import logging

import numpy as np
import pandas as pd
from sentence_transformers import (
    InputExample,
    LoggingHandler,
    SentenceTransformer,
    losses,
)
from sentence_transformers.evaluation import EmbeddingSimilarityEvaluator
from sklearn.model_selection import train_test_split
from torch.utils.data import DataLoader

# Configure logging
logging.basicConfig(
    format="%(asctime)s - %(message)s",
    datefmt="%Y-%m-%d %H:%M:%S",
    level=logging.INFO,
    handlers=[LoggingHandler()],
)

model_name = "sentence-transformers/all-MiniLM-L6-v2"
model = SentenceTransformer(model_name, device="cuda")
# num_examples = 10_000

# Perform train-test split
# Example fake data with right types (for testing)
# import faker
# fake = Faker()
# train_data = [
#     (fake.city(), fake.city(), np.random.rand())
#     for _ in range(num_examples)
# ]
data = pd.read_csv("city_distances_sample.csv")
MAX_DISTANCE = 20_037.5  # global max distance
# MAX_DISTANCE = data["distance"].max()  # about 5k

print(f"{MAX_DISTANCE=}")
train_data = [
    (row["city_from"], row["city_to"], 1 - row["distance"] / MAX_DISTANCE)
    for _, row in data.iterrows()
]

np.random.seed(1992)
np.random.shuffle(train_data)
train_examples = examples = [
    InputExample(texts=[city_from, city_to], label=dist)
    for city_from, city_to, dist in train_data
]

train_examples, val_examples = train_test_split(
    examples, test_size=0.2, random_state=21
)
# validation examples can be something like templated sentences
# that maintain the same distance as the cities (same context)
# should probably add training examples like that too if needed
batch_size = 16
num_examples = len(train_examples)
steps_per_epoch = num_examples // batch_size

print(f"\nHead of training data (size: {num_examples}):")
print(train_data[:10], "\n")

# Create DataLoaders for train and validation datasets
train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=16)

print("TRAINING")
# Configure the training arguments
training_args = {
    "output_path": "./output",
    # "evaluation_steps": steps_per_epoch,  # already evaluates at the end of each epoch
    "epochs": 5,
    "warmup_steps": 500,
    "optimizer_params": {"lr": 2e-5},
    # "weight_decay": 0,  # not sure if this helps but works fine without setting it.
    "scheduler": "WarmupLinear",
    "save_best_model": True,
    "checkpoint_path": "./checkpoints_absmax_split",
    "checkpoint_save_steps": steps_per_epoch,
    "checkpoint_save_total_limit": 20,
}
print(f"TRAINING ARGUMENTS:\n {training_args}")

train_loss = losses.CosineSimilarityLoss(model)

# Create an evaluator for validation dataset
evaluator = EmbeddingSimilarityEvaluator.from_input_examples(
    val_examples, write_csv=True
)

model.fit(
    train_objectives=[(train_dataloader, train_loss)],
    evaluator=evaluator,
    **training_args,
)