fix: pass num_context_turns further in dam constructor (#1029)

* fix: num_context_turns in dam * fix: num_context_turns parameter passing to the parent class, multiple num_context_turns declarations in configs * fix: num_context_turns parameter passing to the parent class, multiple num_context_turns declarations in configs
deeppavlov · Oct 7, 2019 · f3fa6bc · f3fa6bc
1 parent 850c9f8
commit f3fa6bc
Show file tree

Hide file tree

Showing 11 changed files with 48 additions and 48 deletions.
diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v1_mt_word2vec_dam.json b/deeppavlov/configs/ranking/ranking_ubuntu_v1_mt_word2vec_dam.json
@@ -3,7 +3,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v1_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v1_data",
-    "num_context_turns": 10,
+    "num_context_turns": "{NUM_CONTEXT_TURNS}",
     "padding": "pre"
   },
   "dataset_iterator": {
@@ -40,7 +40,7 @@
         "save_path": "{MODELS_PATH}/ubuntu_v1_mt_word2vec_dam/preproc/tok.dict",
         "load_path": "{MODELS_PATH}/ubuntu_v1_mt_word2vec_dam/preproc/tok.dict",
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "embedding_dim": 200,
         "fit_on": ["x"],
@@ -68,7 +68,7 @@
         "class_name": "dam_nn",
         "stack_num": 5,
         "is_positional": true,
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#word2vec_embedder.dim",
         "emb_matrix": "#embeddings.emb_mat",
@@ -110,7 +110,8 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v1_mt_word2vec_dam_transformer.json b/deeppavlov/configs/ranking/ranking_ubuntu_v1_mt_word2vec_dam_transformer.json
@@ -3,7 +3,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v1_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v1_data",
-    "num_context_turns": 10,
+    "num_context_turns": "{NUM_CONTEXT_TURNS}",
     "padding": "pre"
   },
   "dataset_iterator": {
@@ -40,7 +40,7 @@
         "save_path": "{MODELS_PATH}/ubuntu_v1_mt_word2vec_dam_transformer/preproc/tok.dict",
         "load_path": "{MODELS_PATH}/ubuntu_v1_mt_word2vec_dam_transformer/preproc/tok.dict",
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "embedding_dim": 200,
         "add_raw_text": true,
@@ -69,7 +69,7 @@
         "class_name": "dam_nn_use_transformer",
         "stack_num": 5,
         "is_positional": true,
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#word2vec_embedder.dim",
         "emb_matrix": "#embeddings.emb_mat",
@@ -113,7 +113,8 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v1_mt_word2vec_smn.json b/deeppavlov/configs/ranking/ranking_ubuntu_v1_mt_word2vec_smn.json
@@ -3,7 +3,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v1_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v1_data",
-    "num_context_turns": 10,
+    "num_context_turns": "{NUM_CONTEXT_TURNS}",
     "padding": "pre"
   },
   "dataset_iterator": {
@@ -40,7 +40,7 @@
         "save_path": "{MODELS_PATH}/ubuntu_v1_mt_word2vec_smn/preproc/tok.dict",
         "load_path": "{MODELS_PATH}/ubuntu_v1_mt_word2vec_smn/preproc/tok.dict",
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "embedding_dim": 200,
         "fit_on": ["x"],
@@ -66,7 +66,7 @@
         "in_y": ["y"],
         "out": ["y_predicted"],
         "class_name": "smn_nn",
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#word2vec_embedder.dim",
         "emb_matrix": "#embeddings.emb_mat",
@@ -107,7 +107,8 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt.json b/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt.json
@@ -2,7 +2,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v2_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v2_data",
-    "num_context_turns": 10
+    "num_context_turns": "{NUM_CONTEXT_TURNS}"
   },
   "dataset_iterator": {
     "class_name": "siamese_iterator",
@@ -17,7 +17,7 @@
         "class_name": "siamese_preprocessor",
         "use_matrix": true,
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "fit_on": ["x"],
         "in": ["x"],
@@ -55,7 +55,7 @@
         "out": ["y_predicted"],
         "class_name": "bilstm_gru_nn",
         "use_matrix": "#preproc.use_matrix",
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "len_vocab": "#siam_vocab.len",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#siam_embedder.dim",
@@ -90,7 +90,8 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_interact.json b/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_interact.json
@@ -2,7 +2,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v2_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v2_data",
