This time we'll solve a problem of transribing hebrew words in english, also known as g2p (grapheme2phoneme)
Unlike what most deep learning researchers do, we won't only train it to maximize likelihood of correct translation, but also employ reinforcement learning to actually teach it to translate with as few errors as possible.
One notable property of Hebrew is that it's consonant language. That is, there are no wovels in the written language. One could represent wovels with diacritics above consonants, but you don't expect people to do that in everyay life.
Therefore, some hebrew characters will correspond to several english letters and others - to none, so we should use encoder-decoder architecture to figure that out.
Encoder-decoder architectures are about converting anything to anything, including
We chose simplified Hebrew->English machine translation for words and short phrases (character-level), as it is relatively quick to train even without a gpu cluster.
In [ ]:
# If True, only translates phrases shorter than 20 characters (way easier).
EASY_MODE = True
# Please keep it until you're done debugging your code
# If false, works with all phrases (please switch to this mode for homework assignment)
# way we translate. Either "he-to-en" or "en-to-he"
MODE = "he-to-en"
# maximal length of _generated_ output, does not affect training
MAX_OUTPUT_LENGTH = 50 if not EASY_MODE else 20
REPORT_FREQ = 100 # how often to evaluate validation score
We shall store dataset as a dictionary
{ word1:[translation1,translation2,...], word2:[...],...}.
This is mostly due to the fact that many words have several correct translations.
We have implemented this thing for you so that you can focus on more interesting parts.
Attention python2 users! You may want to cast everything to unicode later during homework phase, just make sure you do it everywhere.
In [ ]:
import numpy as np
from collections import defaultdict
word_to_translation = defaultdict(list) # our dictionary
bos = '_'
eos = ';'
with open("main_dataset.txt", encoding='utf8') as fin:
for line in fin:
en, he = line[:-1].lower().replace(bos, ' ').replace(eos,
' ').split('\t')
word, trans = (he, en) if MODE == 'he-to-en' else (en, he)
if len(word) < 3:
continue
if EASY_MODE:
if max(len(word), len(trans)) > 20:
continue
word_to_translation[word].append(trans)
print("size = ", len(word_to_translation))
In [ ]:
# get all unique lines in source language
all_words = np.array(list(word_to_translation.keys()))
# get all unique lines in translation language
all_translations = np.array(
[ts for all_ts in word_to_translation.values() for ts in all_ts])
In [ ]:
from sklearn.model_selection import train_test_split
train_words, test_words = train_test_split(
all_words, test_size=0.1, random_state=42)
In [ ]:
from voc import Vocab
inp_voc = Vocab.from_lines(''.join(all_words), bos=bos, eos=eos, sep='')
out_voc = Vocab.from_lines(''.join(all_translations), bos=bos, eos=eos, sep='')
In [ ]:
# Here's how you cast lines into ids and backwards.
batch_lines = all_words[:5]
batch_ids = inp_voc.to_matrix(batch_lines)
batch_lines_restored = inp_voc.to_lines(batch_ids)
print("lines")
print(batch_lines)
print("\nwords to ids (0 = bos, 1 = eos):")
print(batch_ids)
print("\nback to words")
print(batch_lines_restored)
Draw word/translation length distributions to estimate the scope of the task.
In [ ]:
import matplotlib.pyplot as plt
%matplotlib inline
plt.figure(figsize=[8, 4])
plt.subplot(1, 2, 1)
plt.title("words")
plt.hist(list(map(len, all_words)), bins=20)
plt.subplot(1, 2, 2)
plt.title('translations')
plt.hist(list(map(len, all_translations)), bins=20)
assignment starts here
Our architecture consists of two main blocks:
Than it gets fed into a model that follows this simple interface:
model.symbolic_translate(inp, **flags) -> out, logp - takes symbolic int32 matrix of hebrew words, produces output tokens sampled from the model and output log-probabilities for all possible tokens at each tick.greedy=True, takes most likely next token at each iteration. Otherwise samples with next token probabilities predicted by model.model.symbolic_score(inp, out, **flags) -> logp - takes symbolic int32 matrices of hebrew words and their english translations. Computes the log-probabilities of all possible english characters given english prefices and hebrew word.That's all! It's as hard as it gets. With those two methods alone you can implement all kinds of prediction and training.
