本文整理汇总了Python中config.batch_size方法的典型用法代码示例。如果您正苦于以下问题：Python config.batch_size方法的具体用法？Python config.batch_size怎么用？Python config.batch_size使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在模块config的用法示例。

在下文中一共展示了config.batch_size方法的29个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于我们的系统推荐出更棒的Python代码示例。

示例1: get_loader

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def get_loader(train=False, val=False, test=False, trainval=False):

""" Returns a data loader for the desired split """

split = VQA(

utils.path_for(train=train, val=val, test=test, trainval=trainval, question=True),

utils.path_for(train=train, val=val, test=test, trainval=trainval, answer=True),

config.preprocessed_trainval_path if not test else config.preprocessed_test_path,

answerable_only=train or trainval,

dummy_answers=test,

)

loader = torch.utils.data.DataLoader(

split,

batch_size=config.batch_size,

shuffle=train or trainval, # only shuffle the data in training

pin_memory=True,

num_workers=config.data_workers,

collate_fn=collate_fn,

)

return loader

开发者ID:KaihuaTang，项目名称:VQA2.0-Recent-Approachs-2018.pytorch，代码行数:20，

示例2: __init__

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def __init__(self):

self.vocab = Vocab(config.vocab_path, config.vocab_size)

self.batcher = Batcher(config.train_data_path, self.vocab, mode='train',

batch_size=config.batch_size, single_pass=False)

time.sleep(5)

if not os.path.exists(config.log_root):

os.mkdir(config.log_root)

self.model_dir = os.path.join(config.log_root, 'train_model')

if not os.path.exists(self.model_dir):

os.mkdir(self.model_dir)

self.eval_log = os.path.join(config.log_root, 'eval_log')

if not os.path.exists(self.eval_log):

os.mkdir(self.eval_log)

self.summary_writer = tf.compat.v1.summary.FileWriter(self.eval_log)

开发者ID:wyu-du，项目名称:Reinforce-Paraphrase-Generation，代码行数:19，

示例3: adv_train_generator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def adv_train_generator(self, g_step):

"""

The gen is trained by MLE-like objective.

"""

total_g_loss = 0

for step in range(g_step):

inp, target = GenDataIter.prepare(self.gen.sample(cfg.batch_size, cfg.batch_size), gpu=cfg.CUDA)

# ===Train===

rewards = self.get_mali_reward(target)

adv_loss = self.gen.adv_loss(inp, target, rewards)

self.optimize(self.gen_adv_opt, adv_loss)

total_g_loss += adv_loss.item()

# ===Test===

self.log.info('[ADV-GEN]: g_loss = %.4f, %s' % (total_g_loss, self.cal_metrics(fmt_str=True)))

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:18，

示例4: adv_train_generator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def adv_train_generator(self, g_step):

total_loss = 0

for step in range(g_step):

real_samples = F.one_hot(self.oracle_data.random_batch()['target'], cfg.vocab_size).float()

gen_samples = self.gen.sample(cfg.batch_size, cfg.batch_size, one_hot=True)

if cfg.CUDA:

real_samples, gen_samples = real_samples.cuda(), gen_samples.cuda()

# ===Train===

d_out_real = self.dis(real_samples)

d_out_fake = self.dis(gen_samples)

g_loss, _ = get_losses(d_out_real, d_out_fake, cfg.loss_type)

self.optimize(self.gen_adv_opt, g_loss, self.gen)

total_loss += g_loss.item()

return total_loss / g_step if g_step != 0 else 0

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:19，

示例5: adv_train_discriminator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def adv_train_discriminator(self, d_step):

total_loss = 0

for step in range(d_step):

real_samples = F.one_hot(self.oracle_data.random_batch()['target'], cfg.vocab_size).float()

gen_samples = self.gen.sample(cfg.batch_size, cfg.batch_size, one_hot=True)

if cfg.CUDA:

real_samples, gen_samples = real_samples.cuda(), gen_samples.cuda()

# ===Train===

d_out_real = self.dis(real_samples)

d_out_fake = self.dis(gen_samples)

_, d_loss = get_losses(d_out_real, d_out_fake, cfg.loss_type)

self.optimize(self.dis_opt, d_loss, self.dis)

total_loss += d_loss.item()

return total_loss / d_step if d_step != 0 else 0

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:19，

示例6: cal_metrics

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def cal_metrics(self, fmt_str=False):

