-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
206 lines (159 loc) · 5.73 KB
/
main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
import collections
from absl import app
from absl import flags
from absl import logging
import functools
import flax
from flax import nn
from flax import optim
from flax.training import common_utils
from flax.training.checkpoints import save_checkpoint, restore_checkpoint
import jax
from jax import random
from jax import numpy as jnp
import numpy as np
import nlp
from utils import batchify, get_batch, process_data
from utils import Tokenizer, Vocab
import model
FLAGS = flags.FLAGS
flags.DEFINE_float(
"learning_rate", default=0.001, help=("The learning rate for the Adam optimizer.")
)
flags.DEFINE_integer("batch_size", default=64, help=("Batch size for training."))
flags.DEFINE_integer("num_epochs", default=5, help=("Number of training epochs."))
flags.DEFINE_integer(
"hidden_size", default=200, help=("Hidden size for the GRU and MLP.")
)
flags.DEFINE_integer(
"embedding_size", default=200, help=("Size of the word embeddings.")
)
flags.DEFINE_integer("seq_len", default=35, help=("Sequence length in the dataset."))
flags.DEFINE_integer(
"seed", default=0, help=("Random seed for network initialization.")
)
@jax.jit
def compute_cross_entropy(logits, targets):
onehot_targets = common_utils.onehot(targets, logits.shape[-1])
loss = -jnp.sum(onehot_targets * nn.log_softmax(logits), axis=-1)
return loss
@jax.jit
def train_step(optimizer, inputs, carry, targets, rng):
rng, new_rng = jax.random.split(rng)
def loss_fn(model, carry):
with nn.stochastic(rng):
carry, logits = model(inputs, carry)
loss = jnp.mean(compute_cross_entropy(logits, targets))
return loss, (logits, carry)
(loss, out), grad = jax.value_and_grad(loss_fn, has_aux=True)(
optimizer.target, carry
)
optimizer = optimizer.apply_gradient(grad)
_, carry = out
return optimizer, loss, carry, new_rng
@jax.jit
def eval_step(model, inputs, carry, targets):
carry, logits = model(inputs, carry)
loss = jnp.mean(compute_cross_entropy(logits, targets))
return loss, carry
def log(epoch, train_metrics, valid_metrics):
train_loss = train_metrics["loss"] / train_metrics["total"]
logging.info(
"Epoch %02d train loss %.4f valid loss %.4f",
epoch + 1,
train_loss,
valid_metrics["loss"],
)
def evaluate(model, dataset):
count = 0
total_loss = 0.0
carry = nn.GRUCell.initialize_carry(
jax.random.PRNGKey(0), (FLAGS.batch_size,), FLAGS.hidden_size
)
for i in range(len(dataset) // FLAGS.batch_size):
inputs, targets = get_batch(dataset, FLAGS.batch_size, i)
count = count + inputs.shape[0]
loss, carry = eval_step(model, inputs, carry, targets)
total_loss += loss.item()
loss = total_loss / count
metrics = dict(loss=loss)
return metrics
def train_model(
model, learning_rate, num_epochs, seed, train_data, valid_data, batch_size
):
train_metrics = collections.defaultdict(float)
rng = jax.random.PRNGKey(seed)
optimizer = flax.optim.Adam(learning_rate=learning_rate).create(model)
carry = nn.GRUCell.initialize_carry(
jax.random.PRNGKey(0), (FLAGS.batch_size,), FLAGS.hidden_size
)
for epoch in range(num_epochs):
for i in range(len(train_data) // batch_size):
data, targets = get_batch(train_data, batch_size, i)
optimizer, loss, carry, rng = train_step(
optimizer, data, carry, targets, rng
)
train_metrics["loss"] += loss * data.shape[0]
train_metrics["total"] += data.shape[0]
valid_metrics = evaluate(optimizer.target, valid_data)
log(epoch, train_metrics, valid_metrics)
# save_checkpoint(".", optimizer.target, epoch + 1, keep=1)
return optimizer.target
def generate_text(
model, vocab, max_length=100, temperature=0.5, top_k=3, start_letter="T",
):
output_text = start_letter
carry = nn.GRUCell.initialize_carry(jax.random.PRNGKey(0), (1,), FLAGS.hidden_size)
for i in range(max_length):
input = vocab.numericalize(output_text[-1])
input_t = jnp.array(input, dtype=jnp.int32).reshape(1, 1)
carry, pred = model(input_t, carry)
prob = nn.softmax(pred / temperature, axis=1)
# output_text += vocab.textify(prob.argmax().tolist())[0]
prob_np = np.array(prob)[0]
top_k_index = prob_np.argsort()[-top_k:]
next_char = np.random.choice(
top_k_index.tolist(), 1, prob_np[top_k_index].tolist()
)
output_text += vocab.textify(next_char.item())[0]
return output_text
def main(argv):
data = nlp.load_dataset("tiny_shakespeare")
train_data = data["train"][0]["text"]
valid_data = data["test"][0]["text"]
tokenize = Tokenizer()
vocabulary = Vocab()
train_data, valid_data, vocab_size = process_data(
train_data, valid_data, tokenize, vocabulary, FLAGS.batch_size
)
charnn = model.create_model(
seed=FLAGS.seed,
batch_size=FLAGS.batch_size,
seq_len=FLAGS.batch_size,
model_kwargs=dict(
vocab_size=vocab_size,
embedding_size=FLAGS.embedding_size,
hidden_size=FLAGS.hidden_size,
output_size=vocab_size,
),
)
trained_model = train_model(
model=charnn,
learning_rate=FLAGS.learning_rate,
num_epochs=FLAGS.num_epochs,
seed=FLAGS.seed,
train_data=train_data,
valid_data=valid_data,
batch_size=FLAGS.batch_size,
)
generated_text = generate_text(
trained_model,
vocabulary,
max_length=100,
temperature=0.8,
top_k=3,
start_letter="T",
)
print("Hello Shakespeare: ", generated_text)
if __name__ == "__main__":
app.run(main)