In [ ]:
# set flags here if necessary
import theano
theano.config.floatX = 'float32'
import theano.tensor as T
import lasagne
In [ ]:
from basic_model_theano import BasicTranslationModel
model = BasicTranslationModel(inp_voc, out_voc,
emb_size=64, hid_size=128)
In [ ]:
# Play around with symbolic_translate and symbolic_score
inp = T.constant(np.random.randint(0, 10, [3, 5], dtype='int32'))
out = T.constant(np.random.randint(0, 10, [3, 5], dtype='int32'))
# translate inp (with untrained model)
sampled_out, logp = model.symbolic_translate(inp, greedy=False)
dummy_translate = theano.function([], sampled_out, updates=model.auto_updates)
print("\nSymbolic_translate output:\n", sampled_out, logp)
print("\nSample translations:\n", dummy_translate())
In [ ]:
# score logp(out | inp) with untrained input
logp = model.symbolic_score(inp, out)
dummy_score = theano.function([], logp)
print("\nSymbolic_score output:\n", logp)
print("\nLog-probabilities (clipped):\n", dummy_score()[:, :2, :5])
In [ ]:
# Prepare any operations you want here
inp = T.imatrix("input tokens [batch,time]")
trans, _ = <YOUR CODE: build symbolic translations with greedy = True>
translate_fun = theano.function([inp], trans, updates=model.auto_updates)
def translate(lines):
"""
You are given a list of input lines.
Make your neural network translate them.
:return: a list of output lines
"""
# Convert lines to a matrix of indices
lines_ix = <YOUR CODE>
# Compute translations in form of indices (call your function)
trans_ix = <YOUR CODE>
# Convert translations back into strings
return out_voc.to_lines(trans_ix)
In [ ]:
print("Sample inputs:", all_words[:3])
print("Dummy translations:", translate(all_words[:3]))
assert trans.ndim == 2 and trans.dtype.startswith(
'int'), "trans must be a tensor of integers (token ids)"
assert translate(all_words[:3]) == translate(
all_words[:3]), "make sure translation is deterministic (use greedy=True and disable any noise layers)"
assert type(translate(all_words[:3])) is list and (type(translate(all_words[:1])[0]) is str or type(
translate(all_words[:1])[0]) is unicode), "translate(lines) must return a sequence of strings!"
print("Tests passed!")
LogLikelihood is a poor estimator of model performance.
Therefore, we will use minimal Levenshtein distance. It measures how many characters do we need to add/remove/replace from model translation to make it perfect. Alternatively, one could use character-level BLEU/RougeL or other similar metrics.
The catch here is that Levenshtein distance is not differentiable: it isn't even continuous. We can't train our neural network to maximize it by gradient descent.
In [ ]:
import editdistance # !pip install editdistance
def get_distance(word, trans):
"""
A function that takes word and predicted translation
and evaluates (Levenshtein's) edit distance to closest correct translation
"""
references = word_to_translation[word]
assert len(references) != 0, "wrong/unknown word"
return min(editdistance.eval(trans, ref) for ref in references)
def score(words, bsize=100):
"""a function that computes levenshtein distance for bsize random samples"""
assert isinstance(words, np.ndarray)
batch_words = np.random.choice(words, size=bsize, replace=False)
batch_trans = translate(batch_words)
distances = list(map(get_distance, batch_words, batch_trans))
return np.array(distances, dtype='float32')
In [ ]:
# should be around 5-50 and decrease rapidly after training :)
[score(test_words, 10).mean() for _ in range(5)]
In [ ]:
from agentnet.learning.generic import get_values_for_actions, get_mask_by_eos
class llh_trainer:
# variable for correct answers
input_sequence = T.imatrix("input sequence [batch,time]")
reference_answers = T.imatrix("reference translations [batch, time]")
# Compute log-probabilities of all possible tokens at each step. Use model interface.
logprobs_seq = <YOUR CODE>
# compute mean crossentropy
crossentropy = - get_values_for_actions(logprobs_seq, reference_answers)
mask = get_mask_by_eos(T.eq(reference_answers, out_voc.eos_ix))
loss = T.sum(crossentropy * mask)/T.sum(mask)
