..."}
+    \"\"\"
+    processed_data = []
+    
+    with open(input_file, 'r', encoding='utf-8') as f:
+        for line in f:
+            item = json.loads(line)
+            conversation = item.get('conversation', [])
+            
+            if len(conversation) >= 2:  # At least one exchange
+                formatted_text = format_conversation(conversation)
+                processed_data.append({"text": formatted_text})
+    
+    # Save processed data
+    output_path = Path(output_file)
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    
+    with open(output_file, 'w', encoding='utf-8') as f:
+        for item in processed_data:
+            f.write(json.dumps(item) + '\\n')
+    
+    print(f"Processed {len(processed_data)} conversations")
+    print(f"Saved to {output_file}")
+
+if __name__ == "__main__":
+    # Example usage
+    preprocess_conversational_data(
+        'data/raw/conversations.jsonl',
+        'data/processed/train.jsonl'
+    )
+""")
+    
+    # Create conversation handler utility
+    with open(os.path.join(project_name, "utils", "conversation_handler.py"), "w") as f:
+        f.write("""\"\"\"
+Conversation management utilities
+\"\"\"
+from typing import List, Dict, Optional
+
+class ConversationHistory:
+    \"\"\"Manage conversation history for multi-turn dialogues\"\"\"
+    
+    def __init__(self, max_history: int = 10):
+        \"\"\"
+        Initialize conversation history.
+        
+        Args:
+            max_history: Maximum number of turns to keep in history
+        \"\"\"
+        self.messages: List[Dict[str, str]] = []
+        self.max_history = max_history
+    
+    def add_user_message(self, content: str):
+        \"\"\"Add a user message to history\"\"\"
+        self.messages.append({"role": "user", "content": content})
+        self._trim_history()
+    
+    def add_assistant_message(self, content: str):
+        \"\"\"Add an assistant message to history\"\"\"
+        self.messages.append({"role": "assistant", "content": content})
+        self._trim_history()
+    
+    def _trim_history(self):
+        \"\"\"Keep only the most recent messages\"\"\"
+        if len(self.messages) > self.max_history * 2:  # 2 messages per turn
+            self.messages = self.messages[-(self.max_history * 2):]
+    
+    def get_prompt(self, 
+                   user_token: str = "<|user|>",
+                   assistant_token: str = "<|assistant|>") -> str:
+        \"\"\"
+        Generate prompt from conversation history.
+        
+        Returns:
+            Formatted conversation prompt
+        \"\"\"
+        prompt_parts = []
+        
+        for msg in self.messages:
+            role = msg['role']
+            content = msg['content']
+            
+            if role == 'user':
+                prompt_parts.append(f"{user_token} {content}")
+            elif role == 'assistant':
+                prompt_parts.append(f"{assistant_token} {content}")
+        
+        # Add assistant token for next response
+        prompt_parts.append(assistant_token)
+        
+        return "\\n".join(prompt_parts) + " "
+    
+    def clear(self):
+        \"\"\"Clear conversation history\"\"\"
+        self.messages = []
+    
+    def get_last_n_turns(self, n: int) -> List[Dict[str, str]]:
+        \"\"\"Get last n conversation turns\"\"\"
+        return self.messages[-(n * 2):]
+""")
+    
+    # Create interactive chat script
+    with open(os.path.join(project_name, "inference", "interactive_chat.py"), "w") as f:
+        f.write("""\"\"\"
+Interactive chat with conversational model
+\"\"\"
+import sys
+sys.path.append('..')
