csc525_retrieval_based_chatbot / chatbot_model.py
JoeArmani
chat refinements
c7c1b4e
import os
import numpy as np
from sentence_transformers import SentenceTransformer
import tensorflow as tf
from typing import List, Tuple, Dict, Optional, Union, Any
import math
from dataclasses import dataclass
import json
from pathlib import Path
import datetime
import faiss
import gc
import re
from response_quality_checker import ResponseQualityChecker
from cross_encoder_reranker import CrossEncoderReranker
from conversation_summarizer import DeviceAwareModel, Summarizer
from chatbot_config import ChatbotConfig
from tf_data_pipeline import TFDataPipeline
import absl.logging
from logger_config import config_logger
from tqdm.auto import tqdm
absl.logging.set_verbosity(absl.logging.WARNING)
logger = config_logger(__name__)
logger.setLevel("WARNING")
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
tqdm(disable=True)
class RetrievalChatbot(DeviceAwareModel):
"""
Retrieval-based learning chatbot model.
Uses trained embeddings and FAISS for similarity search.
"""
def __init__(
self,
config: ChatbotConfig,
device: str = None,
strategy=None,
reranker: Optional[CrossEncoderReranker] = None,
summarizer: Optional[Summarizer] = None,
mode: str = 'training'
):
super().__init__()
self.config = config
self.strategy = strategy
self.device = device or self._setup_default_device()
self.mode = mode.lower()
# Initialize reranker, summarizer, tokenizer, and encoder
self.encoder = self._initialize_encoder()
self.tokenizer = self.encoder.tokenizer
self.reranker = reranker or self._initialize_reranker()
self.summarizer = summarizer or self._initialize_summarizer()
# Initialize data pipeline
logger.info("Initializing TFDataPipeline.")
self.data_pipeline = TFDataPipeline(
config=self.config,
tokenizer=self.tokenizer,
encoder=self.encoder,
response_pool=[],
query_embeddings_cache={},
)
# Collect unique responses from dialogues
if self.mode == 'inference':
logger.info("Mode set to 'inference'. Loading FAISS index and response pool.")
self._load_faiss_index_and_responses()
elif self.mode != 'training':
logger.error(f"Unsupported mode in RetrievalChatbot init: {self.mode}")
raise ValueError(f"Unsupported mode in RetrievalChatbot init: {self.mode}")
# Initialize training history
self.history = {
"train_loss": [],
"val_loss": [],
"train_metrics": {},
"val_metrics": {}
}
def _setup_default_device(self) -> str:
"""Set up default device if none is provided."""
if tf.config.list_physical_devices('GPU'):
return 'GPU'
else:
return 'CPU'
def _initialize_reranker(self) -> CrossEncoderReranker:
"""Initialize the CrossEncoderReranker."""
logger.info("Initializing default CrossEncoderReranker...")
return CrossEncoderReranker(model_name=self.config.cross_encoder_model)
def _initialize_summarizer(self) -> Summarizer:
"""Initialize the Summarizer."""
return Summarizer(
tokenizer=self.tokenizer,
model_name=self.config.summarizer_model,
max_summary_length=self.config.max_context_length // 4,
device=self.device,
max_summary_rounds=2
)
def _initialize_encoder(self) -> SentenceTransformer:
"""Initialize the Sentence Transformer model."""
logger.info("Initializing SentenceTransformer encoder model...")
encoder = SentenceTransformer(self.config.pretrained_model)
return encoder
def _load_faiss_index_and_responses(self) -> None:
"""Load FAISS index and response pool for inference."""
try:
logger.info(f"Loading FAISS index from {self.data_pipeline.faiss_index_file_path}...")
self.data_pipeline.load_faiss_index(self.data_pipeline.faiss_index_file_path)
logger.info("FAISS index loaded successfully.")
# Load response pool
response_pool_path = self.data_pipeline.faiss_index_file_path.replace('.index', '_responses.json')
if os.path.exists(response_pool_path):
with open(response_pool_path, 'r', encoding='utf-8') as f:
self.data_pipeline.response_pool = json.load(f)
logger.info(f"Loaded {len(self.data_pipeline.response_pool)} responses from {response_pool_path}.")
else:
logger.error(f"Response pool file not found at {response_pool_path}.")
raise FileNotFoundError(f"Response pool file not found at {response_pool_path}.")
# Validate FAISS index and response pool
self.data_pipeline.validate_faiss_index()
logger.info("FAISS index and response pool validated successfully.")
except Exception as e:
logger.error(f"Failed to load FAISS index and response pool: {e}")
raise
@classmethod
def load_model(cls, load_dir: Union[str, Path], mode: str = 'training') -> 'RetrievalChatbot':
"""Load chatbot model and configuration."""
load_dir = Path(load_dir)
# Load config
config_path = load_dir / "config.json"
if config_path.exists():
with open(config_path, "r") as f:
config = ChatbotConfig.from_dict(json.load(f))
logger.info("Loaded ChatbotConfig from config.json.")
else:
raise FileNotFoundError(f"Config file not found at {config_path}. Please ensure it exists.")