"""

Calculate metrics

:param fmt_str: if return format string for logging

"""

with torch.no_grad():

# Prepare data for evaluation

gen_data = GenDataIter(self.gen.sample(cfg.samples_num, 4 * cfg.batch_size))

# Reset metrics

self.nll_oracle.reset(self.oracle, gen_data.loader)

self.nll_gen.reset(self.gen, self.oracle_data.loader)

self.nll_div.reset(self.gen, gen_data.loader)

if fmt_str:

return ', '.join(['%s = %s' % (metric.get_name(), metric.get_score()) for metric in self.all_metrics])

else:

return [metric.get_score() for metric in self.all_metrics]

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:20，

示例7: train_discriminator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def train_discriminator(self, d_step, d_epoch, phase='MLE'):

"""

Training the discriminator on real_data_samples (positive) and generated samples from gen (negative).

Samples are drawn d_step times, and the discriminator is trained for d_epoch d_epoch.

"""

# prepare loader for validate

global d_loss, train_acc

for step in range(d_step):

# prepare loader for training

pos_samples = self.train_data.target # not re-sample the Oracle data

neg_samples = self.gen.sample(cfg.samples_num, 4 * cfg.batch_size)

dis_data = DisDataIter(pos_samples, neg_samples)

for epoch in range(d_epoch):

# ===Train===

d_loss, train_acc = self.train_dis_epoch(self.dis, dis_data.loader, self.dis_criterion,

self.dis_opt)

# ===Test===

self.log.info('[%s-DIS] d_step %d: d_loss = %.4f, train_acc = %.4f,' % (

phase, step, d_loss, train_acc))

if cfg.if_save and not cfg.if_test:

torch.save(self.dis.state_dict(), cfg.pretrained_dis_path)

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:27，

示例8: adv_train_generator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def adv_train_generator(self, g_step):

total_loss = 0

for step in range(g_step):

real_samples = self.train_data.random_batch()['target']

gen_samples = self.gen.sample(cfg.batch_size, cfg.batch_size, one_hot=True)

if cfg.CUDA:

real_samples, gen_samples = real_samples.cuda(), gen_samples.cuda()

real_samples = F.one_hot(real_samples, cfg.vocab_size).float()

# ===Train===

d_out_real = self.dis(real_samples)

d_out_fake = self.dis(gen_samples)

g_loss, _ = get_losses(d_out_real, d_out_fake, cfg.loss_type)

self.optimize(self.gen_adv_opt, g_loss, self.gen)

total_loss += g_loss.item()

return total_loss / g_step if g_step != 0 else 0

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:20，

示例9: adv_train_generator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def adv_train_generator(self, g_step):

"""

The gen is trained using policy gradients, using the reward from the discriminator.

Training is done for num_batches batches.

"""

rollout_func = rollout.ROLLOUT(self.gen, cfg.CUDA)

total_g_loss = 0

for step in range(g_step):

inp, target = GenDataIter.prepare(self.gen.sample(cfg.batch_size, cfg.batch_size), gpu=cfg.CUDA)

# ===Train===

rewards = rollout_func.get_reward(target, cfg.rollout_num, self.dis)

adv_loss = self.gen.batchPGLoss(inp, target, rewards)

self.optimize(self.gen_adv_opt, adv_loss)

total_g_loss += adv_loss.item()

# ===Test===

self.log.info('[ADV-GEN]: g_loss = %.4f, %s' % (total_g_loss, self.cal_metrics(fmt_str=True)))

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:20，

示例10: train_discriminator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def train_discriminator(self, d_step, d_epoch, phase='MLE'):

"""

Training the discriminator on real_data_samples (positive) and generated samples from gen (negative).

Samples are drawn d_step times, and the discriminator is trained for d_epoch d_epoch.