+
+import gptmed
+from pathlib import Path
+from utils.conversation_handler import ConversationHistory
+
+class ChatBot:
+    \"\"\"Interactive chatbot using trained conversational model\"\"\"
+    
+    def __init__(self, checkpoint_path: str, tokenizer_path: str):
+        self.checkpoint_path = checkpoint_path
+        self.tokenizer_path = tokenizer_path
+        self.history = ConversationHistory(max_history=5)
+        self.user_token = "<|user|>"
+        self.assistant_token = "<|assistant|>"
+        self.end_token = "<|endoftext|>"
+    
+    def generate_response(self, user_input: str, max_length: int = 150) -> str:
+        \"\"\"Generate response to user input\"\"\"
+        
+        # Add user message to history
+        self.history.add_user_message(user_input)
+        
+        # Get prompt from conversation history
+        prompt = self.history.get_prompt(self.user_token, self.assistant_token)
+        
+        # Generate response
+        response = gptmed.generate(
+            checkpoint=self.checkpoint_path,
+            prompt=prompt,
+            tokenizer=self.tokenizer_path,
+            max_length=max_length,
+            temperature=0.8,
+            top_k=50,
+            top_p=0.9
+        )
+        
+        # Extract assistant response
+        response = response.replace(prompt, '').strip()
+        
+        # Remove end token if present
+        if self.end_token in response:
+            response = response.split(self.end_token)[0].strip()
+        
+        # Remove user token if model generated it (shouldn't happen but just in case)
+        if self.user_token in response:
+            response = response.split(self.user_token)[0].strip()
+        
+        # Add assistant response to history
+        self.history.add_assistant_message(response)
+        
+        return response
+    
+    def chat(self):
+        \"\"\"Start interactive chat session\"\"\"
+        print("Conversational AI Chatbot")
+        print("Type 'quit' or 'exit' to end the conversation")
+        print("Type 'clear' to reset conversation history")
+        print("-" * 50)
+        
+        while True:
+            try:
+                user_input = input("You: ").strip()
+                
+                if not user_input:
+                    continue
+                
+                if user_input.lower() in ['quit', 'exit']:
+                    print("Goodbye!")
+                    break
+                
+                if user_input.lower() == 'clear':
+                    self.history.clear()
+                    print("Conversation history cleared.")
+                    continue
+                
+                response = self.generate_response(user_input)
+                print(f"Bot: {response}")
+                
+            except KeyboardInterrupt:
+                print("\\nGoodbye!")
+                break
+            except Exception as e:
+                print(f"Error: {e}")
+
+if __name__ == "__main__":
+    # Initialize chatbot
+    checkpoint = "../models/checkpoints/best_model.pt"
+    tokenizer = "../tokenizer/conv_tokenizer.model"
+    
+    if not Path(checkpoint).exists():
+        print(f"Error: Checkpoint not found at {checkpoint}")
+        print("Please train your model first using main.py")
         sys.exit(1)
+    
+    chatbot = ChatBot(checkpoint, tokenizer)
+    chatbot.chat()
+""")
+    
+    # Create batch inference script
+    with open(os.path.join(project_name, "inference", "batch_inference.py"), "w") as f:
+        f.write("""\"\"\"
+Batch inference for conversational model
+\"\"\"
+import sys
+sys.path.append('..')
+
+import json
+import gptmed
+from pathlib import Path
+from utils.conversation_handler import ConversationHistory
+from typing import List, Dict
+
+def generate_conversation_response(
+    conversation_history: List[Dict[str, str]],
+    checkpoint_path: str,
+    tokenizer_path: str,
+    max_length: int = 150
+) -> str:
+    \"\"\"
+    Generate response for a conversation.
+    
+    Args:
+        conversation_history: List of messages with role and content
+        checkpoint_path: Path to model checkpoint
+        tokenizer_path: Path to tokenizer
+        max_length: Maximum response length
+    
+    Returns:
+        Generated response
+    \"\"\"
+    history = ConversationHistory()
+    
+    # Rebuild conversation history
+    for msg in conversation_history:
+        if msg['role'] == 'user':
+            history.add_user_message(msg['content'])
+        elif msg['role'] == 'assistant':
+            history.add_assistant_message(msg['content'])
+    
+    # Get prompt
+    prompt = history.get_prompt()
+    
+    # Generate response
+    response = gptmed.generate(
+        checkpoint=checkpoint_path,
+        prompt=prompt,
+        tokenizer=tokenizer_path,
+        max_length=max_length,
+        temperature=0.8,
+        top_k=50
+    )
+    
+    # Clean up response
+    response = response.replace(prompt, '').strip()
+    if "<|endoftext|>" in response:
+        response = response.split("<|endoftext|>")[0].strip()
+    
+    return response
+
+def batch_process_conversations(
+    input_file: str,
+    output_file: str,
+    checkpoint_path: str,
+    tokenizer_path: str
+):
+    \"\"\"
+    Process multiple conversations in batch.