# Initialize chatbot
chatbot = cls(config, mode=mode)
# Load Sentence Transformer
model_path = load_dir / "sentence_transformer"
if model_path.exists():
# Load locally saved model
chatbot.encoder = SentenceTransformer(str(model_path))
logger.info("Loaded SentenceTransformer model from local path successfully.")
else:
# Load from pre-trained model hub
chatbot.encoder = SentenceTransformer(config.pretrained_model)
logger.info(f"Loaded SentenceTransformer model '{config.pretrained_model}' from the hub successfully.")
return chatbot
@classmethod
def _prepare_model_for_inference(cls, chatbot: 'RetrievalChatbot', load_dir: Path) -> None:
"""Load inference components."""
try:
# Load FAISS index
faiss_path = load_dir / 'faiss_indices/faiss_index_production.index'
if faiss_path.exists():
chatbot.index = faiss.read_index(str(faiss_path))
logger.info("FAISS index loaded successfully")
else:
raise FileNotFoundError(f"FAISS index not found at {faiss_path}")
# Load response pool
response_pool_path = load_dir / 'faiss_indices/faiss_index_production_responses.json'
if response_pool_path.exists():
with open(response_pool_path, 'r') as f:
chatbot.response_pool = json.load(f)
logger.info(f"Loaded {len(chatbot.response_pool)} responses")
else:
raise FileNotFoundError(f"Response pool not found at {response_pool_path}")
# Verify dimensions match
if chatbot.index.d != chatbot.config.embedding_dim:
raise ValueError(
f"FAISS index dimension {chatbot.index.d} doesn't match "
f"model dimension {chatbot.config.embedding_dim}"
)
except Exception as e:
logger.error(f"Error loading inference components: {e}")
raise
def save_models(self, save_dir: Union[str, Path]):
"""Save SentenceTransformer model and config."""
save_dir = Path(save_dir)
save_dir.mkdir(parents=True, exist_ok=True)
# Save config
with open(save_dir / "config.json", "w") as f:
json.dump(self.config.to_dict(), f, indent=2)
# Save Sentence Transformer
self.encoder.save(save_dir / "sentence_transformer")
logger.info(f"Model and config saved to {save_dir}.")
def retrieve_responses(
self,
query: str,
top_k: int = 10,
reranker: Optional[CrossEncoderReranker] = None,
summarizer: Optional[Summarizer] = None,
summarize_threshold: int = 512,
boost_factor: float = 1.15
) -> List[Tuple[str, float]]:
"""
Retrieve top-k responses using FAISS and cross-encoder re-ranking.
Args:
query: The user's input text.
top_k: Number of responses to return.
reranker: Optional reranker for refined scoring.
summarizer: Optional summarizer for long queries.
summarize_threshold: Threshold to summarize long queries.
boost_factor: Factor to boost scores for keyword matches.
Returns:
List of (response_text, final_score).
"""
def sigmoid(x: float) -> float:
return 1 / (1 + np.exp(-x))
# Summarize long queries
if summarizer and len(query.split()) > summarize_threshold:
logger.info(f"Query is long ({len(query.split())} words). Summarizing...")
query = summarizer.summarize_text(query)
logger.info(f"Summarized query: {query}")
# Detect domain for query
detected_domain = self.detect_domain_from_query(query)
#logger.info(f"Detected domain: {detected_domain}")
# Retrieve candidates from FAISS
#logger.info("Retrieving initial candidates from FAISS...")
faiss_candidates = self.data_pipeline.retrieve_responses(query, top_k=top_k * 10)
if not faiss_candidates:
logger.warning("No candidates retrieved from FAISS.")
return []
# Filter out-of-domain responses
if detected_domain != 'other':
in_domain_candidates = [c for c in faiss_candidates if c[0]["domain"] == detected_domain]
if in_domain_candidates:
faiss_candidates = in_domain_candidates
else:
logger.info(f"No in-domain responses found for '{query}'. Using all candidates.")
# Re-rank candidates using Cross-Encoder
#logger.info("Re-ranking candidates using Cross-Encoder...")
texts = [item[0]["text"] for item in faiss_candidates] # Extract response texts
faiss_scores = [item[1] for item in faiss_candidates]
if reranker is None:
reranker = CrossEncoderReranker(model_name=self.config.cross_encoder_model)
ce_logits = reranker.rerank(query, texts, max_length=256) # Re-rank responses
# Combine FAISS and Cross-Encoder scores
final_candidates = []
for resp_text, faiss_score, logit in zip(texts, faiss_scores, ce_logits):
ce_prob = sigmoid(logit) # Cross-encoder score in range [0, 1]
faiss_norm = (faiss_score + 1) / 2 # Normalize FAISS score to range [0, 1]
combined_score = 0.75 * ce_prob + 0.25 * faiss_norm
# Boost score based on keyword match
query_keywords = self.extract_keywords(query)
if query_keywords and any(kw in resp_text.lower() for kw in query_keywords):
combined_score *= boost_factor
# Adjust score based on length
length_adjusted_score = self.length_adjust_score(resp_text, combined_score)
final_candidates.append((resp_text, length_adjusted_score))
# Sort and return top-k results
final_candidates.sort(key=lambda x: x[1], reverse=True)
#logger.info(f"Returning top-{top_k} re-ranked responses.")
return final_candidates[:top_k]
def extract_keywords(self, query: str) -> List[str]:
"""
Return any domain keywords present in the query (lowercased).
