07 label_semantic_roles #548
Conversation
…-level-api-branch
…gh-level-api-branch
…gh-level-api-branch
…gh-level-api-branch
…gh-level-api-branch
07.label_semantic_roles/README.cn.md
Outdated
|
|
||
| # 标注序列 | ||
| target = paddle.layer.data(name='target', type=d_type(label_dict_len)) | ||
| mark_dict_len = 2 |
There was a problem hiding this comment.
Can we keep the comments from the original versions?
-mark_dict_len = 2 # 谓上下文区域标志的维度，是一个0-1 2值特征，因此维度为2
-word_dim = 32 # 词向量维度
-mark_dim = 5 # 谓词上下文区域通过词表被映射为一个实向量，这个是相邻的维度
-hidden_dim = 512 # LSTM隐层向量的维度 ： 512 / 4
-depth = 8 # 栈式LSTM的深度
07.label_semantic_roles/train.py
Outdated
| def d_type(size): | ||
| return paddle.data_type.integer_value_sequence(size) | ||
| embedding_name = 'emb' | ||
| default_std = 1 / math.sqrt(hidden_dim) / 3.0 # ? |
07.label_semantic_roles/train.py
Outdated
|
|
||
| predicate_embedding = paddle.layer.embedding( | ||
| size=word_dim, | ||
| # def db_lstm(): |
There was a problem hiding this comment.
Why do we have two versions of db_lstm? Are we intend to show the v2 version as well?
There was a problem hiding this comment.
Removed the commented out version
| target = paddle.layer.data(name='target', type=d_type(label_dict_len)) | ||
| crf_cost = paddle.layer.crf( | ||
| size=label_dict_len, | ||
| feature_out = db_lstm(**locals()) |
There was a problem hiding this comment.
What is **locals() ? I don't see this declared elsewhere.
There was a problem hiding this comment.
def just_print(haha, xixi, **whatever):
print 'haha is :', haha
print 'xixi is :', xixi
def demo_it():
haha = 'I am just laughing'
xixi = 'I am just grinning'
just_print(**locals())
demo_it() Output is
$ python show_dict.py
haha is : I am just laughing
xixi is : I am just grinningThere was a problem hiding this comment.
So when we call it, we pass in locals() which is the local variables in the current scope. ignored can be any word.
07.label_semantic_roles/README.cn.md
Outdated
|
|
||
| # 标注序列 | ||
| target = paddle.layer.data(name='target', type=d_type(label_dict_len)) | ||
| mark_dict_len = 2 |
There was a problem hiding this comment.
Can we keep the original comments?
-mark_dict_len = 2 # 谓上下文区域标志的维度，是一个0-1 2值特征，因此维度为2
-word_dim = 32 # 词向量维度
-mark_dim = 5 # 谓词上下文区域通过词表被映射为一个实向量，这个是相邻的维度
-hidden_dim = 512 # LSTM隐层向量的维度 ： 512 / 4
-depth = 8 # 栈式LSTM的深度
07.label_semantic_roles/README.cn.md
Outdated
| for i in range(1, depth): | ||
| mix_hidden = paddle.layer.mixed( | ||
| def db_lstm(word, predicate, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2, mark, | ||
| **ignored): |
There was a problem hiding this comment.
What does the **ignored do? It is not used anywhere.
07.label_semantic_roles/README.cn.md
Outdated
| # Set the threshold low to speed up the CI test | ||
| if float(cost) < 60.0: | ||
| if save_dirname is not None: | ||
| # TODO(liuyiqun): Change the target to crf_decode |
There was a problem hiding this comment.
remove TODO in readme? train.py we can keep it
Eng & Chi README, train.py all done.