+    
+    Input format (JSONL):
+    {"conversation": [{"role": "user", "content": "Hi"}, ...]}
+    
+    Output format (JSONL):
+    {"conversation": [...], "generated_response": "..."}
+    \"\"\"
+    results = []
+    
+    with open(input_file, 'r') as f:
+        for line in f:
+            item = json.loads(line)
+            conversation = item.get('conversation', [])
+            
+            if conversation:
+                response = generate_conversation_response(
+                    conversation,
+                    checkpoint_path,
+                    tokenizer_path
+                )
+                
+                results.append({
+                    "conversation": conversation,
+                    "generated_response": response
+                })
+    
+    # Save results
+    with open(output_file, 'w') as f:
+        for result in results:
+            f.write(json.dumps(result) + '\\n')
+    
+    print(f"Processed {len(results)} conversations")
+    print(f"Results saved to {output_file}")
+
+if __name__ == "__main__":
+    batch_process_conversations(
+        input_file='test_conversations.jsonl',
+        output_file='results.jsonl',
+        checkpoint_path='../models/checkpoints/best_model.pt',
+        tokenizer_path='../tokenizer/conv_tokenizer.model'
+    )
+""")
+    
+    # Create README.md
+    with open(os.path.join(project_name, "README.md"), "w") as f:
+        f.write(f"""# {project_name} - Conversational Model
+
+Multi-turn conversational language model training architecture.
+
+## Directory Structure
+
+```
+{project_name}/
+├── configs/              # Training configurations
+├── data/
+│   ├── raw/             # Raw conversation data
+│   └── processed/       # Preprocessed data
+├── models/
+│   └── checkpoints/     # Model checkpoints
+├── tokenizer/           # Tokenizer files
+├── logs/                # Training logs
+├── inference/           # Inference scripts
+│   ├── interactive_chat.py   # Interactive chat interface
+│   └── batch_inference.py    # Batch processing
+├── utils/               # Utility modules
+│   └── conversation_handler.py
+├── main.py              # Main training script
+├── preprocess.py        # Data preprocessing
+└── README.md
+```
+
+## Getting Started
+
+### 1. Prepare Data
+
+Place your conversational data in JSONL format in `data/raw/conversations.jsonl`:
+
+```json
+{{
+  "conversation": [
+    {{"role": "user", "content": "Hello!"}},
+    {{"role": "assistant", "content": "Hi! How can I help you today?"}},
+    {{"role": "user", "content": "Tell me about AI"}},
+    {{"role": "assistant", "content": "AI stands for Artificial Intelligence..."}}
+  ]
+}}
+```
+
+### 2. Preprocess Data
+
+```bash
+python preprocess.py
+```
+
+This will format conversations with special tokens:
+- `<|user|>` - marks user messages
+- `<|assistant|>` - marks assistant messages
+- `<|endoftext|>` - marks end of conversation
+
+### 3. Configure Training
+
+```bash
+python main.py  # Creates default config
+```
+
+Edit `configs/training_config.yaml` with your settings.
+
+### 4. Train Model
+
+```bash
+python main.py
+```
+
+### 5. Interactive Chat
+
+```bash
+cd inference
+python interactive_chat.py
+```
+
+Commands:
+- Type your message and press Enter
+- Type `clear` to reset conversation history
+- Type `quit` or `exit` to end chat
+
+### 6. Batch Inference
+
+```bash
+cd inference
+python batch_inference.py
+```
+
+## Features
+
+### Multi-turn Conversations
+The model maintains conversation history and generates contextually relevant responses.
+
+### Conversation Management
+The `ConversationHistory` class manages:
+- Message history tracking
+- Automatic history trimming
+- Prompt generation from history
+
+### Interactive Chat
+Real-time chat interface with:
+- Multi-turn conversation support
+- History management
+- User-friendly commands
+
+### Batch Processing
+Process multiple conversations:
+- Evaluate model on test sets
+- Generate responses for datasets
+- Performance benchmarking
+
+## Conversation Format
+
+The model uses special tokens to structure conversations:
+
+```
+<|user|> Hello, how are you?
+<|assistant|> I'm doing well, thank you! How can I assist you today?
+<|user|> I need help with Python
+<|assistant|> I'd be happy to help with Python! What specific topic?