"""

# prepare loader for validate

global d_loss, train_acc

for step in range(d_step):

# prepare loader for training

pos_samples = self.train_data.target

neg_samples = self.gen.sample(cfg.samples_num, 4 * cfg.batch_size)

dis_data = DisDataIter(pos_samples, neg_samples)

for epoch in range(d_epoch):

# ===Train===

d_loss, train_acc = self.train_dis_epoch(self.dis, dis_data.loader, self.dis_criterion,

self.dis_opt)

# ===Test===

self.log.info('[%s-DIS] d_step %d: d_loss = %.4f, train_acc = %.4f,' % (

phase, step, d_loss, train_acc))

if cfg.if_save and not cfg.if_test:

torch.save(self.dis.state_dict(), cfg.pretrained_dis_path)

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:26，

示例11: train_discriminator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def train_discriminator(self, d_step, d_epoch, phase='MLE'):

"""

Training the discriminator on real_data_samples (positive) and generated samples from gen (negative).

Samples are drawn d_step times, and the discriminator is trained for d_epoch d_epoch.

"""

d_loss, train_acc = 0, 0

for step in range(d_step):

# prepare loader for training

pos_samples = self.train_data.target

neg_samples = self.gen.sample(cfg.samples_num, cfg.batch_size, self.dis)

dis_data = DisDataIter(pos_samples, neg_samples)

for epoch in range(d_epoch):

# ===Train===

d_loss, train_acc = self.train_dis_epoch(self.dis, dis_data.loader, self.dis_criterion,

self.dis_opt)

# ===Test===

self.log.info('[%s-DIS] d_step %d: d_loss = %.4f, train_acc = %.4f,' % (

phase, step, d_loss, train_acc))

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:22，

示例12: cal_metrics

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def cal_metrics(self, fmt_str=False):

with torch.no_grad():

# Prepare data for evaluation

eval_samples = self.gen.sample(cfg.samples_num, cfg.batch_size, self.dis)

gen_data = GenDataIter(eval_samples)

gen_tokens = tensor_to_tokens(eval_samples, self.idx2word_dict)

gen_tokens_s = tensor_to_tokens(self.gen.sample(200, cfg.batch_size, self.dis), self.idx2word_dict)

# Reset metrics

self.bleu.reset(test_text=gen_tokens, real_text=self.test_data.tokens)

self.nll_gen.reset(self.gen, self.train_data.loader, leak_dis=self.dis)

self.nll_div.reset(self.gen, gen_data.loader, leak_dis=self.dis)

self.self_bleu.reset(test_text=gen_tokens_s, real_text=gen_tokens)

self.ppl.reset(gen_tokens)

if fmt_str:

return ', '.join(['%s = %s' % (metric.get_name(), metric.get_score()) for metric in self.all_metrics])

else:

return [metric.get_score() for metric in self.all_metrics]

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:21，

示例13: adv_train_generator

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def adv_train_generator(self, g_step):

"""

The gen is trained using policy gradients, using the reward from the discriminator.

Training is done for num_batches batches.

"""

for i in range(cfg.k_label):

rollout_func = rollout.ROLLOUT(self.gen_list[i], cfg.CUDA)

total_g_loss = 0

for step in range(g_step):

inp, target = GenDataIter.prepare(self.gen_list[i].sample(cfg.batch_size, cfg.batch_size), gpu=cfg.CUDA)

# ===Train===

rewards = rollout_func.get_reward(target, cfg.rollout_num, self.dis, current_k=i)

adv_loss = self.gen_list[i].batchPGLoss(inp, target, rewards)

self.optimize(self.gen_opt_list[i], adv_loss)

total_g_loss += adv_loss.item()

# ===Test===

self.log.info('[ADV-GEN]: %s', self.comb_metrics(fmt_str=True))

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:21，

示例14: cal_metrics_with_label

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def cal_metrics_with_label(self, label_i):

assert type(label_i) == int, 'missing label'

with torch.no_grad():

# Prepare data for evaluation

eval_samples = self.gen_list[label_i].sample(cfg.samples_num, 8 * cfg.batch_size)

gen_data = GenDataIter(eval_samples)

gen_tokens = tensor_to_tokens(eval_samples, self.idx2word_dict)

gen_tokens_s = tensor_to_tokens(self.gen_list[label_i].sample(200, 200), self.idx2word_dict)

clas_data = CatClasDataIter([eval_samples], label_i)