"""
domain_keywords = {
'restaurant': ['restaurant', 'dining', 'food', 'dine', 'reservation', 'table', 'menu', 'cuisine', 'eat', 'place to eat', 'hungry', 'chef', 'dish', 'meal', 'brunch', 'bistro', 'buffet', 'catering', 'gourmet', 'fast food', 'fine dining', 'takeaway', 'delivery', 'restaurant booking'],
'movie': ['movie', 'cinema', 'film', 'ticket', 'showtime', 'showing', 'theater', 'flick', 'screening', 'film ticket', 'film show', 'blockbuster', 'premiere', 'trailer', 'director', 'actor', 'actress', 'plot', 'genre', 'screen', 'sequel', 'animation', 'documentary'],
'ride_share': ['ride', 'taxi', 'uber', 'lyft', 'car service', 'pickup', 'dropoff', 'driver', 'cab', 'hailing', 'rideshare', 'ride hailing', 'carpool', 'chauffeur', 'transit', 'transportation', 'hail ride'],
'coffee': ['coffee', 'café', 'cafe', 'starbucks', 'espresso', 'latte', 'mocha', 'americano', 'barista', 'brew', 'cappuccino', 'macchiato', 'iced coffee', 'cold brew', 'espresso machine', 'coffee shop', 'tea', 'chai', 'java', 'bean', 'roast', 'decaf'],
'pizza': ['pizza', 'delivery', 'order food', 'pepperoni', 'topping', 'pizzeria', 'slice', 'pie', 'margherita', 'deep dish', 'thin crust', 'cheese', 'oven', 'tossed', 'sauce', 'garlic bread', 'calzone'],
'auto': ['car', 'vehicle', 'repair', 'maintenance', 'mechanic', 'oil change', 'garage', 'auto shop', 'tire', 'check engine', 'battery', 'transmission', 'brake', 'engine diagnostics', 'carwash', 'detail', 'alignment', 'exhaust', 'spark plug', 'dashboard'],
}
query_lower = query.lower()
found = set()
for domain, kw_list in domain_keywords.items():
for kw in kw_list:
if kw in query_lower:
found.add(kw)
return list(found)
def length_adjust_score(self, text: str, base_score: float) -> float:
"""
Penalize very short lines, reward longer lines.
"""
words = text.split()
wcount = len(words)
# Penalty if under 4 words
if wcount < 4:
return base_score * 0.8
# Bonus for lines > 15 words
if wcount > 15:
bonus = min(0.03, 0.001 * (wcount - 15))
base_score += bonus
return base_score
def detect_domain_from_query(self, query: str) -> str:
"""
Detect the domain of the query based on keywords. Used for filtering FAISS search.
"""
domain_patterns = {
'restaurant': r'\b(restaurant|restaurants?|dining|food|foods?|dine|reservation|reservations?|table|tables?|menu|menus?|cuisine|cuisines?|eat|eats?|place\s?to\s?eat|places\s?to\s?eat|hungry|chef|chefs?|dish|dishes?|meal|meals?|fork|forks?|knife|knives?|spoon|spoons?|brunch|bistro|buffet|buffets?|catering|caterings?|gourmet|fast\s?food|fine\s?dining|takeaway|takeaways?|delivery|deliveries|restaurant\s?booking)\b',
'movie': r'\b(movie|movies?|cinema|cinemas?|film|films?|ticket|tickets?|showtime|showtimes?|showing|showings?|theater|theaters?|flick|flicks?|screening|screenings?|film\s?ticket|film\s?tickets?|film\s?show|film\s?shows?|blockbuster|blockbusters?|premiere|premieres?|trailer|trailers?|director|directors?|actor|actors?|actress|actresses?|plot|plots?|genre|genres?|screen|screens?|sequel|sequels?|animation|animations?|documentary|documentaries)\b',
'ride_share': r'\b(ride|rides?|taxi|taxis?|uber|lyft|car\s?service|car\s?services?|pickup|pickups?|dropoff|dropoffs?|driver|drivers?|cab|cabs?|hailing|hailings?|rideshare|rideshares?|ride\s?hailing|ride\s?hailings?|carpool|carpools?|chauffeur|chauffeurs?|transit|transits?|transportation|transportations?|hail\s?ride|hail\s?rides?)\b',
'coffee': r'\b(coffee|coffees?|café|cafés?|cafe|cafes?|starbucks|espresso|espressos?|latte|lattes?|mocha|mochas?|americano|americanos?|barista|baristas?|brew|brews?|cappuccino|cappuccinos?|macchiato|macchiatos?|iced\s?coffee|iced\s?coffees?|cold\s?brew|cold\s?brews?|espresso\s?machine|espresso\s?machines?|coffee\s?shop|coffee\s?shops?|tea|teas?|chai|chais?|java|javas?|bean|beans?|roast|roasts?|decaf)\b',
'pizza': r'\b(pizza|pizzas?|delivery|deliveries|order\s?food|order\s?foods?|pepperoni|pepperonis?|topping|toppings?|pizzeria|pizzerias?|slice|slices?|pie|pies?|margherita|margheritas?|deep\s?dish|deep\s?dishes?|thin\s?crust|thin\s?crusts?|cheese|cheeses?|oven|ovens?|tossed|tosses?|sauce|sauces?|garlic\s?bread|garlic\s?breads?|calzone|calzones?)\b',
'auto': r'\b(car|cars?|vehicle|vehicles?|repair|repairs?|maintenance|maintenances?|mechanic|mechanics?|oil\s?change|oil\s?changes?|garage|garages?|auto\s?shop|auto\s?shops?|tire|tires?|check\s?engine|check\s?engines?|battery|batteries?|transmission|transmissions?|brake|brakes?|engine\s?diagnostics|engine\s?diagnostic|carwash|carwashes?|detail|details?|alignment|alignments?|exhaust|exhausts?|spark\s?plug|spark\s?plugs?|dashboard|dashboards?)\b',
}
# Check for matches
for domain, pattern in domain_patterns.items():
if re.search(pattern, query.lower()):
return domain
return 'other'
def is_numeric_response(self, text: str) -> bool:
"""
Return True if `text` is purely digits and/or spaces.