+<|endoftext|>
+```
+
+## Training Tips
+
+1. **Data Quality**: Ensure conversations are natural and coherent
+2. **History Length**: Adjust `max_history` based on your use case
+3. **Temperature**: Lower (0.6-0.8) for focused responses, higher (0.8-1.0) for creative
+4. **Model Size**: Start with tiny/small, scale up as needed
+
+## Inference Parameters
+
+Adjust generation parameters in inference scripts:
+
+```python
+response = gptmed.generate(
+    checkpoint=checkpoint_path,
+    prompt=prompt,
+    tokenizer=tokenizer_path,
+    max_length=150,      # Maximum response length
+    temperature=0.8,     # Randomness (0.0-1.0)
+    top_k=50,            # Top-k sampling
+    top_p=0.9            # Nucleus sampling
+)
+```
+
+## Example Usage
+
+```python
+from utils.conversation_handler import ConversationHistory
+import gptmed
+
+# Initialize conversation
+history = ConversationHistory(max_history=5)
+
+# Add messages
+history.add_user_message("What is machine learning?")
+
+# Generate prompt
+prompt = history.get_prompt()
+
+# Generate response
+response = gptmed.generate(
+    checkpoint='models/checkpoints/best_model.pt',
+    prompt=prompt,
+    tokenizer='tokenizer/conv_tokenizer.model'
+)
+
+# Add response to history
+history.add_assistant_message(response)
+```
+""")
+
+def create_basic_project(project_name):
+    """Create basic project structure (original behavior)"""
     os.makedirs(os.path.join(project_name, "configs"))
     os.makedirs(os.path.join(project_name, "tasks"))
     os.makedirs(os.path.join(project_name, "models"))
     os.makedirs(os.path.join(project_name, "data"))
     with open(os.path.join(project_name, "main.py"), "w") as f:
         f.write("import gptmed\n\n# Your project entrypoint\n")
-    print(f"Project '{project_name}' created.")
+
+def startproject(project_name, project_type=None):
+    """
+    Create a new gptmed project.
+    
+    Args:
+        project_name: Name of the project
+        project_type: Type of project ('qna', 'conversational', or None for basic)
+    """
+    if not project_name.isidentifier():
+        print("Invalid project name. Your project name must be a valid Python identifier. "
+              "Do not use hyphens or spaces. Use underscores instead.")
+        sys.exit(1)
+    
+    if os.path.exists(project_name):
+        print(f"Directory '{project_name}' already exists.")
+        sys.exit(1)
+    
+    # Create project based on type
+    if project_type == "qna":
+        create_qna_templates(project_name)
+        print(f"QNA project '{project_name}' created successfully!")
+        print(f"\nNext steps:")
+        print(f"1. cd {project_name}")
+        print(f"2. Place your QNA data in data/raw/qna_data.jsonl")
+        print(f"3. Run: python preprocess.py")
+        print(f"4. Run: python main.py")
+        
+    elif project_type == "conversational":
+        create_conversational_templates(project_name)
+        print(f"Conversational project '{project_name}' created successfully!")
+        print(f"\nNext steps:")
+        print(f"1. cd {project_name}")
+        print(f"2. Place your conversation data in data/raw/conversations.jsonl")
+        print(f"3. Run: python preprocess.py")
+        print(f"4. Run: python main.py")
+        print(f"5. For interactive chat: cd inference && python interactive_chat.py")
+        
+    else:
+        create_basic_project(project_name)
+        print(f"Project '{project_name}' created.")
+

{gptmed-0.5.2.dist-info → gptmed-0.5.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gptmed
-Version: 0.5.2
+Version: 0.5.5
 Summary: A lightweight GPT-based language model framework for training custom question-answering models on any domain
 Author-email: Sanjog Sigdel <sigdelsanjog@gmail.com>, Sanjog Sigdel <sanjog.sigdel@ku.edu.np>
 License-Expression: MIT
@@ -438,7 +438,7 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
 
 ## Changelog
 
-See [CHANGELOG.md](CHANGELOG.md) for version history.