# Reset metrics

self.bleu.reset(test_text=gen_tokens, real_text=self.test_data_list[label_i].tokens)

self.nll_gen.reset(self.gen_list[label_i], self.train_data_list[label_i].loader)

self.nll_div.reset(self.gen_list[label_i], gen_data.loader)

self.self_bleu.reset(test_text=gen_tokens_s, real_text=gen_tokens)

self.clas_acc.reset(self.clas, clas_data.loader)

self.ppl.reset(gen_tokens)

return [metric.get_score() for metric in self.all_metrics]

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:22，

示例15: cal_metrics_with_label

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def cal_metrics_with_label(self, label_i):

assert type(label_i) == int, 'missing label'

with torch.no_grad():

# Prepare data for evaluation

eval_samples = self.gen.sample(cfg.samples_num, 8 * cfg.batch_size, label_i=label_i)

gen_data = GenDataIter(eval_samples)

gen_tokens = tensor_to_tokens(eval_samples, self.idx2word_dict)

gen_tokens_s = tensor_to_tokens(self.gen.sample(200, 200, label_i=label_i), self.idx2word_dict)

clas_data = CatClasDataIter([eval_samples], label_i)

# Reset metrics

self.bleu.reset(test_text=gen_tokens, real_text=self.test_data_list[label_i].tokens)

self.nll_gen.reset(self.gen, self.train_data_list[label_i].loader, label_i)

self.nll_div.reset(self.gen, gen_data.loader, label_i)

self.self_bleu.reset(test_text=gen_tokens_s, real_text=gen_tokens)

self.clas_acc.reset(self.clas, clas_data.loader)

self.ppl.reset(gen_tokens)

return [metric.get_score() for metric in self.all_metrics]

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:22，

示例16: create_oracle

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def create_oracle():

"""Create a new Oracle model and Oracle's samples"""

import config as cfg

from models.Oracle import Oracle

print('Creating Oracle...')

oracle = Oracle(cfg.gen_embed_dim, cfg.gen_hidden_dim, cfg.vocab_size,

cfg.max_seq_len, cfg.padding_idx, gpu=cfg.CUDA)

if cfg.CUDA:

oracle = oracle.cuda()

torch.save(oracle.state_dict(), cfg.oracle_state_dict_path)

big_samples = oracle.sample(cfg.samples_num, 4 * cfg.batch_size)

# large

torch.save(big_samples, cfg.oracle_samples_path.format(cfg.samples_num))

# small

torch.save(oracle.sample(cfg.samples_num // 2, 4 * cfg.batch_size),

cfg.oracle_samples_path.format(cfg.samples_num // 2))

oracle_data = GenDataIter(big_samples)

mle_criterion = nn.NLLLoss()

groud_truth = NLL.cal_nll(oracle, oracle_data.loader, mle_criterion)

print('NLL_Oracle Groud Truth: %.4f' % groud_truth)

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:26，

示例17: pretrain_loss

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def pretrain_loss(self, target, dis, start_letter=cfg.start_letter):

"""

Returns the pretrain_generator Loss for predicting target sequence.

Inputs: target, dis, start_letter

- target: batch_size * seq_len

"""

batch_size, seq_len = target.size()

_, feature_array, goal_array, leak_out_array = self.forward_leakgan(target, dis, if_sample=False, no_log=False,

start_letter=start_letter)

# Manager loss

mana_cos_loss = self.manager_cos_loss(batch_size, feature_array,

goal_array) # batch_size * (seq_len / step_size)

manager_loss = -torch.sum(mana_cos_loss) / (batch_size * (seq_len // self.step_size))

# Worker loss

work_nll_loss = self.worker_nll_loss(target, leak_out_array) # batch_size * seq_len

work_loss = torch.sum(work_nll_loss) / (batch_size * seq_len)

return manager_loss, work_loss

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:24，

示例18: adversarial_loss

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def adversarial_loss(self, target, rewards, dis, start_letter=cfg.start_letter):

"""

Returns a pseudo-loss that gives corresponding policy gradients (on calling .backward()).

