Classical pipeline

import os
os.environ["TOKENIZERS_PARALLELISM"] = "false"

import torch

BATCH_SIZE = 30
EPOCHS = 25
LEARNING_RATE = 3e-2
SEED = 0

Input data

def read_data(filename):
    labels, sentences = [], []
    with open(filename) as f:
        for line in f:
            t = float(line[0])
            labels.append([t, 1-t])
            sentences.append(line[1:].strip())
    return labels, sentences


train_labels, train_data = read_data('datasets/mc_train_data.txt')
dev_labels, dev_data = read_data('datasets/mc_dev_data.txt')
test_labels, test_data = read_data('datasets/mc_test_data.txt')

TESTING = int(os.environ.get('TEST_NOTEBOOKS', '0'))

if TESTING:
    train_labels, train_data = train_labels[:2], train_data[:2]
    dev_labels, dev_data = dev_labels[:2], dev_data[:2]
    test_labels, test_data = test_labels[:2], test_data[:2]
    EPOCHS = 1

Create diagrams

from lambeq import BobcatParser

reader = BobcatParser(verbose='text')

train_diagrams = reader.sentences2diagrams(train_data)
dev_diagrams = reader.sentences2diagrams(dev_data)
test_diagrams = reader.sentences2diagrams(test_data)

Tagging sentences.
Parsing tagged sentences.
Turning parse trees to diagrams.
Tagging sentences.
Parsing tagged sentences.
Turning parse trees to diagrams.
Tagging sentences.
Parsing tagged sentences.
Turning parse trees to diagrams.

Create circuits

from lambeq.backend.tensor import Dim

from lambeq import AtomicType, SpiderAnsatz

ansatz = SpiderAnsatz({AtomicType.NOUN: Dim(2),
                       AtomicType.SENTENCE: Dim(2)})

train_circuits = [ansatz(diagram) for diagram in train_diagrams]
dev_circuits =  [ansatz(diagram) for diagram in dev_diagrams]
test_circuits = [ansatz(diagram) for diagram in test_diagrams]

train_circuits[-1].draw(figsize=(7, 1))

Parameterise

from lambeq import PytorchModel
all_circuits = train_circuits + dev_circuits + test_circuits
model = PytorchModel.from_diagrams(all_circuits)

Define Evaluation Metric

sig = torch.sigmoid

def accuracy(y_hat, y):
    return torch.sum(torch.eq(torch.round(sig(y_hat)), y))/len(y)/2  # half due to double-counting

Initialize Trainer

from lambeq import PytorchTrainer

trainer = PytorchTrainer(
        model=model,
        loss_function=torch.nn.BCEWithLogitsLoss(),
        optimizer=torch.optim.AdamW,    # type: ignore
        learning_rate=LEARNING_RATE,
        epochs=EPOCHS,
        evaluate_functions={"acc": accuracy},
        evaluate_on_train=True,
        verbose='text',
        seed=SEED)

from lambeq import Dataset

train_dataset = Dataset(
            train_circuits,
            train_labels,
            batch_size=BATCH_SIZE)

dev_dataset = Dataset(dev_circuits, dev_labels)

Train

trainer.fit(train_dataset, dev_dataset, log_interval=5)

Epoch 5:   train/loss: 0.6386   valid/loss: 0.7189   train/time: 0.68s   valid/time: 1.47s   train/acc: 0.5786   valid/acc: 0.5333
Epoch 10:  train/loss: 0.5280   valid/loss: 0.6392   train/time: 0.49s   valid/time: 0.15s   train/acc: 0.5857   valid/acc: 0.5833
Epoch 15:  train/loss: 0.4138   valid/loss: 0.4924   train/time: 0.38s   valid/time: 0.27s   train/acc: 0.7500   valid/acc: 0.7500
Epoch 20:  train/loss: 0.1306   valid/loss: 0.2794   train/time: 0.60s   valid/time: 0.14s   train/acc: 0.9857   valid/acc: 0.9500
Epoch 25:  train/loss: 0.0120   valid/loss: 0.0595   train/time: 0.37s   valid/time: 0.21s   train/acc: 0.9929   valid/acc: 0.9833

Training completed!
train/time: 2.52s   train/time_per_epoch: 0.10s   train/time_per_step: 0.03s   valid/time: 2.23s   valid/time_per_eval: 0.09s

Show results

import matplotlib.pyplot as plt
import numpy as np

fig1, ((ax_tl, ax_tr), (ax_bl, ax_br)) = plt.subplots(2, 2, sharey='row', figsize=(10, 6))

ax_tl.set_title('Training set')
ax_tr.set_title('Development set')
ax_bl.set_xlabel('Epochs')
ax_br.set_xlabel('Epochs')
ax_bl.set_ylabel('Accuracy')
ax_tl.set_ylabel('Loss')

colours = iter(plt.rcParams['axes.prop_cycle'].by_key()['color'])
range_ = np.arange(1, trainer.epochs + 1)
ax_tl.plot(range_, trainer.train_epoch_costs, color=next(colours))
ax_bl.plot(range_, trainer.train_eval_results['acc'], color=next(colours))
ax_tr.plot(range_, trainer.val_costs, color=next(colours))
ax_br.plot(range_, trainer.val_eval_results['acc'], color=next(colours))

# print test accuracy
test_acc = accuracy(model.forward(test_circuits), torch.tensor(test_labels))
print('Test accuracy:', test_acc.item())

Test accuracy: 0.9833333492279053