未验证 提交 bccff22a 编写于 作者: Webbley's avatar Webbley 提交者: GitHub

Merge pull request #5 from PaddlePaddle/main

merge
...@@ -20,7 +20,7 @@ evaluator = Evaluator(name='ogbn-arxiv') ...@@ -20,7 +20,7 @@ evaluator = Evaluator(name='ogbn-arxiv')
def get_config(): def get_config():
parser = argparse.ArgumentParser() parser = argparse.ArgumentParser()
## 基本模型参数 ## model_arg
model_group=parser.add_argument_group('model_base_arg') model_group=parser.add_argument_group('model_base_arg')
model_group.add_argument('--num_layers', default=3, type=int) model_group.add_argument('--num_layers', default=3, type=int)
model_group.add_argument('--hidden_size', default=128, type=int) model_group.add_argument('--hidden_size', default=128, type=int)
...@@ -28,7 +28,7 @@ def get_config(): ...@@ -28,7 +28,7 @@ def get_config():
model_group.add_argument('--dropout', default=0.3, type=float) model_group.add_argument('--dropout', default=0.3, type=float)
model_group.add_argument('--attn_dropout', default=0, type=float) model_group.add_argument('--attn_dropout', default=0, type=float)
## label embedding模型参数 ## label_embed_arg
embed_group=parser.add_argument_group('embed_arg') embed_group=parser.add_argument_group('embed_arg')
embed_group.add_argument('--use_label_e', action='store_true') embed_group.add_argument('--use_label_e', action='store_true')
embed_group.add_argument('--label_rate', default=0.625, type=float) embed_group.add_argument('--label_rate', default=0.625, type=float)
...@@ -81,17 +81,17 @@ def eval_test(parser, program, model, test_exe, graph, y_true, split_idx): ...@@ -81,17 +81,17 @@ def eval_test(parser, program, model, test_exe, graph, y_true, split_idx):
def train_loop(parser, start_program, main_program, test_program, def train_loop(parser, start_program, main_program, test_program,
model, graph, label, split_idx, exe, run_id, wf=None): model, graph, label, split_idx, exe, run_id, wf=None):
#启动上文构建的训练器 #build up training program
exe.run(start_program) exe.run(start_program)
max_acc=0 # 最佳test_acc max_acc=0 # best test_acc
max_step=0 # 最佳test_acc 对应step max_step=0 # step for best test_acc
max_val_acc=0 # 最佳val_acc max_val_acc=0 # best val_acc
max_cor_acc=0 # 最佳val_acc对应test_acc max_cor_acc=0 # test_acc for best val_acc
max_cor_step=0 # 最佳val_acc对应step max_cor_step=0 # step for best val_acc
#训练循环 #training loop
for epoch_id in tqdm(range(parser.epochs)): for epoch_id in tqdm(range(parser.epochs)):
#运行训练器 #start training
if parser.use_label_e: if parser.use_label_e:
feed_dict=model.gw.to_feed(graph) feed_dict=model.gw.to_feed(graph)
...@@ -115,7 +115,7 @@ def train_loop(parser, start_program, main_program, test_program, ...@@ -115,7 +115,7 @@ def train_loop(parser, start_program, main_program, test_program,
# print(loss[1][0]) # print(loss[1][0])
loss = loss[0] loss = loss[0]
#测试结果 #eval result
result = eval_test(parser, test_program, model, exe, graph, label, split_idx) result = eval_test(parser, test_program, model, exe, graph, label, split_idx)
train_acc, valid_acc, test_acc = result train_acc, valid_acc, test_acc = result
...@@ -191,11 +191,7 @@ if __name__ == '__main__': ...@@ -191,11 +191,7 @@ if __name__ == '__main__':
test_prog=train_prog.clone(for_test=True) test_prog=train_prog.clone(for_test=True)
model.train_program() model.train_program()
adam_optimizer = optimizer_func(parser.lr)#optimizer
# ave_loss = train_program(pred_output)#训练程序
# lr, global_step= linear_warmup_decay(parser.lr, parser.epochs*0.1, parser.epochs)
# adam_optimizer = optimizer_func(lr)#训练优化函数
adam_optimizer = optimizer_func(parser.lr)#训练优化函数
adam_optimizer.minimize(model.avg_cost) adam_optimizer.minimize(model.avg_cost)
exe = F.Executor(place) exe = F.Executor(place)
......
