| import os |
| os.system("pip install gradio==2.8.0b22") |
| os.system("pip install -r requirements.txt") |
| os.system("pip freeze") |
| from huggingface_hub import from_pretrained_keras |
| import numpy as np |
| import pandas as pd |
| import tensorflow as tf |
| import tensorflow_hub as hub |
| import tensorflow_text as text |
| from tensorflow import keras |
| import gradio as gr |
|
|
|
|
| def make_bert_preprocessing_model(sentence_features, seq_length=128): |
| """Returns Model mapping string features to BERT inputs. |
| |
| Args: |
| sentence_features: A list with the names of string-valued features. |
| seq_length: An integer that defines the sequence length of BERT inputs. |
| |
| Returns: |
| A Keras Model that can be called on a list or dict of string Tensors |
| (with the order or names, resp., given by sentence_features) and |
| returns a dict of tensors for input to BERT. |
| """ |
| |
| input_segments = [ |
| tf.keras.layers.Input(shape=(), dtype=tf.string, name=ft) |
| for ft in sentence_features |
| ] |
| |
| |
| bert_preprocess = hub.load(bert_preprocess_path) |
| tokenizer = hub.KerasLayer(bert_preprocess.tokenize, |
| name="tokenizer") |
| |
| segments = [tokenizer(s) for s in input_segments] |
| |
| truncated_segments = segments |
| |
| packer = hub.KerasLayer(bert_preprocess.bert_pack_inputs, |
| arguments=dict(seq_length=seq_length), |
| name="packer") |
| model_inputs = packer(truncated_segments) |
| return keras.Model(input_segments, model_inputs) |
|
|
|
|
| def preprocess_image(image_path, resize): |
| extension = tf.strings.split(image_path)[-1] |
| |
| image = tf.io.read_file(image_path) |
| if extension == b"jpg": |
| image = tf.image.decode_jpeg(image, 3) |
| else: |
| image = tf.image.decode_png(image, 3) |
| |
| image = tf.image.resize(image, resize) |
| return image |
|
|
| def preprocess_text(text_1, text_2): |
| |
| text_1 = tf.convert_to_tensor([text_1]) |
| text_2 = tf.convert_to_tensor([text_2]) |
| |
| output = bert_preprocess_model([text_1, text_2]) |
| |
| output = {feature: tf.squeeze(output[feature]) for feature in bert_input_features} |
| |
| return output |
|
|
| def preprocess_text_and_image(sample, resize): |
| |
| image_1 = preprocess_image(sample['image_1_path'], resize) |
| image_2 = preprocess_image(sample['image_2_path'], resize) |
| |
| text = preprocess_text(sample['text_1'], sample['text_2']) |
| |
| return {"image_1": image_1, "image_2": image_2, "text": text} |
|
|
|
|
| def classify_info(image_1, text_1, image_2, text_2): |
|
|
| sample = dict() |
| sample['image_1_path'] = image_1 |
| sample['image_2_path'] = image_2 |
| sample['text_1'] = text_1 |
| sample['text_2'] = text_2 |
|
|
| dataframe = pd.DataFrame(sample, index=[0]) |
|
|
| ds = tf.data.Dataset.from_tensor_slices((dict(dataframe), [0])) |
| ds = ds.map(lambda x, y: (preprocess_text_and_image(x, resize), y)).cache() |
| batch_size = 1 |
| auto = tf.data.AUTOTUNE |
| ds = ds.batch(batch_size).prefetch(auto) |
| output = model.predict(ds) |
|
|
| outputs = dict() |
| |
| outputs[labels[0]] = float(output[0][0]) |
| outputs[labels[1]] = float(output[0][1]) |
| outputs[labels[2]] = float(output[0][2]) |
| |
| return outputs |
|
|
|
|
| model = from_pretrained_keras("keras-io/multimodal-entailment") |
| resize = (128, 128) |
| bert_input_features = ["input_word_ids", "input_type_ids", "input_mask"] |
| bert_model_path = ("https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-2_H-256_A-4/1") |
| bert_preprocess_path = "https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3" |
| bert_preprocess_model = make_bert_preprocessing_model(['text_1', 'text_2']) |
|
|
| labels = {0: "Contradictory", 1: "Implies", 2: "No Entailment"} |
|
|
| block = gr.Blocks() |
|
|
| examples = [['examples/image_1.png', '#IndiaFightsCorona:\n\nNearly 4.5 million beneficiaries vaccinated against #COVID19 in 19 days.\n\nIndia is the fastest country to cross landmark of vaccinating 4 million beneficiaries in merely 18 days.\n\n#StaySafe #IndiaWillWin #Unite2FightCorona https://t.co/beGDQfd06S', 'examples/image_2.jpg', '#IndiaFightsCorona:\n\nIndia has become the fastest nation to reach 4 million #COVID19 vaccinations ; it took only 18 days to administer the first 4 million #vaccines\n\n:@MoHFW_INDIA Secretary\n\n#StaySafe #IndiaWillWin #Unite2FightCorona https://t.co/9GENQlqtn3']] |
|
|
|
|
| with block: |
| gr.Markdown("Multimodal Entailment") |
| with gr.Tab("Hypothesis"): |
| with gr.Row(): |
| gr.Markdown("Upload hypothesis image:") |
| image_1 = gr.inputs.Image(type="filepath") |
| text_1 = gr.inputs.Textbox(lines=5) |
| |
| with gr.Tab("Premise"): |
| with gr.Row(): |
| gr.Markdown("Upload premise image:") |
| image_2 = gr.inputs.Image(type="filepath") |
| text_2 = gr.inputs.Textbox(lines=5) |
|
|
|
|
| run = gr.Button("Run") |
| label = gr.outputs.Label() |
| run.click(classify_info, inputs=[image_1, text_1, image_2, text_2], outputs=label) |
|
|
| block.launch() |
|
|