Merge pull request #1784 from luotao1/beam

add seqtext_print for seqToseq demo

Author: luotao1

demo/seqToseq/api_train_v2.py

@@ -126,7 +126,7 @@ def gru_decoder_with_attention(enc_vec, enc_proj, current_word):
 
 def main():
     paddle.init(use_gpu=False, trainer_count=1)
-    is_generating = True
+    is_generating = False
 
     # source and target dict dim.
     dict_size = 30000
@@ -167,16 +167,47 @@ def event_handler(event):
 
     # generate a english sequence to french
     else:
-        gen_creator = paddle.dataset.wmt14.test(dict_size)
+        # use the first 3 samples for generation
+        gen_creator = paddle.dataset.wmt14.gen(dict_size)
         gen_data = []
+        gen_num = 3
         for item in gen_creator():
             gen_data.append((item[0], ))
-            if len(gen_data) == 3:
+            if len(gen_data) == gen_num:
                 break
 
         beam_gen = seqToseq_net(source_dict_dim, target_dict_dim, is_generating)
+        # get the pretrained model, whose bleu = 26.92
         parameters = paddle.dataset.wmt14.model()
-        trg_dict = paddle.dataset.wmt14.trg_dict(dict_size)
+        # prob is the prediction probabilities, and id is the prediction word. 
+        beam_result = paddle.infer(
+            output_layer=beam_gen,
+            parameters=parameters,
+            input=gen_data,
+            field=['prob', 'id'])
+
+        # get the dictionary
+        src_dict, trg_dict = paddle.dataset.wmt14.get_dict(dict_size)
+
+        # the delimited element of generated sequences is -1,
+        # the first element of each generated sequence is the sequence length
+        seq_list = []
+        seq = []
+        for w in beam_result[1]:
+            if w != -1:
+                seq.append(w)
+            else:
+                seq_list.append(' '.join([trg_dict.get(w) for w in seq[1:]]))
+                seq = []
+
+        prob = beam_result[0]
+        beam_size = 3
+        for i in xrange(gen_num):
+            print "\n*******************************************************\n"
+            print "src:", ' '.join(
+                [src_dict.get(w) for w in gen_data[i][0]]), "\n"
+            for j in xrange(beam_size):
+                print "prob = %f:" % (prob[i][j]), seq_list[i * beam_size + j]
 
 
 if __name__ == '__main__':

python/paddle/v2/dataset/wmt14.py

@@ -26,7 +26,7 @@
 MD5_DEV_TEST = '7d7897317ddd8ba0ae5c5fa7248d3ff5'
 # this is a small set of data for test. The original data is too large and will be add later.
 URL_TRAIN = 'http://paddlepaddle.cdn.bcebos.com/demo/wmt_shrinked_data/wmt14.tgz'
-MD5_TRAIN = 'a755315dd01c2c35bde29a744ede23a6'
+MD5_TRAIN = '0791583d57d5beb693b9414c5b36798c'
 # this is the pretrained model, whose bleu = 26.92
 URL_MODEL = 'http://paddlepaddle.bj.bcebos.com/demo/wmt_14/wmt14_model.tar.gz'
 MD5_MODEL = '4ce14a26607fb8a1cc23bcdedb1895e4'
@@ -108,17 +108,27 @@ def test(dict_size):
         download(URL_TRAIN, 'wmt14', MD5_TRAIN), 'test/test', dict_size)
 
 
+def gen(dict_size):
+    return reader_creator(
+        download(URL_TRAIN, 'wmt14', MD5_TRAIN), 'gen/gen', dict_size)
+
+
 def model():
     tar_file = download(URL_MODEL, 'wmt14', MD5_MODEL)
     with gzip.open(tar_file, 'r') as f:
         parameters = Parameters.from_tar(f)
     return parameters
 
 
-def trg_dict(dict_size):
+def get_dict(dict_size, reverse=True):
+    # if reverse = False, return dict = {'a':'001', 'b':'002', ...}
+    # else reverse = true, return dict = {'001':'a', '002':'b', ...}
     tar_file = download(URL_TRAIN, 'wmt14', MD5_TRAIN)
     src_dict, trg_dict = __read_to_dict__(tar_file, dict_size)
-    return trg_dict
+    if reverse:
+        src_dict = {v: k for k, v in src_dict.items()}
+        trg_dict = {v: k for k, v in trg_dict.items()}
+    return src_dict, trg_dict
 
 
 def fetch():