# Build weight updates. Use model.weights to get all trainable params.
updates = <YOUR CODE>
train_step = theano.function(
[input_sequence, reference_answers], loss, updates=updates)
Actually run training on minibatches
In [ ]:
import random
def sample_batch(words, word_to_translation, batch_size):
"""
sample random batch of words and random correct translation for each word
example usage:
batch_x,batch_y = sample_batch(train_words, word_to_translations,10)
"""
# choose words
batch_words = np.random.choice(words, size=batch_size)
# choose translations
batch_trans_candidates = list(map(word_to_translation.get, batch_words))
batch_trans = list(map(random.choice, batch_trans_candidates))
return inp_voc.to_matrix(batch_words), out_voc.to_matrix(batch_trans)
In [ ]:
bx, by = sample_batch(train_words, word_to_translation, batch_size=3)
print("Source:")
print(bx)
print("Target:")
print(by)
In [ ]:
from IPython.display import clear_output
from tqdm import tqdm, trange # or use tqdm_notebook,tnrange
loss_history = []
editdist_history = []
for i in trange(25000):
loss = llh_trainer.train_step(
*sample_batch(train_words, word_to_translation, 32))
loss_history.append(loss)
if (i+1) % REPORT_FREQ == 0:
clear_output(True)
current_scores = score(test_words)
editdist_history.append(current_scores.mean())
plt.figure(figsize=(12, 4))
plt.subplot(131)
plt.title('train loss / traning time')
plt.plot(loss_history)
plt.grid()
plt.subplot(132)
plt.title('val score distribution')
plt.hist(current_scores, bins=20)
plt.subplot(133)
plt.title('val score / traning time')
plt.plot(editdist_history)
plt.grid()
plt.show()
print("llh=%.3f, mean score=%.3f" %
(np.mean(loss_history[-10:]), np.mean(editdist_history[-10:])))
In [ ]:
for word in train_words[:10]:
print("%s -> %s" % (word, translate([word])[0]))
In [ ]:
test_scores = []
for start_i in trange(0, len(test_words), 32):
batch_words = test_words[start_i:start_i+32]
batch_trans = translate(batch_words)
distances = list(map(get_distance, batch_words, batch_trans))
test_scores.extend(distances)
print("Supervised test score:", np.mean(test_scores))
First we need to define loss function as a custom theano operation.
The simple way to do so is
@theano.compile.as_op(input_types,output_type(s),infer_shape)
def my_super_function(inputs):
return outputsYour task is to implement _compute_levenshtein function that takes matrices of words and translations, along with input masks, then converts those to actual words and phonemes and computes min-levenshtein via get_distance function above.
In [ ]:
@theano.compile.as_op([T.imatrix]*2, [T.fvector], lambda _, shapes: [shapes[0][:1]])
def _compute_levenshtein(words_ix, trans_ix):
"""
A custom theano operation that computes levenshtein loss for predicted trans.
Params:
- words_ix - a matrix of letter indices, shape=[batch_size,word_length]
- words_mask - a matrix of zeros/ones,
1 means "word is still not finished"
0 means "word has already finished and this is padding"
- trans_mask - a matrix of output letter indices, shape=[batch_size,translation_length]
- trans_mask - a matrix of zeros/ones, similar to words_mask but for trans_ix
Please implement the function and make sure it passes tests from the next cell.
"""
# convert words to strings
words = <YOUR CODE: restore words(a list of strings) from words_ix>
assert type(words) is list and type(
words[0]) is str and len(words) == len(words_ix)
# convert translations to lists
translations = <YOUR CODE: restore trans (a list of lists of phonemes) from trans_ix
assert type(translations) is list and type(
translations[0]) is str and len(translations) == len(trans_ix)
# computes levenstein distances. can be arbitrary python code.
distances = <YOUR CODE: apply get_distance to each pair of[words, translations]>
assert type(distances) in (list, tuple, np.ndarray) and len(
distances) == len(words_ix)
distances = np.array(list(distances), dtype='float32')
return distances
# forbid gradient
from theano.gradient import disconnected_grad
def compute_levenshtein(*args):
return disconnected_grad(_compute_levenshtein(*[arg.astype('int32') for arg in args]))
Simple test suite to make sure your implementation is correct. Hint: if you run into any bugs, feel free to use print from inside _compute_levenshtein.
In [ ]:
# test suite
# sample random batch of (words, correct trans, wrong trans)
batch_words = np.random.choice(train_words, size=100)
batch_trans = list(map(random.choice, map(
word_to_translation.get, batch_words)))
batch_trans_wrong = np.random.choice(all_translations, size=100)
batch_words_ix = T.constant(inp_voc.to_matrix(batch_words))
batch_trans_ix = T.constant(out_voc.to_matrix(batch_trans))
batch_trans_wrong_ix = T.constant(out_voc.to_matrix(batch_trans_wrong))
In [ ]:
# assert compute_levenshtein is zero for ideal translations
correct_answers_score = compute_levenshtein(
batch_words_ix, batch_trans_ix).eval()
assert np.all(correct_answers_score ==
0), "a perfect translation got nonzero levenshtein score!"
print("Everything seems alright!")
In [ ]:
# assert compute_levenshtein matches actual scoring function
wrong_answers_score = compute_levenshtein(
batch_words_ix, batch_trans_wrong_ix).eval()
true_wrong_answers_score = np.array(
list(map(get_distance, batch_words, batch_trans_wrong)))
assert np.all(wrong_answers_score ==
true_wrong_answers_score), "for some word symbolic levenshtein is different from actual levenshtein distance"
print("Everything seems alright!")
In this section you'll implement algorithm called self-critical sequence training (here's an article).
The algorithm is a vanilla policy gradient with a special baseline.
$$ \nabla J = E_{x \sim p(s)} E_{y \sim \pi(y|x)} \nabla log \pi(y|x) \cdot (R(x,y) - b(x)) $$Here reward R(x,y) is a negative levenshtein distance (since we minimize it). The baseline b(x) represents how well model fares on word x.
