def test_forward(self):
model_config = self.config.model_attributes.cnn_lstm
cnn_lstm = CNNLSTM(model_config)
cnn_lstm.build()
cnn_lstm.init_losses_and_metrics()
self.assertTrue(isinstance(cnn_lstm, torch.nn.Module))
test_sample = Sample()
test_sample.text = torch.randint(1, 79, (10, ), dtype=torch.long)
test_sample.image = torch.randn(3, 320, 480)
test_sample.targets = torch.randn(32)
test_sample_list = SampleList([test_sample])
test_sample_list.dataset_type = "train"
test_sample_list.dataset_name = "clevr"
output = cnn_lstm(test_sample_list)
scores = output["scores"]
loss = output["losses"]["train/clevr/logit_bce"]
accuracy = output["metrics"]["train/clevr/accuracy"]
np.testing.assert_almost_equal(loss.item(), 19.2635, decimal=4)
np.testing.assert_almost_equal(accuracy.item(), 0)
self.assertEqual(scores.size(), torch.Size((1, 32)))
expected_scores = [
-0.7598285675048828, -0.07029829174280167, -0.20382611453533173,
-0.06990239024162292, 0.7965695858001709, 0.4730074405670166,
-0.30569902062416077, 0.4244227707386017, 0.6511023044586182,
0.2480515092611313, -0.5087617635726929, -0.7675772905349731,
0.4361543357372284, 0.0018743239343166351, 0.6774630546569824,
0.30618518590927124, -0.398895800113678, -0.13120117783546448,
-0.4433199465274811, -0.25969570875167847, 0.6798790097236633,
-0.34090861678123474, 0.0384102463722229, 0.2484571784734726,
0.0456063412129879, -0.428459107875824, -0.026385333389043808,
-0.1570669412612915, -0.2377825379371643, 0.3231588304042816,
0.21098048985004425, -0.712349534034729
]
np.testing.assert_almost_equal(scores[0].tolist(),
expected_scores,
decimal=5)
def test_forward(self):
model_config = self.config.model_attributes.cnn_lstm
cnn_lstm = CNNLSTM(model_config)
cnn_lstm.build()
cnn_lstm.init_losses_and_metrics()
self.assertTrue(isinstance(cnn_lstm, torch.nn.Module))
test_sample = Sample()
test_sample.text = torch.randint(1, 79, (10, ), dtype=torch.long)
test_sample.image = torch.randn(3, 320, 480)
test_sample.targets = torch.randn(32)
test_sample_list = SampleList([test_sample])
test_sample_list.dataset_type = "train"
test_sample_list.dataset_name = "clevr"
output = cnn_lstm(test_sample_list)
scores = output["scores"]
loss = output["losses"]["train/logit_bce"]
accuracy = output["metrics"]["train/accuracy"]
np.testing.assert_almost_equal(loss.item(), 23.4751, decimal=4)
np.testing.assert_almost_equal(accuracy.item(), 0)
self.assertEqual(scores.size(), torch.Size((1, 32)))
expected_scores = [
2.2298e-02, -2.4975e-01, -1.1960e-01, -5.0868e-01, -9.3013e-02,
1.3202e-02, -1.7536e-01, -3.1180e-01, 1.5369e-01, 1.4900e-01,
1.9006e-01, -1.9457e-01, 1.4924e-02, -1.1032e-01, 1.3777e-01,
-3.6255e-01, -2.9327e-01, 5.6247e-04, -4.8732e-01, 4.0949e-01,
-1.1069e-01, 2.9696e-01, 4.1903e-02, 6.7062e-02, 7.0094e-01,
-1.9898e-01, -2.9502e-03, -3.9040e-01, 1.2218e-01, 3.7895e-02,
2.4472e-02, 1.7213e-01
]
np.testing.assert_almost_equal(scores[0].tolist(),
expected_scores,
decimal=5)
def get_item(self, idx):
data = self.questions[idx]
# Each call to get_item from dataloader returns a Sample class object which
# collated by our special batch collator to a SampleList which is basically
# a attribute based batch in layman terms
current_sample = Sample()
question = data["question"]
tokens = tokenize(question, keep=[";", ","], remove=["?", "."])
processed = self.text_processor({"tokens": tokens})
current_sample.text = processed["text"]
processed = self.answer_processor({"answers": [data["answer"]]})
current_sample.answers = processed["answers"]
current_sample.targets = processed["answers_scores"]
image_path = os.path.join(self.image_path, data["image_filename"])
image = np.true_divide(Image.open(image_path).convert("RGB"), 255)
image = image.astype(np.float32)
current_sample.image = torch.from_numpy(image.transpose(2, 0, 1))
return current_sample