...@@ -22,14 +22,14 @@ evaluator = Evaluator(name='ogbn-products') ...@@ -22,14 +22,14 @@ evaluator = Evaluator(name='ogbn-products')
def get_config(): def get_config():
parser = argparse.ArgumentParser() parser = argparse.ArgumentParser()
## 采样参数 ## data_sampling_arg
data_group= parser.add_argument_group('data_arg') data_group= parser.add_argument_group('data_arg')
data_group.add_argument('--batch_size', default=1500, type=int) data_group.add_argument('--batch_size', default=1500, type=int)
data_group.add_argument('--num_workers', default=12, type=int) data_group.add_argument('--num_workers', default=12, type=int)
data_group.add_argument('--sizes', default=[10, 10, 10], type=int, nargs='+' ) data_group.add_argument('--sizes', default=[10, 10, 10], type=int, nargs='+' )
data_group.add_argument('--buf_size', default=1000, type=int) data_group.add_argument('--buf_size', default=1000, type=int)
## 基本模型参数 ## model_arg
model_group=parser.add_argument_group('model_base_arg') model_group=parser.add_argument_group('model_base_arg')
model_group.add_argument('--num_layers', default=3, type=int) model_group.add_argument('--num_layers', default=3, type=int)
model_group.add_argument('--hidden_size', default=128, type=int) model_group.add_argument('--hidden_size', default=128, type=int)
...@@ -37,7 +37,7 @@ def get_config(): ...@@ -37,7 +37,7 @@ def get_config():
model_group.add_argument('--dropout', default=0.3, type=float) model_group.add_argument('--dropout', default=0.3, type=float)
model_group.add_argument('--attn_dropout', default=0, type=float) model_group.add_argument('--attn_dropout', default=0, type=float)
## label embedding模型参数 ## label_embed_arg
embed_group=parser.add_argument_group('embed_arg') embed_group=parser.add_argument_group('embed_arg')
embed_group.add_argument('--use_label_e', action='store_true') embed_group.add_argument('--use_label_e', action='store_true')
embed_group.add_argument('--label_rate', default=0.625, type=float) embed_group.add_argument('--label_rate', default=0.625, type=float)
...@@ -113,19 +113,19 @@ def eval_test(parser, test_p_list, model, test_exe, dataset, split_idx): ...@@ -113,19 +113,19 @@ def eval_test(parser, test_p_list, model, test_exe, dataset, split_idx):
def train_loop(parser, start_program, main_program, test_p_list, def train_loop(parser, start_program, main_program, test_p_list,
model, feat_init, place, dataset, split_idx, exe, run_id, wf=None): model, feat_init, place, dataset, split_idx, exe, run_id, wf=None):
#启动上文构建的训练器 #build up training program
exe.run(start_program) exe.run(start_program)
feat_init(place) feat_init(place)
max_acc=0 # 最佳test_acc max_acc=0 # best test_acc
max_step=0 # 最佳test_acc 对应step max_step=0 # step for best test_acc
max_val_acc=0 # 最佳val_acc max_val_acc=0 # best val_acc
max_cor_acc=0 # 最佳val_acc对应test_acc max_cor_acc=0 # test_acc for best val_acc
max_cor_step=0 # 最佳val_acc对应step max_cor_step=0 # step for best val_acc
#训练循环 #training loop
for epoch_id in range(parser.epochs): for epoch_id in range(parser.epochs):
#运行训练器 #start training
if parser.use_label_e: if parser.use_label_e:
train_idx_temp=copy.deepcopy(split_idx['train']) train_idx_temp=copy.deepcopy(split_idx['train'])
...@@ -158,8 +158,7 @@ def train_loop(parser, start_program, main_program, test_p_list, ...@@ -158,8 +158,7 @@ def train_loop(parser, start_program, main_program, test_p_list,
print('acc: ', (acc_num/unlabel_idx.shape[0])*100) print('acc: ', (acc_num/unlabel_idx.shape[0])*100)
#测试结果 #eval result
# total=0.0
if (epoch_id+1)>=50 and (epoch_id+1)%10==0: if (epoch_id+1)>=50 and (epoch_id+1)%10==0:
result = eval_test(parser, test_p_list, model, exe, dataset, split_idx) result = eval_test(parser, test_p_list, model, exe, dataset, split_idx)
train_acc, valid_acc, test_acc = result train_acc, valid_acc, test_acc = result