In practice, this means that we compute baseline as a score of greedy translation, $b(x) = R(x,y_{greedy}(x)) $.
Luckily, we already obtained the required outputs: model.greedy_translations, model.greedy_mask and we only need to compute levenshtein using compute_levenshtein function.
In [ ]:
class trainer:
input_sequence = T.imatrix("input tokens [batch,time]")
# use model to __sample__ symbolic translations given input_sequence
sample_translations, sample_logp = <YOUR CODE>
auto_updates = model.auto_updates
# use model to __greedy__ symbolic translations given input_sequence
greedy_translations, greedy_logp = <YOUR CODE>
greedy_auto_updates = model.auto_updates
# Note: you can use model.symbolic_translate(...,unroll_scan=True,max_len=MAX_OUTPUT_LENGTH)
# to run much faster at a cost of longer compilation
rewards = - compute_levenshtein(input_sequence, sample_translations)
baseline = <YOUR CODE: compute __negative__ levenshtein for greedy mode>
# compute advantage using rewards and baseline
advantage = <YOUR CODE: compute advantage>
# compute log_pi(a_t|s_t), shape = [batch, seq_length]
logprobs_phoneme = get_values_for_actions(sample_logp, sample_translations)
# policy gradient
J = logprobs_phoneme*advantage[:, None]
mask = get_mask_by_eos(T.eq(sample_translations, out_voc.eos_ix))
loss = - T.sum(J*mask) / T.sum(mask)
# regularize with negative entropy. Don't forget the sign!
# note: for entropy you need probabilities for all tokens (sample_logp), not just phoneme_logprobs
entropy = <YOUR CODE: compute entropy matrix of shape[batch, seq_length], H = -sum(p*log_p), don't forget the sign!>
assert entropy.ndim == 2, "please make sure elementwise entropy is of shape [batch,time]"
loss -= 0.01*T.sum(entropy*mask) / T.sum(mask)
# compute weight updates, clip by norm
grads = T.grad(loss, model.weights)
grads = lasagne.updates.total_norm_constraint(grads, 50)
updates = lasagne.updates.adam(grads, model.weights, learning_rate=1e-5)
train_step = theano.function([input_sequence], loss,
updates=auto_updates+greedy_auto_updates+updates)
In [ ]:
for i in trange(100000):
loss_history.append(
trainer.train_step(sample_batch(
train_words, word_to_translation, 32)[0])
)
if (i+1) % REPORT_FREQ == 0:
clear_output(True)
current_scores = score(test_words)
editdist_history.append(current_scores.mean())
plt.figure(figsize=(8, 4))
plt.subplot(121)
plt.title('val score distribution')
plt.hist(current_scores, bins=20)
plt.subplot(122)
plt.title('val score / traning time')
plt.plot(editdist_history)
plt.grid()
plt.show()
print("J=%.3f, mean score=%.3f" %
(np.mean(loss_history[-10:]), np.mean(editdist_history[-10:])))
In [ ]:
model.translate("EXAMPLE;")
In [ ]:
for word in train_words[:10]:
print("%s -> %s" % (word, translate([word])[0]))
In [ ]:
test_scores = []
for start_i in trange(0, len(test_words), 32):
batch_words = test_words[start_i:start_i+32]
batch_trans = translate(batch_words)
distances = list(map(get_distance, batch_words, batch_trans))
test_scores.extend(distances)
print("Supervised test score:", np.mean(test_scores))
# ^^ If you get Out Of Memory, please replace this with batched computation
In this section we want you to finally restart with EASY_MODE=False and experiment to find a good model/curriculum for that task.
We recommend you to start with the following architecture
encoder---decoder
P(y|h)
^
LSTM -> LSTM
^ ^
biLSTM -> LSTM
^ ^
input y_prevNote: you can fit all 4 state tensors of both LSTMs into a in a single state - just assume that it contains, for example, [h0, c0, h1, c1] - pack it in encode and update in decode.
Here are some cool ideas on what you can do then.
General tips & tricks:
tf.variable_scope(unique_name, reuse=False).Formal criteria: To get 5 points we want you to build an architecture that:
There's more than one way to connect decoder to encoder
The most effective (and cool) of those is, of course, attention. You can read more about attention in this nice blog post. The easiest way to begin is to use "soft" attention with "additive" or "dot-product" intermediate layers.
Tips
Once your model made it through several epochs, it is a good idea to visualize attention maps to understand what your model has actually learned
There's more stuff here
he-pron-wiktionary.txt.
In [ ]:
assert not EASY_MODE, "make sure you set EASY_MODE = False at the top of the notebook."
[your report/log here or anywhere you please]
Contributions: This notebook is brought to you by