Add model and configuration files; implement GPT-2 tokenizer and model initialization in main.py

logs/poetry_generation.log

@@ -3,3 +3,105 @@
 2024-11-16 23:21:32,229 - main - INFO - Loading model...
 2024-11-16 23:21:32,229 - main - ERROR - Model file not found at ./models\poeticagpt-quantized-new.pth
 2024-11-16 23:21:32,231 - main - ERROR - Failed to initialize model manager
+2024-11-16 23:30:46,037 - main - INFO - Loading tokenizer...
+2024-11-16 23:30:46,798 - main - WARNING - Could not load custom vocabulary: property 'vocab' of 'GPT2TokenizerFast' object has no setter
+2024-11-16 23:30:46,799 - main - INFO - Loading model...
+2024-11-16 23:30:46,799 - main - ERROR - Error initializing model: Incorrect path_or_model_id: './models/poeticagpt.pth'. Please provide either the path to a local folder or the repo_id of a model on the Hub.
+2024-11-16 23:30:46,800 - main - ERROR - Detailed traceback:
+Traceback (most recent call last):
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\utils\hub.py", line 402, in cached_file
+    resolved_file = hf_hub_download(
+                    ^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\huggingface_hub\utils\_validators.py", line 106, in _inner_fn
+    validate_repo_id(arg_value)
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\huggingface_hub\utils\_validators.py", line 154, in validate_repo_id
+    raise HFValidationError(
+huggingface_hub.errors.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': './models/poeticagpt.pth'. Use `repo_type` argument if needed.
+
+The above exception was the direct cause of the following exception:
+
+Traceback (most recent call last):
+  File "E:\Self Work\My Projects\Poetica HuggingFace Server\poetica\main.py", line 88, in initialize
+    self.model = AutoModelForCausalLM.from_pretrained(model_path, local_files_only=True)
+                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\models\auto\auto_factory.py", line 485, in from_pretrained
+    resolved_config_file = cached_file(
+                           ^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\utils\hub.py", line 466, in cached_file
+    raise EnvironmentError(
+OSError: Incorrect path_or_model_id: './models/poeticagpt.pth'. Please provide either the path to a local folder or the repo_id of a model on the Hub.
+2024-11-16 23:30:46,803 - main - ERROR - Failed to initialize model manager
+2024-11-16 23:33:40,483 - main - INFO - Loading tokenizer...
+2024-11-16 23:33:41,621 - main - WARNING - Could not load custom vocabulary: property 'vocab' of 'GPT2TokenizerFast' object has no setter
+2024-11-16 23:33:41,622 - main - INFO - Loading model...
+2024-11-16 23:33:43,332 - main - ERROR - Error initializing model: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory ./models/.
+2024-11-16 23:33:43,333 - main - ERROR - Detailed traceback:
+Traceback (most recent call last):
+  File "E:\Self Work\My Projects\Poetica HuggingFace Server\poetica\main.py", line 88, in initialize
+    self.model = AutoModelForCausalLM.from_pretrained(model_path, local_files_only=True)
+                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\models\auto\auto_factory.py", line 564, in from_pretrained
+    return model_class.from_pretrained(
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\modeling_utils.py", line 3447, in from_pretrained
+    raise EnvironmentError(
+OSError: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory ./models/.
+2024-11-16 23:33:43,335 - main - ERROR - Failed to initialize model manager
+2024-11-16 23:34:18,283 - main - INFO - Loading tokenizer...
+2024-11-16 23:34:18,966 - main - WARNING - Could not load custom vocabulary: property 'vocab' of 'GPT2TokenizerFast' object has no setter