"""
pattern = r'^[\s]*[\d]+([\s.,\d]+)*[\s]*$'
return bool(re.match(pattern, text.strip()))
def introduction_message(self) -> None:
"""Print an introduction message to introduce the chatbot."""
print(
"\nAssistant: Hello! I'm a simple chatbot assistant. I've been trained to answer "
"basic questions about topics including restaurants, movies, ride sharing, coffee, and pizza. "
"Please ask me a question and I'll do my best to assist you."
)
def run_interactive_chat(self, quality_checker, show_alternatives=False):
"""Separate function for interactive chat loop."""
# Chatbot introduction
self.introduction_message()
# Chat loop
while True:
try:
user_input = input("\nYou: ")
except (KeyboardInterrupt, EOFError):
print("\nAssistant: Goodbye!")
break
if user_input.lower() in ["quit", "exit", "bye"]:
print("\nAssistant: Goodbye!")
break
response, candidates, metrics, top_response_score = self.chat(
query=user_input,
conversation_history=None,
quality_checker=quality_checker,
top_k=10
)
print(f"\nAssistant: {response}")
if show_alternatives and candidates and metrics.get("is_confident", False):
print("\n Alternative responses:")
for resp, score in candidates[1:4]:
print(f" Score: {score:.4f} - {resp}")
elif top_response_score < 0.7:
print("\n[Low Confidence]: Consider rephrasing your query for better assistance.")
def chat(
self,
query: str,
conversation_history: Optional[List[Tuple[str, str]]] = None,
quality_checker: Optional['ResponseQualityChecker'] = None,
top_k: int = 10,
) -> Tuple[str, List[Tuple[str, float]], Dict[str, Any]]:
"""
Live chat with the chatbot. Uses same processing flow as validation, except for context handling and quality checking.
"""
@self.run_on_device
def get_response(self_arg, query_arg):
# Build conversation context string
conversation_str = self_arg._build_conversation_context(query_arg, conversation_history)
# Retrieve and re-rank
responses = self_arg.retrieve_responses(
query=conversation_str,
top_k=top_k,
reranker=self_arg.reranker,
summarizer=self_arg.summarizer,
summarize_threshold=512
)
# Handle low confidence or empty responses
if not responses:
return ("I'm sorry, but I couldn't find a relevant response.", [], {})
# Analyze is_confident and computed score when returning the top response
metrics = quality_checker.check_response_quality(query_arg, responses)
is_confident = metrics.get('is_confident', False)
top_response_score = responses[0][1]
# if uncertain, ask for clarification
if not is_confident or top_response_score < 0.5:
return ("I need more information to provide a good answer. Could you please clarify?", responses, metrics, top_response_score)
# Return the top response
return responses[0][0], responses, metrics, top_response_score
return get_response(self, query)
def _build_conversation_context(
self,
query: str,
conversation_history: Optional[List[Tuple[str, str]]]
) -> str:
"""
Build conversation context string from conversation history,
using literal <USER> and <ASSISTANT> tokens (no tokenizer special index).
"""
USER_TOKEN = "<USER>"
ASSISTANT_TOKEN = "<ASSISTANT>"
if not conversation_history:
return f"{USER_TOKEN} {query}"
conversation_parts = []
for user_txt, assistant_txt in conversation_history:
# Insert literal tokens
conversation_parts.append(f"{USER_TOKEN} {user_txt}")
conversation_parts.append(f"{ASSISTANT_TOKEN} {assistant_txt}")
conversation_parts.append(f"{USER_TOKEN} {query}")
return "\n".join(conversation_parts)
def train_model(
self,
tfrecord_file_path: str,
epochs: int = 20,
batch_size: int = 16,
validation_split: float = 0.2,
checkpoint_dir: str = "checkpoints/",
use_lr_schedule: bool = True,
peak_lr: float = 1e-5,
warmup_steps_ratio: float = 0.1,
early_stopping_patience: int = 3,
min_delta: float = 1e-4,
test_mode: bool = False,
initial_epoch: int = 0
) -> None:
"""
Train the retrieval model using a pre-prepared TFRecord dataset.