-    "num_context_turns": 10
+    "num_context_turns": "{NUM_CONTEXT_TURNS}"
   },
   "dataset_iterator": {
     "class_name": "siamese_iterator",
@@ -17,7 +17,7 @@
         "class_name": "siamese_preprocessor",
         "use_matrix": true,
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "fit_on": ["x"],
         "in": ["x"],
@@ -53,7 +53,7 @@
         "id": "model",
         "class_name": "bilstm_gru_nn",
         "use_matrix": "#preproc.use_matrix",
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "len_vocab": "#siam_vocab.len",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#siam_embedder.dim",
@@ -72,7 +72,7 @@
         "out": ["y_predicted"],
         "class_name": "siamese_predictor",
         "model": "#model",
-        "num_context_turns": "#model.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "batch_size": "#model.batch_size",
         "responses": "#siam_sent_vocab",
         "preproc_func": "#preproc.__call__"
@@ -99,7 +99,9 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
+
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_word2vec_dam.json b/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_word2vec_dam.json
@@ -3,7 +3,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v2_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v2_data_clean",
-    "num_context_turns": 10,
+    "num_context_turns": "{NUM_CONTEXT_TURNS}",
     "padding": "pre"
   },
   "dataset_iterator": {
@@ -40,7 +40,7 @@
         "save_path": "{MODELS_PATH}/ubuntu_v2_mt_word2vec_dam/preproc/tok.dict",
         "load_path": "{MODELS_PATH}/ubuntu_v2_mt_word2vec_dam/preproc/tok.dict",
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "embedding_dim": 200,
         "fit_on": ["x"],
@@ -68,7 +68,7 @@
         "class_name": "dam_nn",
         "stack_num": 5,
         "filters2_conv3d": 32,
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#word2vec_embedder.dim",
         "emb_matrix": "#embeddings.emb_mat",
@@ -110,7 +110,8 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_word2vec_dam_transformer.json b/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_word2vec_dam_transformer.json
@@ -3,7 +3,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v2_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v2_data_clean",
-    "num_context_turns": 10,
+    "num_context_turns": "{NUM_CONTEXT_TURNS}",
     "padding": "pre"
   },
   "dataset_iterator": {
@@ -40,7 +40,7 @@
         "save_path": "{MODELS_PATH}/ubuntu_v2_mt_word2vec_dam_transformer/preproc/tok.dict",
         "load_path": "{MODELS_PATH}/ubuntu_v2_mt_word2vec_dam_transformer/preproc/tok.dict",
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "embedding_dim": 200,
         "add_raw_text": true,
@@ -69,7 +69,7 @@
         "class_name": "dam_nn_use_transformer",
         "stack_num": 5,
         "is_positional": true,
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#word2vec_embedder.dim",
         "emb_matrix": "#embeddings.emb_mat",
@@ -113,7 +113,8 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_word2vec_smn.json b/deeppavlov/configs/ranking/ranking_ubuntu_v2_mt_word2vec_smn.json
@@ -3,7 +3,7 @@
   "dataset_reader": {
     "class_name": "ubuntu_v2_mt_reader",
     "data_path": "{DOWNLOADS_PATH}/ubuntu_v2_data_clean",
-    "num_context_turns": 10,
+    "num_context_turns": "{NUM_CONTEXT_TURNS}",
     "padding": "pre"
   },
   "dataset_iterator": {
@@ -40,7 +40,7 @@
         "save_path": "{MODELS_PATH}/ubuntu_v2_mt_word2vec_smn/preproc/tok.dict",
         "load_path": "{MODELS_PATH}/ubuntu_v2_mt_word2vec_smn/preproc/tok.dict",
         "num_ranking_samples": 10,
-        "num_context_turns": 10,
+        "num_context_turns": "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": 50,
         "embedding_dim": 200,
         "fit_on": ["x"],
@@ -66,7 +66,7 @@
         "in_y": ["y"],
         "out": ["y_predicted"],
         "class_name": "smn_nn",
-        "num_context_turns": "#preproc.num_context_turns",
+        "num_context_turns":  "{NUM_CONTEXT_TURNS}",
         "max_sequence_length": "#preproc.max_sequence_length",
         "embedding_dim": "#word2vec_embedder.dim",
         "emb_matrix": "#embeddings.emb_mat",
@@ -106,7 +106,8 @@
     "variables": {
       "ROOT_PATH": "~/.deeppavlov",
       "DOWNLOADS_PATH": "{ROOT_PATH}/downloads",
-      "MODELS_PATH": "{ROOT_PATH}/models"
+      "MODELS_PATH": "{ROOT_PATH}/models",
+      "NUM_CONTEXT_TURNS": 10
     },
     "requirements": [
       "{DEEPPAVLOV_PATH}/requirements/tf.txt",