+2024-11-16 23:34:18,966 - main - INFO - Loading model...
+2024-11-16 23:34:20,499 - main - ERROR - Error initializing model: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory ./models/.
+2024-11-16 23:34:20,500 - main - ERROR - Detailed traceback:
+Traceback (most recent call last):
+  File "E:\Self Work\My Projects\Poetica HuggingFace Server\poetica\main.py", line 88, in initialize
+    self.model = AutoModelForCausalLM.from_pretrained(model_path, local_files_only=True)
+                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\models\auto\auto_factory.py", line 564, in from_pretrained
+    return model_class.from_pretrained(
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\modeling_utils.py", line 3447, in from_pretrained
+    raise EnvironmentError(
+OSError: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory ./models/.
+2024-11-16 23:34:20,502 - main - ERROR - Failed to initialize model manager
+2024-11-16 23:35:15,983 - main - INFO - Loading tokenizer...
+2024-11-16 23:35:17,111 - main - WARNING - Could not load custom vocabulary: property 'vocab' of 'GPT2TokenizerFast' object has no setter
+2024-11-16 23:35:17,111 - main - INFO - Loading model...
+2024-11-16 23:35:18,795 - main - ERROR - Error initializing model: Unable to load weights from pytorch checkpoint file for './models/pytorch_model.bin' at './models/pytorch_model.bin'. If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True.
+2024-11-16 23:35:18,796 - main - ERROR - Detailed traceback:
+Traceback (most recent call last):
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\modeling_utils.py", line 575, in load_state_dict
+    return torch.load(
+           ^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\torch\serialization.py", line 1024, in load
+    raise pickle.UnpicklingError(UNSAFE_MESSAGE + str(e)) from None
+_pickle.UnpicklingError: Weights only load failed. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution.Do it only if you get the file from a trusted source. WeightsUnpickler error: Unsupported class torch.qint8
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\modeling_utils.py", line 584, in load_state_dict
+    if f.read(7) == "version":
+       ^^^^^^^^^
+  File "D:\Program Files\Python\Lib\encodings\cp1252.py", line 23, in decode
+    return codecs.charmap_decode(input,self.errors,decoding_table)[0]
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+UnicodeDecodeError: 'charmap' codec can't decode byte 0x81 in position 1651: character maps to <undefined>
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "E:\Self Work\My Projects\Poetica HuggingFace Server\poetica\main.py", line 88, in initialize
+    self.model = AutoModelForCausalLM.from_pretrained(model_path, local_files_only=True)
+                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\models\auto\auto_factory.py", line 564, in from_pretrained
+    return model_class.from_pretrained(
+           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\modeling_utils.py", line 3703, in from_pretrained
+    state_dict = load_state_dict(resolved_archive_file)
+                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "e:\Self Work\My Projects\Poetica HuggingFace Server\.venv\Lib\site-packages\transformers\modeling_utils.py", line 596, in load_state_dict
+    raise OSError(
+OSError: Unable to load weights from pytorch checkpoint file for './models/pytorch_model.bin' at './models/pytorch_model.bin'. If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True.
+2024-11-16 23:35:18,815 - main - ERROR - Failed to initialize model manager
+2024-11-16 23:37:05,649 - main - INFO - Loading tokenizer...
+2024-11-16 23:37:06,372 - main - INFO - Loading model...