Inspired by the example in http://karpathy.github.io/2016/05/31/rl/

Inputs: target, rewards, dis, start_letter

- target: batch_size * seq_len

- rewards: batch_size * seq_len (discriminator rewards for each token)

"""

batch_size, seq_len = target.size()

_, feature_array, goal_array, leak_out_array = self.forward_leakgan(target, dis, if_sample=False, no_log=False,

start_letter=start_letter, train=True)

# Manager Loss

t0 = time.time()

mana_cos_loss = self.manager_cos_loss(batch_size, feature_array,

goal_array) # batch_size * (seq_len / step_size)

mana_loss = -torch.sum(rewards * mana_cos_loss) / (batch_size * (seq_len // self.step_size))

# Worker Loss

work_nll_loss = self.worker_nll_loss(target, leak_out_array) # batch_size * seq_len

work_cos_reward = self.worker_cos_reward(feature_array, goal_array) # batch_size * seq_len

work_loss = -torch.sum(work_nll_loss * work_cos_reward) / (batch_size * seq_len)

return mana_loss, work_loss

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:27，

示例19: worker_cos_reward

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def worker_cos_reward(self, feature_array, goal_array):

"""

Get reward for worker (cosine distance)

:return: cos_loss: batch_size * seq_len

"""

for i in range(int(self.max_seq_len / self.step_size)):

real_feature = feature_array[:, i * self.step_size, :].unsqueeze(1).expand((-1, self.step_size, -1))

feature_array[:, i * self.step_size:(i + 1) * self.step_size, :] = real_feature

if i > 0:

sum_goal = torch.sum(goal_array[:, (i - 1) * self.step_size:i * self.step_size, :], dim=1, keepdim=True)

else:

sum_goal = goal_array[:, 0, :].unsqueeze(1)

goal_array[:, i * self.step_size:(i + 1) * self.step_size, :] = sum_goal.expand((-1, self.step_size, -1))

offset_feature = feature_array[:, 1:, :] # f_{t+1}, batch_size * seq_len * goal_out_size

goal_array = goal_array[:, :self.max_seq_len, :] # batch_size * seq_len * goal_out_size

sub_feature = offset_feature - goal_array

# L2 normalization

sub_feature = F.normalize(sub_feature, p=2, dim=-1)

all_goal = F.normalize(goal_array, p=2, dim=-1)

cos_loss = F.cosine_similarity(sub_feature, all_goal, dim=-1) # batch_size * seq_len

return cos_loss

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:27，

示例20: step

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def step(self, inp, hidden):

"""

RelGAN step forward

:param inp: [batch_size]

:param hidden: memory size

:return: pred, hidden, next_token, next_token_onehot, next_o

- pred: batch_size * vocab_size, use for adversarial training backward

- hidden: next hidden

- next_token: [batch_size], next sentence token

- next_token_onehot: batch_size * vocab_size, not used yet

- next_o: batch_size * vocab_size, not used yet

"""

emb = self.embeddings(inp).unsqueeze(1)

out, hidden = self.lstm(emb, hidden)

gumbel_t = self.add_gumbel(self.lstm2out(out.squeeze(1)))

next_token = torch.argmax(gumbel_t, dim=1).detach()

# next_token_onehot = F.one_hot(next_token, cfg.vocab_size).float() # not used yet

next_token_onehot = None

pred = F.softmax(gumbel_t * self.temperature, dim=-1) # batch_size * vocab_size

# next_o = torch.sum(next_token_onehot * pred, dim=1) # not used yet

next_o = None

return pred, hidden, next_token, next_token_onehot, next_o

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:26，

示例21: sample

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def sample(self, num_samples, batch_size, start_letter=cfg.start_letter):

"""

Samples the network and returns num_samples samples of length max_seq_len.

:return samples: num_samples * max_seq_length (a sampled sequence in each row)

"""

num_batch = num_samples // batch_size + 1 if num_samples != batch_size else 1

samples = torch.zeros(num_batch * batch_size, self.max_seq_len).long()

# Generate sentences with multinomial sampling strategy

for b in range(num_batch):

hidden = self.init_hidden(batch_size)

inp = torch.LongTensor([start_letter] * batch_size)

if self.gpu:

inp = inp.cuda()

for i in range(self.max_seq_len):

out, hidden = self.forward(inp, hidden, need_hidden=True) # out: batch_size * vocab_size

next_token = torch.multinomial(torch.exp(out), 1) # batch_size * 1 (sampling from each row)

samples[b * batch_size:(b + 1) * batch_size, i] = next_token.view(-1)

inp = next_token.view(-1)

samples = samples[:num_samples]