+[Full Changelog](https://github.com/sigdelsanjog/gptmed/blob/main/CHANGELOG.md)
 
 ---
 

{gptmed-0.5.2.dist-info → gptmed-0.5.5.dist-info}/RECORD

@@ -10,8 +10,9 @@ gptmed/data/parsers/__init__.py,sha256=BgVzXuZgeE5DUCC4SzN7vflL40wQ4Q4_4DmJ1Y43_
 gptmed/data/parsers/medquad_parser.py,sha256=g3QCRiVBdcq8RdyuYH_qKFrHgU5KkHY59WfWxUwspP0,7974
 gptmed/data/parsers/text_formatter.py,sha256=tVmnDBT54BbxX9BPKMXSPzzLmM39frDxKRKuz_HoRag,4072
 gptmed/framework/__init__.py,sha256=TlzM7NS_n0KQnm9PQTJRrb5pEb6rBXC1pqGPhbSO_bQ,25
-gptmed/framework/cli/__init__.py,sha256=Cc_rehr6bPwBr_4vlGoOaci9W9fxQLvZOE6uNYukc7Y,230
-gptmed/framework/cli/startproject.py,sha256=n74vivw33rUE7aB1XC5n0999DGR3ifNKjHz6iB9xG40,661
+gptmed/framework/cli/__init__.py,sha256=oBUmoaWLCjFs3_aod-hcMCcC11UP4t4SohDnZ7Sdmx0,729
+gptmed/framework/cli/__main__.py,sha256=rLBZjEi695ZgOW8pypqpg2kLgtcDhrI_9_QcrUO3WkU,103
+gptmed/framework/cli/startproject.py,sha256=l73Isqtp5MTMxPBfsy4sIpVp2ClA9p9_cRG7Lg7QERY,25489
 gptmed/inference/__init__.py,sha256=NDPViXhOgpItC8n13T9axX4UH1E7mrjt6kJ5OfIwvMs,25
 gptmed/inference/decoding_utils.py,sha256=zTDZYdl2jcGwSrcINXMw-5uoYuF4A9TSushhPxJi1o0,5041
 gptmed/inference/generation_config.py,sha256=hpPyZUk1K6qGSBAoQx3Jm0_ZrrYld77ACxbIlCCCcVU,2813
@@ -45,9 +46,9 @@ gptmed/training/utils.py,sha256=pJxCwneNr2STITIYwIDCxRzIICDFOxOMzK8DT7ck2oQ,5651
 gptmed/utils/__init__.py,sha256=XuMhIqOXF7mjnog_6Iky-hSbwvFb0iK42B4iDUpgi0U,44
 gptmed/utils/checkpoints.py,sha256=jPKJtO0YRZieGmpwqotgDkBzd__s_raDxS1kLpfjBJE,7113
 gptmed/utils/logging.py,sha256=7dJc1tayMxCBjFSDXe4r9ACUTpoPTTGsJ0UZMTqZIDY,5303
-gptmed-0.5.2.dist-info/licenses/LICENSE,sha256=v2spsd7N1pKFFh2G8wGP_45iwe5S0DYiJzG4im8Rupc,1066
-gptmed-0.5.2.dist-info/METADATA,sha256=wjn14NVIVMstJr6tanFiD3z0kveJzMODspGhzyesV9M,13816
-gptmed-0.5.2.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-gptmed-0.5.2.dist-info/entry_points.txt,sha256=ZZeYg2kOQuHHvRvQYRvq5L-RpClnBMHSpUom9DxQW0c,145
-gptmed-0.5.2.dist-info/top_level.txt,sha256=mhyEq3rG33t21ziJz5w3TPgx0RjPf4zXMNUx2JTiNmE,7
-gptmed-0.5.2.dist-info/RECORD,,
+gptmed-0.5.5.dist-info/licenses/LICENSE,sha256=v2spsd7N1pKFFh2G8wGP_45iwe5S0DYiJzG4im8Rupc,1066
+gptmed-0.5.5.dist-info/METADATA,sha256=-7Drfeaxy2SWQ3nlnHqwP4pj1q9bsrmTxe99pIWMePk,13842
+gptmed-0.5.5.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+gptmed-0.5.5.dist-info/entry_points.txt,sha256=ZZeYg2kOQuHHvRvQYRvq5L-RpClnBMHSpUom9DxQW0c,145
+gptmed-0.5.5.dist-info/top_level.txt,sha256=mhyEq3rG33t21ziJz5w3TPgx0RjPf4zXMNUx2JTiNmE,7
+gptmed-0.5.5.dist-info/RECORD,,