File size: 2,990 Bytes
0240c6e |
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 |
import tensorflow as tf
import tf_utils as tfu
from graph.base import GraphPartBase
class Inflect(GraphPartBase):
def __init__(self, for_usage, global_settings, current_settings, optimiser, reset_optimiser):
super().__init__(for_usage, global_settings, current_settings, optimiser, reset_optimiser, 'inflect', ["Loss", "AccuracyByChar", "Accuracy"])
self.chars_count = self.chars_count + 1
self.start_char_index = global_settings['start_token']
self.end_char_index = global_settings['end_token']
self.results = []
self.x_cls = []
self.ys = []
self.y_seq_lens = []
self.y_cls = []
self.keep_drops = []
self.decoder_keep_drops = []
def __build_graph_for_device__(self, x, x_seq_len, batch_size, cls=None):
self.xs.append(x)
self.x_seq_lens.append(x_seq_len)
x_cls = tf.placeholder(dtype=tf.int32, shape=(None,), name='XClass')
self.x_cls.append(x_cls)
if batch_size is None:
batch_size = self.settings['batch_size']
y = tf.placeholder(dtype=tf.int32, shape=(None, None), name='Y')
self.ys.append(y)
y_seq_len = tf.placeholder(dtype=tf.int32, shape=(None,), name='YSeqLen')
self.y_seq_lens.append(y_seq_len)
y_cls = tf.placeholder(dtype=tf.int32, shape=(None,), name='YClass')
self.y_cls.append(y_cls)
tfu.seq2seq(self,
batch_size,
x,
y_cls,
x_seq_len,
y,
x_cls,
y_seq_len)
def __update_feed_dict__(self, op_name, feed_dict, batch, dev_num):
feed_dict[self.x_cls[dev_num]] = batch['x_cls']
feed_dict[self.y_cls[dev_num]] = batch['y_cls']
feed_dict[self.ys[dev_num]] = batch['y']
feed_dict[self.y_seq_lens[dev_num]] = batch['y_seq_len']
feed_dict[self.keep_drops[dev_num]] = self.settings['keep_drop']
feed_dict[self.decoder_keep_drops[dev_num]] = self.settings['decoder']['keep_drop']
def __load_dataset__(self, operation_name):
items = list(tfu.load_inflect_dataset(
self.dataset_path,
self.devices_count,
operation_name,
self.settings['batch_size']
))
return items
def transfer_learning_init(self, sess):
my_prefix = f"{self.main_scope_name}/"
vars = {
var.name[len(my_prefix):]: var
for var in tf.global_variables(my_prefix)
if "Adam" not in var.name
}
lem_prefix = f"Lemm/"
lem_vars = {
var.name[len(lem_prefix):]: var
for var in tf.global_variables(lem_prefix)
if "Adam" not in var.name
}
for key in vars:
my_var = vars[key]
lem_var = lem_vars[key]
value = sess.run(lem_var)
sess.run(my_var.assign(value))
print()
|