Installing Tensorflow

[Sport Driver]

Hello. 

I'm trying to install TensorFlow on Windows 10 and it's just not having it. Tried many things but would go anywhere.

I first used this this video and this video but it wouldn't do anything. Does anyone have any good recommendation what to do?

[WiiManic]

I've found the Puget systems guide on it helpful when I hit some issues getting CUDA installed for it on a Linux machine, so you could try their Windows 10 guide: https://www.pugetsystems.com/labs/hpc/How-to-Install-TensorFlow-with-GPU-Support-on-Windows-10-Without-Installing-CUDA-UPDATED-1419/.

 

Skip out the CUDA/Nvidia driver stuff and install `tensorflow` not `tensorflow-gpu` if you don't have an Nvidia GPU.

[wasab]

you sure it doesn't do anything? tensorflow requires long time to install and training a model takes even longer. I think it would be best to do a real world project to get a grasp of how it works. 

 

Follow this guide. It is about training a neural network that can tell if a user likes a movie or not by what comments they posted.

https://github.com/bentrevett/pytorch-sentiment-analysis/blob/master/3 - Faster Sentiment Analysis.ipynb 

 

You need to install all the dependencies first

pip install torchtext;
python -m spacy download en;
pip install transformers;

 

This is a python script that is mostly a copy and paste using the code snippets from the github documentation link above. I used it in a flask app I made. 

#!/usr/bin/python

import torch
from torchtext import data
from torchtext import datasets
import random
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import time
import spacy
import flask

app = flask.Flask(__name__)


# Set to false to disable training, important to turn it off after training it once or you will wait another hour
TO_TRAIN = True

# The more epochs the more training will be done. Note that without a GPU, each epoch will take upwards 10 minutes.
N_EPOCHS = 5

# File name for the model
FILE_NAME = 'avg_model.pt'

def generate_bigrams(x):
    n_grams = set(zip(*[x[i:] for i in range(2)]))
    for n_gram in n_grams:
        x.append(' '.join(n_gram))
    return x

SEED = 1234
torch.manual_seed(SEED)
torch.backends.cudnn.deterministic = True

TEXT = data.Field(tokenize = 'spacy', preprocessing = generate_bigrams)
LABEL = data.LabelField(dtype = torch.float)

train_data, test_data = datasets.IMDB.splits(TEXT, LABEL)
train_data, valid_data = train_data.split(random_state = random.seed(SEED))

MAX_VOCAB_SIZE = 25_000

TEXT.build_vocab(train_data,
                 max_size = MAX_VOCAB_SIZE,
                 vectors = "glove.6B.100d",
                 unk_init = torch.Tensor.normal_)

LABEL.build_vocab(train_data)

BATCH_SIZE = 64

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

train_iterator, valid_iterator, test_iterator = data.BucketIterator.splits(
    (train_data, valid_data, test_data),
    batch_size = BATCH_SIZE,
    device = device)

class FastText(nn.Module):
    def __init__(self, vocab_size, embedding_dim, output_dim, pad_idx):

        super().__init__()

        self.embedding = nn.Embedding(vocab_size, embedding_dim, padding_idx=pad_idx)

        self.fc = nn.Linear(embedding_dim, output_dim)

    def forward(self, text):

        #text = [sent len, batch size]

        embedded = self.embedding(text)

        #embedded = [sent len, batch size, emb dim]

        embedded = embedded.permute(1, 0, 2)

        #embedded = [batch size, sent len, emb dim]

        pooled = F.avg_pool2d(embedded, (embedded.shape[1], 1)).squeeze(1)

        #pooled = [batch size, embedding_dim]

        return self.fc(pooled)

INPUT_DIM = len(TEXT.vocab)
EMBEDDING_DIM = 100
OUTPUT_DIM = 1
PAD_IDX = TEXT.vocab.stoi[TEXT.pad_token]

model = FastText(INPUT_DIM, EMBEDDING_DIM, OUTPUT_DIM, PAD_IDX)

def count_parameters(model):
    return sum(p.numel() for p in model.parameters() if p.requires_grad)

print(f'The model has {count_parameters(model):,} trainable parameters')

pretrained_embeddings = TEXT.vocab.vectors

model.embedding.weight.data.copy_(pretrained_embeddings)