- Checkpoint loading/restoring
- LR scheduling
- Epoch/iteration tracking
- Training-history logging
- Early stopping
- Custom loss function (Contrastive loss with hard negative sampling))
"""
logger.info("Starting training with pre-prepared TFRecord dataset...")
def parse_tfrecord_fn(example_proto, max_length, neg_samples):
"""
Parses a single TFRecord example.
"""
feature_description = {
'query_ids': tf.io.FixedLenFeature([max_length], tf.int64),
'positive_ids': tf.io.FixedLenFeature([max_length], tf.int64),
'negative_ids': tf.io.FixedLenFeature([neg_samples * max_length], tf.int64),
}
parsed_features = tf.io.parse_single_example(example_proto, feature_description)
query_ids = tf.cast(parsed_features['query_ids'], tf.int32)
positive_ids = tf.cast(parsed_features['positive_ids'], tf.int32)
negative_ids = tf.cast(parsed_features['negative_ids'], tf.int32)
negative_ids = tf.reshape(negative_ids, [neg_samples, max_length])
return query_ids, positive_ids, negative_ids
# Count total records in TFRecord
raw_dataset = tf.data.TFRecordDataset(tfrecord_file_path)
total_pairs = sum(1 for _ in raw_dataset)
logger.info(f"Total pairs in TFRecord: {total_pairs}")
train_size = int(total_pairs * (1 - validation_split))
val_size = total_pairs - train_size
steps_per_epoch = math.ceil(train_size / batch_size)
val_steps = math.ceil(val_size / batch_size)
total_steps = steps_per_epoch * epochs
buffer_size = max(1, total_pairs // 2) # 50% of the dataset for shuffling
logger.info(f"Training pairs: {train_size}")
logger.info(f"Validation pairs: {val_size}")
logger.info(f"Steps per epoch: {steps_per_epoch}")
logger.info(f"Validation steps: {val_steps}")
logger.info(f"Total steps: {total_steps}")
# Set up optimizer & LR schedule
if use_lr_schedule:
warmup_steps = int(total_steps * warmup_steps_ratio)
lr_schedule = self._get_lr_schedule(
total_steps=total_steps,
peak_lr=tf.cast(peak_lr, tf.float32),
warmup_steps=warmup_steps
)
self.optimizer = tf.keras.optimizers.Adam(learning_rate=lr_schedule)
logger.info("Using custom learning rate schedule.")
else:
self.optimizer = tf.keras.optimizers.Adam(learning_rate=tf.cast(peak_lr, tf.float32))
logger.info("Using fixed learning rate.")
# Dummy step to force initialization
dummy_input = tf.zeros((1, self.config.max_context_length), dtype=tf.int32)
with tf.GradientTape() as tape:
dummy_output = self.encoder(dummy_input)
dummy_loss = tf.cast(tf.reduce_mean(dummy_output), tf.float32)
dummy_grads = tape.gradient(dummy_loss, self.encoder.trainable_variables)
self.optimizer.apply_gradients(zip(dummy_grads, self.encoder.trainable_variables))
# Create checkpoint and manager
checkpoint = tf.train.Checkpoint(
epoch=tf.Variable(0, dtype=tf.int32),
optimizer=self.optimizer,
model=self.encoder
)
# Create a CheckpointManager
manager = tf.train.CheckpointManager(
checkpoint,
directory=checkpoint_dir,
max_to_keep=3,
checkpoint_name='ckpt'
)
# Restore from existing checkpoint if one is provided
latest_checkpoint = manager.latest_checkpoint
history_path = Path(checkpoint_dir) / 'training_history.json'
# Log epoch losses across runs, including restore from checkpoint
if not hasattr(self, 'history'):
self.history = {'train_loss': [], 'val_loss': [], 'learning_rate': []}
if latest_checkpoint and not test_mode:
# Debug checkpoint loading
# logger.info(f"\nTrying to load checkpoint from: {latest_checkpoint}")
# reader = tf.train.load_checkpoint(latest_checkpoint)
# shape_from_key = reader.get_variable_to_shape_map()
# dtype_from_key = reader.get_variable_to_dtype_map()
# logger.info("\nCheckpoint Variables:")
# for key in shape_from_key:
# logger.info(f"{key}: dtype={dtype_from_key[key]} - Shape: {shape_from_key[key]}")
status = checkpoint.restore(latest_checkpoint)
status.assert_consumed()
logger.info(f"Restored from checkpoint: {latest_checkpoint}")
logger.info(f"Optimizer iterations after restore: {self.optimizer.iterations.numpy()}")
# Verify learning rate after restore
if use_lr_schedule:
current_lr = float(lr_schedule(self.optimizer.iterations))
else:
current_lr = float(self.optimizer.learning_rate.numpy())
logger.info(f"Current learning rate after restore: {current_lr:.2e}")
# Derive initial_epoch from checkpoint name if not passed in
ckpt_number = int(latest_checkpoint.split('ckpt-')[-1])
if initial_epoch == 0:
initial_epoch = ckpt_number
# Assign to checkpoint.epoch for counting
checkpoint.epoch.assign(tf.cast(initial_epoch, tf.int32))
logger.info(f"Resuming from epoch {initial_epoch}")
# Load history from file:
if history_path.exists():
try:
with open(history_path, 'r') as f:
self.history = json.load(f)
logger.info(f"Loaded previous training history from {history_path}")
except Exception as e:
logger.warning(f"Could not load history, starting fresh: {e}")