...@@ -242,17 +241,14 @@ if __name__ == '__main__': ...@@ -242,17 +241,14 @@ if __name__ == '__main__':
# test_prog=train_prog.clone(for_test=True) # test_prog=train_prog.clone(for_test=True)
model.train_program() model.train_program()
# ave_loss = train_program(pred_output)#训练程序 adam_optimizer = optimizer_func(parser.lr)#optimizer
# lr, global_step= linear_warmup_decay(0.01, 50, 500)
# adam_optimizer = optimizer_func(lr)#训练优化函数
adam_optimizer = optimizer_func(parser.lr)#训练优化函数
adam_optimizer.minimize(model.avg_cost) adam_optimizer.minimize(model.avg_cost)
test_p_list=[] test_p_list=[]
with F.unique_name.guard(): with F.unique_name.guard():
## input层 ## build up eval program
test_p=F.Program() test_p=F.Program()
with F.program_guard(test_p, ): with F.program_guard(test_p, ):
gw_test=pgl.graph_wrapper.GraphWrapper( gw_test=pgl.graph_wrapper.GraphWrapper(
...@@ -281,7 +277,7 @@ if __name__ == '__main__': ...@@ -281,7 +277,7 @@ if __name__ == '__main__':
with F.program_guard(test_p, ): with F.program_guard(test_p, ):
gw_test=pgl.graph_wrapper.GraphWrapper( gw_test=pgl.graph_wrapper.GraphWrapper(
name="product_"+str(0)) name="product_"+str(0))
# feature_batch=model.get_batch_feature(label_feature, test=True) # 把图在CPU存起 # feature_batch=model.get_batch_feature(label_feature, test=True)
feature_batch = F.data( 'hidden_node_feat', feature_batch = F.data( 'hidden_node_feat',
shape=[None, model.num_heads*model.hidden_size], shape=[None, model.num_heads*model.hidden_size],
dtype='float32') dtype='float32')
......
...@@ -23,7 +23,7 @@ evaluator = Evaluator(name='ogbn-proteins') ...@@ -23,7 +23,7 @@ evaluator = Evaluator(name='ogbn-proteins')
def get_config(): def get_config():
parser = argparse.ArgumentParser() parser = argparse.ArgumentParser()
## 基本模型参数 ## model_arg
model_group=parser.add_argument_group('model_base_arg') model_group=parser.add_argument_group('model_base_arg')
model_group.add_argument('--num_layers', default=7, type=int) model_group.add_argument('--num_layers', default=7, type=int)
model_group.add_argument('--hidden_size', default=64, type=int) model_group.add_argument('--hidden_size', default=64, type=int)
...@@ -31,7 +31,7 @@ def get_config(): ...@@ -31,7 +31,7 @@ def get_config():
model_group.add_argument('--dropout', default=0.1, type=float) model_group.add_argument('--dropout', default=0.1, type=float)
model_group.add_argument('--attn_dropout', default=0, type=float) model_group.add_argument('--attn_dropout', default=0, type=float)
## label embedding模型参数 ## label_embed_arg
embed_group=parser.add_argument_group('embed_arg') embed_group=parser.add_argument_group('embed_arg')
embed_group.add_argument('--use_label_e', action='store_true') embed_group.add_argument('--use_label_e', action='store_true')
embed_group.add_argument('--label_rate', default=0.5, type=float) embed_group.add_argument('--label_rate', default=0.5, type=float)
...@@ -90,15 +90,16 @@ def eval_test(parser, program, model, test_exe, graph, y_true, split_idx): ...@@ -90,15 +90,16 @@ def eval_test(parser, program, model, test_exe, graph, y_true, split_idx):
def train_loop(parser, start_program, main_program, test_program, def train_loop(parser, start_program, main_program, test_program,
model, graph, label, split_idx, exe, run_id, wf=None): model, graph, label, split_idx, exe, run_id, wf=None):
#启动上文构建的训练器 #build up training program
exe.run(start_program) exe.run(start_program)
max_acc=0 # 最佳test_acc max_acc=0 # best test_acc
max_step=0 # 最佳test_acc 对应step max_step=0 # step for best test_acc
max_val_acc=0 # 最佳val_acc max_val_acc=0 # best val_acc
max_cor_acc=0 # 最佳val_acc对应test_acc max_cor_acc=0 # test_acc for best val_acc
max_cor_step=0 # 最佳val_acc对应step max_cor_step=0 # step for best val_acc
#训练循环 #training loop