return samples

开发者ID:williamSYSU，项目名称:TextGAN-PyTorch，代码行数:25，

示例22: __init__

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def __init__(self,features,

augmentations,

batch_size=batch_size,

input1=(image_size_h_p,image_size_w_p,nchannels),

input2=(image_size_h_c,image_size_w_c,nchannels),

type1=None,

metadata_dict=None,

metadata_length=0,

with_paths=False):

self.features = features

self.batch_size = batch_size

self.vec_size = input1

self.vec_size2 = input2

self.type = type1

self.metadata_dict = metadata_dict

self.metadata_length = metadata_length

self.augment = augmentations

self.with_paths = with_paths

开发者ID:icarofua，项目名称:vehicle-ReId，代码行数:20，

示例23: forward

​点赞 6

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def forward(self, state, tau, num_quantiles):

input_size = state.size()[0] # batch_size(train) or 1(get_action)

tau = tau.expand(input_size * num_quantiles, quantile_embedding_dim)

pi_mtx = torch.Tensor(np.pi * np.arange(0, quantile_embedding_dim)).expand(input_size * num_quantiles, quantile_embedding_dim)

cos_tau = torch.cos(tau * pi_mtx)

phi = self.phi(cos_tau)

phi = F.relu(phi)

state_tile = state.expand(input_size, num_quantiles, self.num_inputs)

state_tile = state_tile.flatten().view(-1, self.num_inputs)

x = F.relu(self.fc1(state_tile))

x = self.fc2(x * phi)

z = x.view(-1, num_quantiles, self.num_outputs)

z = z.transpose(1, 2) # [input_size, num_output, num_quantile]

return z

开发者ID:g6ling，项目名称:Reinforcement-Learning-Pytorch-Cartpole，代码行数:20，

示例24: config_initialization

​点赞 5

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def config_initialization():

# image shape and feature layers shape inference

image_shape = (FLAGS.train_image_height, FLAGS.train_image_width)

if not FLAGS.dataset_dir:

raise ValueError('You must supply the dataset directory with --dataset_dir')

tf.logging.set_verbosity(tf.logging.DEBUG)

util.init_logger(log_file = 'log_train_seglink_%d_%d.log'%image_shape, log_path = FLAGS.train_dir, stdout = False, mode = 'a')

config.init_config(image_shape,

batch_size = FLAGS.batch_size,

weight_decay = FLAGS.weight_decay,

num_gpus = FLAGS.num_gpus,

train_with_ignored = FLAGS.train_with_ignored,

seg_loc_loss_weight = FLAGS.seg_loc_loss_weight,

link_cls_loss_weight = FLAGS.link_cls_loss_weight,

)

batch_size = config.batch_size

batch_size_per_gpu = config.batch_size_per_gpu

tf.summary.scalar('batch_size', batch_size)

tf.summary.scalar('batch_size_per_gpu', batch_size_per_gpu)

util.proc.set_proc_name(FLAGS.model_name + '_' + FLAGS.dataset_name)

dataset = dataset_factory.get_dataset(FLAGS.dataset_name, FLAGS.dataset_split_name, FLAGS.dataset_dir)

config.print_config(FLAGS, dataset)

return dataset

开发者ID:dengdan，项目名称:seglink，代码行数:32，

示例25: config_initialization

​点赞 5

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def config_initialization():

if not FLAGS.dataset_dir:

raise ValueError('You must supply the dataset directory with --dataset_dir')

tf.logging.set_verbosity(tf.logging.DEBUG)

# image shape and feature layers shape inference

image_shape = (FLAGS.train_image_height, FLAGS.train_image_width)

config.init_config(image_shape, batch_size = FLAGS.batch_size)

util.proc.set_proc_name(FLAGS.model_name + '_' + FLAGS.dataset_name)

dataset = dataset_factory.get_dataset(FLAGS.dataset_name, FLAGS.dataset_split_name, FLAGS.dataset_dir)