main.py

@@ -1,39 +1,32 @@
 import os
 from typing import Optional, Dict, Any
-from enum import Enum
 from fastapi import FastAPI, HTTPException, status
-from pathlib import Path
 import logging
 import sys
 from pydantic import BaseModel, Field
 import torch
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import GPT2Tokenizer, GPT2LMHeadModel
 import json
 
 # Define base model directory
 BASE_MODEL_DIR = "./models/"
+MODEL_PATH = os.path.join(BASE_MODEL_DIR, "poeticagpt.pth")
 
-# Configure logging with fallback to stdout if file writing fails
 def setup_logging():
     logger = logging.getLogger(__name__)
     logger.setLevel(logging.DEBUG)
     
-    # Create formatter
     formatter = logging.Formatter(
         '%(asctime)s - %(name)s - %(levelname)s - %(message)s'
     )
     
-    # Always add stdout handler
     stdout_handler = logging.StreamHandler(sys.stdout)
     stdout_handler.setFormatter(formatter)
     logger.addHandler(stdout_handler)
     
-    # Try to add file handler, but don't fail if we can't
     try:
-        # First try logs directory in current working directory
         log_dir = os.path.join(os.getcwd(), 'logs')
-        if not os.path.exists(log_dir):
-            os.makedirs(log_dir, exist_ok=True)
+        os.makedirs(log_dir, exist_ok=True)
         
         file_handler = logging.FileHandler(os.path.join(log_dir, 'poetry_generation.log'))
         file_handler.setFormatter(formatter)
@@ -43,7 +36,6 @@ def setup_logging():
     
     return logger
 
-# Set up logging
 logger = setup_logging()
 
 class GenerateRequest(BaseModel):
@@ -63,31 +55,25 @@ class ModelManager:
         """Initialize the model and tokenizer"""
         try:
             logger.info("Loading tokenizer...")
-            
-            # First, let's try to load the base GPT-2 tokenizer
-            self.tokenizer = AutoTokenizer.from_pretrained("gpt2")
-            
-            # Now customize it with your vocabulary if needed
-            vocab_path = os.path.join(BASE_MODEL_DIR, "vocab.json")
-            if os.path.exists(vocab_path):
-                try:
-                    with open(vocab_path, 'r', encoding='utf-8') as f:
-                        custom_vocab = json.load(f)
-                    self.tokenizer.vocab = custom_vocab
-                    self.tokenizer.ids_to_tokens = {v: k for k, v in custom_vocab.items()}
-                except Exception as e:
-                    logger.warning(f"Could not load custom vocabulary: {str(e)}")
+            # Load the base GPT-2 tokenizer
+            self.tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
+            self.tokenizer.pad_token = self.tokenizer.eos_token
             
             logger.info("Loading model...")
-            model_path = os.path.join(BASE_MODEL_DIR, "poeticagpt.pth")
-            if not os.path.exists(model_path):
-                logger.error(f"Model file not found at {model_path}")
+            if not os.path.exists(MODEL_PATH):
+                logger.error(f"Model file not found at {MODEL_PATH}")
                 return False
-                
-            # Load the model weights
-            self.model = AutoModelForCausalLM.from_pretrained(model_path, local_files_only=True)
             
-            # Force model to CPU
+            # Initialize a GPT2 model with default configuration
+            self.model = GPT2LMHeadModel.from_pretrained('gpt2')
+            
+            # Load your trained weights
+            state_dict = torch.load(MODEL_PATH, map_location='cpu')
+            
+            # Load the state dictionary into the model
+            self.model.load_state_dict(state_dict)
+            
+            # Force model to CPU and eval mode
             self.model.to('cpu')
             self.model.eval()
             
@@ -127,6 +113,7 @@ class ModelManager:
                     pad_token_id=self.tokenizer.eos_token_id,
                 )
             
+            # Decode the generated text
             generated_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
             
             return {
@@ -157,9 +144,6 @@ async def startup():
     """Initialize the model during startup"""
     if not model_manager.initialize():
         logger.error("Failed to initialize model manager")
-        # In production, we might want to continue running even if model fails to load
-        # Instead of exiting, we'll just log the error
-        # sys.exit(1)
 
 @app.get("/health")
 async def health_check():

models/config.json

@@ -0,0 +1,21 @@
+{
+    "architectures": [
+      "GPT2LMHeadModel"
+    ],
+    "model_type": "gpt2",
+    "activation_function": "gelu_new",
+    "attn_pdrop": 0.1,
+    "bos_token_id": 50256,
+    "embd_pdrop": 0.1,
+    "eos_token_id": 50256,
+    "initializer_range": 0.02,
+    "layer_norm_epsilon": 1e-5,
+    "n_ctx": 1024,
+    "n_embd": 768,
+    "n_head": 12,
+    "n_layer": 12,
+    "n_positions": 1024,
+    "resid_pdrop": 0.1,
+    "vocab_size": 50257,
+    "use_cache": true
+  }

models/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f77da9534fcf01b36f4780cd24ebe46e4d7f8740a1b17b66d5173d8694d6a62e
+size 139310252