UNK_IDX = TEXT.vocab.stoi[TEXT.unk_token]

model.embedding.weight.data[UNK_IDX] = torch.zeros(EMBEDDING_DIM)
model.embedding.weight.data[PAD_IDX] = torch.zeros(EMBEDDING_DIM)

optimizer = optim.Adam(model.parameters())

criterion = nn.BCEWithLogitsLoss()

model = model.to(device)
criterion = criterion.to(device)

def binary_accuracy(preds, y):
    """
    Returns accuracy per batch, i.e. if you get 8/10 right, this returns 0.8, NOT 8
    """

    #round predictions to the closest integer
    rounded_preds = torch.round(torch.sigmoid(preds))
    correct = (rounded_preds == y).float() #convert into float for division
    acc = correct.sum() / len(correct)
    return acc

def train(model, iterator, optimizer, criterion):

    epoch_loss = 0
    epoch_acc = 0

    model.train()

    for batch in iterator:

        optimizer.zero_grad()

        predictions = model(batch.text).squeeze(1)

        loss = criterion(predictions, batch.label)

        acc = binary_accuracy(predictions, batch.label)

        loss.backward()

        optimizer.step()

        epoch_loss += loss.item()
        epoch_acc += acc.item()

    return epoch_loss / len(iterator), epoch_acc / len(iterator)

def evaluate(model, iterator, criterion):

    epoch_loss = 0
    epoch_acc = 0

    model.eval()

    with torch.no_grad():

        for batch in iterator:

            predictions = model(batch.text).squeeze(1)

            loss = criterion(predictions, batch.label)

            acc = binary_accuracy(predictions, batch.label)

            epoch_loss += loss.item()
            epoch_acc += acc.item()

    return epoch_loss / len(iterator), epoch_acc / len(iterator)

def epoch_time(start_time, end_time):
    elapsed_time = end_time - start_time
    elapsed_mins = int(elapsed_time / 60)
    elapsed_secs = int(elapsed_time - (elapsed_mins * 60))
    return elapsed_mins, elapsed_secs

if TO_TRAIN:
    print("BEGIN TRAINING FOR " + str(N_EPOCHS) + " epochs")
    best_valid_loss = float('inf')

    for epoch in range(N_EPOCHS):

        start_time = time.time()

        train_loss, train_acc = train(model, train_iterator, optimizer, criterion)
        valid_loss, valid_acc = evaluate(model, valid_iterator, criterion)

        end_time = time.time()

        epoch_mins, epoch_secs = epoch_time(start_time, end_time)

        if valid_loss < best_valid_loss:
            best_valid_loss = valid_loss
            torch.save(model.state_dict(), FILE_NAME)

        print(f'Epoch: {epoch+1:02} | Epoch Time: {epoch_mins}m {epoch_secs}s')
        print(f'\tTrain Loss: {train_loss:.3f} | Train Acc: {train_acc*100:.2f}%')
        print(f'\t Val. Loss: {valid_loss:.3f} |  Val. Acc: {valid_acc*100:.2f}%')

model.load_state_dict(torch.load(FILE_NAME))

test_loss, test_acc = evaluate(model, test_iterator, criterion)

print(f'Test Loss: {test_loss:.3f} | Test Acc: {test_acc*100:.2f}%')

nlp = spacy.load('en')

def predict_sentiment(model, sentence):
    model.eval()
    tokenized = generate_bigrams([tok.text for tok in nlp.tokenizer(sentence)])
    indexed = [TEXT.vocab.stoi[t] for t in tokenized]
    tensor = torch.LongTensor(indexed).to(device)
    tensor = tensor.unsqueeze(1)
    prediction = torch.sigmoid(model(tensor))
    return prediction.item()

@app.route("/predict", methods=["POST"])
def viewResults():
    data = flask.request.json
    sentence = data["sentence"]
    result = predict_sentiment(model, sentence)
    return flask.jsonify(result=result)

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=3000, threaded=True)

 

For me, it took literally one hour to train the neural network. I have no idea of the theory and science behind it but surprisingly it does work. I can tell it something like "This move is so hip!" and it will correctly tell me that it means I liked the film. Pretty neat. 

 

Edit: here are tutorials relevant to tensorflow. I didn't realize pytorch is a different ML framework. Pick whichever one is to your liking and try them out. 

https://github.com/Hvass-Labs/TensorFlow-Tutorials/blob/master/README.md

 

 

[querneu]

Which version of python are you trying to run it?

 

On windows the 3.7.x is practically unstable

 

TensorFlow only ever worked for me in while using python up to 3.6.x

[gwj99]

Hello, from my personal experience, I reccomend installing it by using Anaconda Navigator. Start by Creating a new environment and install tensorflow and desired packages using the navigator. Much more simple than other methods.