# config.print_config(FLAGS, dataset)

return dataset

开发者ID:dengdan，项目名称:seglink，代码行数:17，

示例26: main

​点赞 5

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def main(_):

util.init_logger()

dump_path = util.io.get_absolute_path('~/temp/no-use/seglink/')

dataset = config_initialization()

batch_queue = create_dataset_batch_queue(dataset)

batch_size = config.batch_size

summary_op = tf.summary.merge_all()

with tf.Session() as sess:

tf.train.start_queue_runners(sess)

b_image, b_seg_label, b_seg_offsets, b_link_label = batch_queue.dequeue()

batch_idx = 0;

while True: #batch_idx < 50:

image_data_batch, seg_label_data_batch, seg_offsets_data_batch, link_label_data_batch = \

sess.run([b_image, b_seg_label, b_seg_offsets, b_link_label])

for image_idx in xrange(batch_size):

image_data = image_data_batch[image_idx, ...]

seg_label_data = seg_label_data_batch[image_idx, ...]

seg_offsets_data = seg_offsets_data_batch[image_idx, ...]

link_label_data = link_label_data_batch[image_idx, ...]

image_data = image_data + [123, 117, 104]

image_data = np.asarray(image_data, dtype = np.uint8)

# decode the encoded ground truth back to bboxes

bboxes = seglink.seglink_to_bbox(seg_scores = seg_label_data,

link_scores = link_label_data,

seg_offsets_pred = seg_offsets_data)

# draw bboxes on the image

for bbox_idx in xrange(len(bboxes)):

bbox = bboxes[bbox_idx, :]

draw_bbox(image_data, bbox)

image_path = util.io.join_path(dump_path, '%d_%d.jpg'%(batch_idx, image_idx))

util.plt.imwrite(image_path, image_data)

print 'Make sure that the text on the image are correctly bounded\

with oriented boxes:', image_path

batch_idx += 1

开发者ID:dengdan，项目名称:seglink，代码行数:41，

示例27: __len__

​点赞 5

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def __len__(self):

return int(np.ceil(len(self.samples) / float(batch_size)))

开发者ID:foamliu，项目名称:FaceNet，代码行数:4，

示例28: __getitem__

​点赞 5

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def __getitem__(self, idx):

i = idx * batch_size

length = min(batch_size, (len(self.samples) - i))

batch_inputs = np.empty((3, length, img_size, img_size, channel), dtype=np.float32)

batch_dummy_target = np.zeros((length, embedding_size * 3), dtype=np.float32)

for i_batch in range(length):

sample = self.samples[i + i_batch]

for j, role in enumerate(['a', 'p', 'n']):

image_name = sample[role]

filename = os.path.join(self.image_folder, image_name)

image = cv.imread(filename) # BGR

image = image[:, :, ::-1] # RGB

dets = self.detector(image, 1)

num_faces = len(dets)

if num_faces > 0:

# Find the 5 face landmarks we need to do the alignment.

faces = dlib.full_object_detections()

for detection in dets:

faces.append(self.sp(image, detection))

image = dlib.get_face_chip(image, faces[0], size=img_size)

else:

image = cv.resize(image, (img_size, img_size), cv.INTER_CUBIC)

if self.usage == 'train':

image = aug_pipe.augment_image(image)

batch_inputs[j, i_batch] = preprocess_input(image)

return [batch_inputs[0], batch_inputs[1], batch_inputs[2]], batch_dummy_target

开发者ID:foamliu，项目名称:FaceNet，代码行数:34，

示例29: parse_args

​点赞 5

​

# 需要导入模块: import config [as 别名]

# 或者: from config import batch_size [as 别名]

def parse_args():

# Parse input arguments

desc = "Implementation for AugMix paper"

parser = argparse.ArgumentParser(description=desc)

parser.add_argument('--batch_size',

type=int,

required=False)

parser.add_argument('--epochs',

help='number of training epochs',

type=int,

required=True)

parser.add_argument('--max_lr',

help='maxium learning rate for lr scheduler',

default=1.0,

type=float)

parser.add_argument('--min_lr',

help='minimum learning rate for lr scheduler',

default=1e-5,

type=float)

parser.add_argument('--img_size',

help='size of the images',

default=32,

type=int)

parser.add_argument("--save_dir_path",

type=str,

help="dir path to save output results",

default="",

required=False)

parser.add_argument("--plot_name",

type=str,

help="filename for the plots",

default="history.png",

required=False)

args = vars(parser.parse_args())

return args

###########################################################################

开发者ID:AakashKumarNain，项目名称:AugMix_TF2，代码行数:39，

注：本文中的config.batch_size方法示例整理自Github/MSDocs等源码及文档管理平台，相关代码片段筛选自各路编程大神贡献的开源项目，源码版权归原作者所有，传播和使用请参考对应项目的License；未经允许，请勿转载。

python batch_size_Python config.batch_size方法代码示例相关推荐

  1. python里config_Python config.get_config方法代码示例

    本文整理汇总了Python中config.get_config方法的典型用法代码示例.如果您正苦于以下问题:Python config.get_config方法的具体用法?Python config. ...