# Save custom weights not being saved in the full model.
# This was a bugfix to extract weights from a checkpoint without retraining.
# Before updating save_models, only Distilbert weights were being saved (custom layers were missed).
# Not needed, also not harmful.
self.save_models(Path(checkpoint_dir) / "pretrained_full_model")
logger.info(f"Manually saved custom weights after restore.")
else:
logger.info("Starting training from scratch")
checkpoint.epoch.assign(tf.cast(0, tf.int32))
initial_epoch = 0
# Set up TensorBoard
log_dir = Path(checkpoint_dir) / "tensorboard_logs"
log_dir.mkdir(parents=True, exist_ok=True)
current_time = datetime.datetime.now().strftime("%Y%m%d-%H%M%S")
train_log_dir = str(log_dir / f"train_{current_time}")
val_log_dir = str(log_dir / f"val_{current_time}")
train_summary_writer = tf.summary.create_file_writer(train_log_dir)
val_summary_writer = tf.summary.create_file_writer(val_log_dir)
logger.info(f"TensorBoard logs will be saved in {log_dir}")
# Parse dataset
dataset = tf.data.TFRecordDataset(tfrecord_file_path)
# Debug mode uses small subset. Useful for CPU debugging.
if test_mode:
subset_size = 200
dataset = dataset.take(subset_size)
logger.info(f"TEST MODE: Using only {subset_size} examples")
# Recompute sizes, steps, epochs, etc., as needed
total_pairs = subset_size
train_size = int(total_pairs * (1 - validation_split))
val_size = total_pairs - train_size
batch_size = min(batch_size, val_size)
steps_per_epoch = math.ceil(train_size / batch_size)
val_steps = math.ceil(val_size / batch_size)
total_steps = steps_per_epoch * epochs
buffer_size = max(1, total_pairs // 10)
epochs = min(epochs, 5) # For quick debug
early_stopping_patience = 2
logger.info(f"New training pairs: {train_size}")
logger.info(f"New validation pairs: {val_size}")
dataset = dataset.map(
lambda x: parse_tfrecord_fn(x, self.config.max_context_length, self.data_pipeline.neg_samples),
num_parallel_calls=tf.data.AUTOTUNE
)
# Train/val split
train_dataset = dataset.take(train_size)
val_dataset = dataset.skip(train_size).take(val_size)
# Shuffle and batch
train_dataset = train_dataset.shuffle(buffer_size=buffer_size)
train_dataset = train_dataset.batch(batch_size, drop_remainder=True)
train_dataset = train_dataset.prefetch(tf.data.AUTOTUNE)
val_dataset = val_dataset.batch(batch_size, drop_remainder=False)
val_dataset = val_dataset.prefetch(tf.data.AUTOTUNE)
val_dataset = val_dataset.cache()
# Training loop
best_val_loss = float("inf")
epochs_no_improve = 0
for epoch in range(int(checkpoint.epoch.numpy()) + 1, epochs + 1):
checkpoint.epoch.assign(epoch)
logger.info(f"Starting Epoch {epoch}...")