graph.node_feat["label"] = label graph.node_feat["label"] = label
graph.node_feat["nid"] = np.arange(0, graph.num_nodes) graph.node_feat["nid"] = np.arange(0, graph.num_nodes)
...@@ -112,7 +113,7 @@ def train_loop(parser, start_program, main_program, test_program, ...@@ -112,7 +113,7 @@ def train_loop(parser, start_program, main_program, test_program,
for epoch_id in tqdm(range(parser.epochs)): for epoch_id in tqdm(range(parser.epochs)):
for subgraph in random_partition(num_clusters=9, graph=graph, shuffle=True): for subgraph in random_partition(num_clusters=9, graph=graph, shuffle=True):
#运行训练器 #start training
if parser.use_label_e: if parser.use_label_e:
feed_dict = model.gw.to_feed(subgraph) feed_dict = model.gw.to_feed(subgraph)
sub_idx = set(subgraph.node_feat["nid"]) sub_idx = set(subgraph.node_feat["nid"])
...@@ -139,7 +140,7 @@ def train_loop(parser, start_program, main_program, test_program, ...@@ -139,7 +140,7 @@ def train_loop(parser, start_program, main_program, test_program,
fetch_list=[model.avg_cost]) fetch_list=[model.avg_cost])
loss = loss[0] loss = loss[0]
#测试结果 #eval result
if (epoch_id+1) > parser.epochs*0.9: if (epoch_id+1) > parser.epochs*0.9:
result = eval_test(parser, test_program, model, exe, graph, label, split_idx) result = eval_test(parser, test_program, model, exe, graph, label, split_idx)
train_acc, valid_acc, test_acc = result train_acc, valid_acc, test_acc = result
...@@ -221,7 +222,7 @@ if __name__ == '__main__': ...@@ -221,7 +222,7 @@ if __name__ == '__main__':
model.train_program() model.train_program()
adam_optimizer = optimizer_func(parser.lr)#训练优化函数 adam_optimizer = optimizer_func(parser.lr)#optimizer
adam_optimizer.minimize(model.avg_cost) adam_optimizer.minimize(model.avg_cost)
exe = F.Executor(place) exe = F.Executor(place)
......
...@@ -27,24 +27,34 @@ import time ...@@ -27,24 +27,34 @@ import time
import paddle.fluid as fluid import paddle.fluid as fluid
from multiprocessing import Queue from multiprocessing import Queue
import threading import threading
from collections import namedtuple
_np_serialized_data = namedtuple("_np_serialized_data", ["value", "shape", "dtype"])
def serialize_data(data): def serialize_data(data):
"""serialize_data""" """serialize_data"""
if data is None: if data is None:
return None return None
return numpy_serialize_data(data) #, ensure_ascii=False) return numpy_serialize_data(data) #, ensure_ascii=False)
def index_iter(data):
"""return indexing iter"""
if isinstance(data, list):
return range(len(data))
elif isinstance(data, dict):
return data.keys()
def numpy_serialize_data(data): def numpy_serialize_data(data):
"""serialize_data""" """serialize_data"""
ret_data = {} ret_data = copy.deepcopy(data)
for key in data:
if isinstance(data[key], np.ndarray): if isinstance(ret_data, (dict, list)):
ret_data[key] = (data[key].tobytes(), list(data[key].shape), for key in index_iter(ret_data):
"%s" % data[key].dtype) if isinstance(ret_data[key], np.ndarray):
else: ret_data[key] = _np_serialized_data(value=ret_data[key].tobytes(),
ret_data[key] = data[key] shape=list(ret_data[key].shape), dtype="%s" % ret_data[key].dtype)
return ret_data return ret_data
...@@ -52,11 +62,12 @@ def numpy_deserialize_data(data): ...@@ -52,11 +62,12 @@ def numpy_deserialize_data(data):
"""deserialize_data""" """deserialize_data"""
if data is None: if data is None:
return None return None
for key in data:
if isinstance(data[key], tuple): if isinstance(data, (dict, list)):
value = np.frombuffer( for key in index_iter(data):
data[key][0], dtype=data[key][2]).reshape(data[key][1]) if isinstance(data[key], _np_serialized_data):
data[key] = value data[key] = np.frombuffer(buffer=data[key].value,
dtype=data[key].dtype).reshape(data[key].shape)
return data return data
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册