  2. python dropout_Python slim.dropout方法代码示例

    本文整理汇总了Python中tensorflow.contrib.slim.dropout方法的典型用法代码示例.如果您正苦于以下问题:Python slim.dropout方法的具体用法?Pytho ...

  3. python data_Python data.Data方法代码示例

    本文整理汇总了Python中data.Data方法的典型用法代码示例.如果您正苦于以下问题:Python data.Data方法的具体用法?Python data.Data怎么用?Python dat ...

  4. python makedirs_Python gfile.MakeDirs方法代码示例

    本文整理汇总了Python中tensorflow.gfile.MakeDirs方法的典型用法代码示例.如果您正苦于以下问题:Python gfile.MakeDirs方法的具体用法?Python gf ...

  5. python fonttool_Python wx.Font方法代码示例

    本文整理汇总了Python中wx.Font方法的典型用法代码示例.如果您正苦于以下问题:Python wx.Font方法的具体用法?Python wx.Font怎么用?Python wx.Font使用 ...

  6. python nextpow2_Python signal.hann方法代码示例

    本文整理汇总了Python中scipy.signal.hann方法的典型用法代码示例.如果您正苦于以下问题:Python signal.hann方法的具体用法?Python signal.hann怎么 ...

  7. python renamer_Python sys.meta_path方法代码示例

    本文整理汇总了Python中sys.meta_path方法的典型用法代码示例.如果您正苦于以下问题:Python sys.meta_path方法的具体用法?Python sys.meta_path怎么 ...

  8. python info_Python logger.info方法代码示例

    本文整理汇总了Python中utils.logger.info方法的典型用法代码示例.如果您正苦于以下问题:Python logger.info方法的具体用法?Python logger.info怎么 ...

  9. python dateformatter_Python dates.DateFormatter方法代码示例

    本文整理汇总了Python中matplotlib.dates.DateFormatter方法的典型用法代码示例.如果您正苦于以下问题:Python dates.DateFormatter方法的具体用法 ...

最新文章

  1. 某程序员吐槽：提离职后领导开始演戏，假装不知道我工资低，对我进行挽留，怎么办？...
  2. ceph的数据存储之路(6) -----pg的创建
  3. 安卓流行布局开源库_如何使用流行度在开源库之间进行选择
  4. [html] img中的src加载失败时如何用默认图片来替换呢？
  5. 前端学习（235）:css HACK
  6. 解决微信小程序新建项目没有样式问题，以及官方demo
  7. python hackrf_HackRF固件更新及编译环境搭建
  8. python基础语法实验要求_Python基础语法-关于条件
  9. XSS漏洞原理/方式/防御
  10. sdh管理单元指针_「干货三」SDH技术重点知识分布（附小技巧）
  11. CS229 Lecture Note 1(监督学习、线性回归)
  12. Idea翻译插件google翻译失败超时
  13. 深度学习入门：基于Python的理论与实现——第一章Python入门
  14. 桥本木分式（使用回溯法求解）
  15. C语言公交车线路信息查询系统
  16. java算法int型整数反转的另类解法
  17. 渗透前期学习资源分享
  18. 西瓜决策树-sklearn实现
  19. java模拟抛硬币的结果
  20. 【CSS3】设置背景样式

热门文章

  1. 关于session校验在项目中的使用
  2. 随机森林分类算法python代码_随机森林的原理及Python代码实现
  3. Win11高效日历推荐
  4. Docker容器commit安装kali工具集
  5. android之weex之component插件开发
  6. Springboot之GetMapping参数
  7. VALSE学习（十一）：弱监督图像语义分割
  8. centos 环境变量_Centos7：Linux环境变量配置文件
  9. opencv安装教程python3.7_Python3.7中安装openCV库的方法
  10. js 判断支持webgl_「WebGL基础」：第一部分