epoch_loss_avg = tf.keras.metrics.Mean(dtype=tf.float32)
batches_processed = 0
try:
train_pbar = tqdm(
total=steps_per_epoch,
desc=f"Training Epoch {epoch}",
unit="batch"
)
is_tqdm_train = True
except ImportError:
train_pbar = None
is_tqdm_train = False
# --- Training ---
for q_batch, p_batch, n_batch in train_dataset:
loss, grad_norm, post_clip_norm = self.train_step(q_batch, p_batch, n_batch)
epoch_loss_avg(loss)
batches_processed += 1
# Log to TensorBoard
with train_summary_writer.as_default():
step = (epoch - 1) * steps_per_epoch + batches_processed
tf.summary.scalar("loss", tf.cast(loss, tf.float32), step=step)
tf.summary.scalar("gradient_norm_pre_clip", tf.cast(grad_norm, tf.float32), step=step)
tf.summary.scalar("gradient_norm_post_clip", tf.cast(post_clip_norm, tf.float32), step=step)
# Update progress bar
if use_lr_schedule:
current_lr = float(lr_schedule(self.optimizer.iterations))
else:
current_lr = float(self.optimizer.learning_rate.numpy())
if is_tqdm_train:
train_pbar.update(1)
train_pbar.set_postfix({
"loss": f"{loss.numpy():.4f}",
"pre_clip": f"{grad_norm.numpy():.2e}",
"post_clip": f"{post_clip_norm.numpy():.2e}",
"lr": f"{current_lr:.2e}",
"batches": f"{batches_processed}/{steps_per_epoch}"
})
gc.collect()
# End the epoch early if we've processed all steps
if batches_processed >= steps_per_epoch:
break
if is_tqdm_train and train_pbar:
train_pbar.close()
# --- Validation ---
val_loss_avg = tf.keras.metrics.Mean(dtype=tf.float32)
val_batches_processed = 0
try:
val_pbar = tqdm(total=val_steps, desc="Validation", unit="batch")
is_tqdm_val = True
except ImportError:
val_pbar = None
is_tqdm_val = False
last_valid_val_loss = None
valid_batches = False
for q_batch, p_batch, n_batch in val_dataset:
# If batch is too small, skip
if tf.shape(q_batch)[0] < 2:
logger.warning(f"Skipping validation batch of size {tf.shape(q_batch)[0]}")
continue
valid_batches = True
val_loss = self.validation_step(q_batch, p_batch, n_batch)
val_loss_avg(val_loss)
last_valid_val_loss = val_loss
val_batches_processed += 1
if is_tqdm_val:
val_pbar.update(1)
val_pbar.set_postfix({
"val_loss": f"{val_loss.numpy():.4f}",
"batches": f"{val_batches_processed}/{val_steps}"
})
gc.collect()
if val_batches_processed >= val_steps:
break
if not valid_batches:
# If no valid batch is found, fallback
logger.warning("No valid validation batches in this epoch")
if last_valid_val_loss is not None:
val_loss = last_valid_val_loss
val_loss_avg(val_loss)
else:
val_loss = epoch_loss_avg.result()
val_loss_avg(val_loss)
if is_tqdm_val and val_pbar:
val_pbar.close()
# End of epoch: final stats
train_loss = epoch_loss_avg.result().numpy()
val_loss = val_loss_avg.result().numpy()
logger.info(f"Epoch {epoch} Complete: Train Loss={train_loss:.4f}, Val Loss={val_loss:.4f}")
# TensorBoard epoch logs
with train_summary_writer.as_default():
tf.summary.scalar("epoch_loss", train_loss, step=epoch)
with val_summary_writer.as_default():
tf.summary.scalar("val_loss", val_loss, step=epoch)
# Save checkpoint
manager.save()
# Save model for iterative testing/inference
model_save_path = Path(checkpoint_dir) / f"model_epoch_{epoch}"
self.save_models(model_save_path)
logger.info(f"Saved model for epoch {epoch} at {model_save_path}")
# Update local history
self.history['train_loss'].append(train_loss)
self.history['val_loss'].append(val_loss)
self.history.setdefault('learning_rate', []).append(current_lr)
def convert_to_py_floats(obj):
if isinstance(obj, dict):
return {k: convert_to_py_floats(v) for k, v in obj.items()}
elif isinstance(obj, list):
return [convert_to_py_floats(x) for x in obj]
elif isinstance(obj, (np.float32, np.float64)):
return float(obj)
elif tf.is_tensor(obj):
return float(obj.numpy())
else:
return obj
json_history = convert_to_py_floats(self.history)
# Save training history to file every epoch
with open(history_path, 'w') as f:
json.dump(json_history, f)
logger.info(f"Saved training history to {history_path}")
# Early stopping
if val_loss < best_val_loss - min_delta:
best_val_loss = val_loss
epochs_no_improve = 0
logger.info(f"Validation loss improved to {val_loss:.4f}. Reset patience.")
else:
epochs_no_improve += 1
logger.info(f"No improvement this epoch. Patience: {epochs_no_improve}/{early_stopping_patience}")
if epochs_no_improve >= early_stopping_patience:
logger.info("Early stopping triggered.")
break
logger.info("Training completed!")
@tf.function
def train_step(
self,
q_batch: tf.Tensor,
p_batch: tf.Tensor,
n_batch: tf.Tensor
) -> tf.Tensor:
"""
Single training step using queries, positives, and hard negatives.
"""
with tf.GradientTape() as tape:
# Encode queries, positives, and negatives
q_enc = self.encoder(q_batch, training=True) # [batch_size, embed_dim]
p_enc = self.encoder(p_batch, training=True) # [batch_size, embed_dim]
shape = tf.shape(n_batch)
bs = shape[0]
neg_samples = shape[1]
# Flatten negatives to feed them in one pass: [batch_size * neg_samples, max_length]
n_batch_flat = tf.reshape(n_batch, [bs * neg_samples, shape[2]])
n_enc_flat = self.encoder(n_batch_flat, training=True) # [bs*neg_samples, embed_dim]
# Reshape back => [batch_size, neg_samples, embed_dim]
n_enc = tf.reshape(n_enc_flat, [bs, neg_samples, -1])
# Combine the positive embedding and negative embeddings along dim=1: shape [batch_size, 1 + neg_samples, embed_dim]
# Col 1 is the pos, subsequent cols are negatives
combined_p_n = tf.concat([tf.expand_dims(p_enc, axis=1), n_enc], axis=1) # [bs, (1+neg_samples), embed_dim]
# Compute scores: dot product of q_enc with each column in combined_p_n. `tf.einsum` handles the batch dimension
dot_products = tf.cast(tf.einsum('bd,bkd->bk', q_enc, combined_p_n), tf.float32)
labels = tf.zeros([bs], dtype=tf.int32) # Keep labels as int32
loss = tf.nn.sparse_softmax_cross_entropy_with_logits(
labels=labels,
logits=dot_products
)
loss = tf.cast(tf.reduce_mean(loss), tf.float32)
# Calculate gradients and clip
gradients = tape.gradient(loss, self.encoder.trainable_variables)
gradients_norm = tf.cast(tf.linalg.global_norm(gradients), tf.float32)
max_grad_norm = tf.constant(1.5, dtype=tf.float32)
gradients, _ = tf.clip_by_global_norm(gradients, max_grad_norm, gradients_norm)
post_clip_norm = tf.cast(tf.linalg.global_norm(gradients), tf.float32)
self.optimizer.apply_gradients(zip(gradients, self.encoder.trainable_variables))
return loss, gradients_norm, post_clip_norm
@tf.function
def validation_step(
self,
q_batch: tf.Tensor,
p_batch: tf.Tensor,
n_batch: tf.Tensor
) -> tf.Tensor:
"""
Single validation step using queries, positives, and hard negatives.
Same idea as train_step, but without gradient updates.
"""
q_enc = self.encoder(q_batch, training=False)
p_enc = self.encoder(p_batch, training=False)
shape = tf.shape(n_batch)
bs = shape[0]
neg_samples = shape[1]
n_batch_flat = tf.reshape(n_batch, [bs * neg_samples, shape[2]])
n_enc_flat = self.encoder(n_batch_flat, training=False)
n_enc = tf.reshape(n_enc_flat, [bs, neg_samples, -1])
combined_p_n = tf.concat(
[tf.expand_dims(p_enc, axis=1), n_enc],
axis=1
)
dot_products = tf.cast(tf.einsum('bd,bkd->bk', q_enc, combined_p_n), tf.float32)
labels = tf.zeros([bs], dtype=tf.int32)
loss = tf.nn.sparse_softmax_cross_entropy_with_logits(
labels=labels,
logits=dot_products
)
loss = tf.cast(tf.reduce_mean(loss), tf.float32)
return loss
def _get_lr_schedule(
self,
total_steps: int,
peak_lr: float,
warmup_steps: int
) -> tf.keras.optimizers.schedules.LearningRateSchedule:
"""
Custom learning rate schedule with warmup and cosine decay.
"""
class CustomSchedule(tf.keras.optimizers.schedules.LearningRateSchedule):
def __init__(
self,
total_steps: int,
peak_lr: float,
warmup_steps: int
):
super().__init__()
self.total_steps = tf.cast(total_steps, tf.float32)
self.peak_lr = tf.cast(peak_lr, tf.float32)
# warmup_steps 10% of total_steps
adjusted_warmup_steps = min(warmup_steps, max(1, total_steps // 10))
self.warmup_steps = tf.cast(adjusted_warmup_steps, tf.float32)
# Calculate constants
self.initial_lr = tf.cast(self.peak_lr * 0.1, tf.float32)
self.min_lr = tf.cast(self.peak_lr * 0.01, tf.float32)
logger.info(f"Learning rate schedule initialized:")
logger.info(f" Initial LR: {float(self.initial_lr):.6f}")
logger.info(f" Peak LR: {float(self.peak_lr):.6f}")
logger.info(f" Min LR: {float(self.min_lr):.6f}")
logger.info(f" Warmup steps: {int(self.warmup_steps)}")
logger.info(f" Total steps: {int(self.total_steps)}")
def __call__(self, step):
step = tf.cast(step, tf.float32)
# Warmup
warmup_factor = tf.cast(tf.minimum(1.0, step / self.warmup_steps), tf.float32)
warmup_lr = self.initial_lr + (self.peak_lr - self.initial_lr) * warmup_factor
# Decay
decay_steps = tf.cast(tf.maximum(1.0, self.total_steps - self.warmup_steps), tf.float32)
decay_factor = tf.cast((step - self.warmup_steps) / decay_steps, tf.float32)
decay_factor = tf.cast(tf.minimum(tf.maximum(0.0, decay_factor), 1.0), tf.float32)
cosine_decay = tf.cast(0.5 * (1.0 + tf.cos(tf.constant(math.pi, dtype=tf.float32) * decay_factor)), tf.float32)
decay_lr = self.min_lr + (self.peak_lr - self.min_lr) * cosine_decay
final_lr = tf.where(step < self.warmup_steps, warmup_lr, decay_lr)
# Ensure valid lr
final_lr = tf.maximum(self.min_lr, final_lr)
final_lr = tf.where(tf.math.is_finite(final_lr), final_lr, self.min_lr)
return final_lr
def get_config(self):
return {
"total_steps": self.total_steps,
"peak_lr": self.peak_lr,
"warmup_steps": self.warmup_steps,
}
return CustomSchedule(total_steps, peak_lr, warmup_steps)