diff --git a/deeppavlov/models/ranking/deep_attention_matching_network.py b/deeppavlov/models/ranking/deep_attention_matching_network.py
@@ -62,7 +62,6 @@ class DAMNetwork(TensorflowBaseMatchingModel):
 
     def __init__(self,
                  embedding_dim: int = 200,
-                 num_context_turns: int = 10,
                  max_sequence_length: int = 50,
                  learning_rate: float = 1e-3,
                  emb_matrix: Optional[np.ndarray] = None,
@@ -78,7 +77,6 @@ def __init__(self,
         self.seed = seed
         tf.set_random_seed(self.seed)
 
-        self.num_context_turns = num_context_turns
         self.max_sentence_len = max_sequence_length
         self.word_embedding_size = embedding_dim
         self.trainable = trainable_embeddings
@@ -90,14 +88,14 @@ def __init__(self,
         self.emb_matrix = emb_matrix
         self.decay_steps = decay_steps
 
+        super(DAMNetwork, self).__init__(*args, **kwargs)
+
         self.sess_config = tf.ConfigProto(allow_soft_placement=True)
         self.sess_config.gpu_options.allow_growth = True
         self.sess = tf.Session(config=self.sess_config)
         self._init_graph()
         self.sess.run(tf.global_variables_initializer())
 
-        super(DAMNetwork, self).__init__(*args, **kwargs)
-
         if self.load_path is not None:
             self.load()
 

diff --git a/deeppavlov/models/ranking/deep_attention_matching_network_use_transformer.py b/deeppavlov/models/ranking/deep_attention_matching_network_use_transformer.py
@@ -69,9 +69,7 @@ class DAMNetworkUSETransformer(TensorflowBaseMatchingModel):
     """
 
     def __init__(self,
-                 batch_size: int,
                  embedding_dim: int = 200,
-                 num_context_turns: int = 10,
                  max_sequence_length: int = 50,
                  learning_rate: float = 1e-3,
                  emb_matrix: Optional[np.ndarray] = None,
@@ -86,8 +84,6 @@ def __init__(self,
         self.seed = seed
         tf.set_random_seed(self.seed)
 
-        self.batch_size = batch_size
-        self.num_context_turns = num_context_turns
         self.max_sentence_len = max_sequence_length
         self.word_embedding_size = embedding_dim
         self.trainable = trainable_embeddings
@@ -97,6 +93,8 @@ def __init__(self,
         self.emb_matrix = emb_matrix
         self.decay_steps = decay_steps
 
+        super(DAMNetworkUSETransformer, self).__init__(*args, **kwargs)
+
         ##############################################################################
         self._init_graph()
         self.sess_config = tf.ConfigProto(allow_soft_placement=True)
@@ -105,9 +103,6 @@ def __init__(self,
         self.sess.run([tf.global_variables_initializer(), tf.tables_initializer()])
         ##############################################################################
 
-        super(DAMNetworkUSETransformer, self).__init__(
-            batch_size=batch_size, num_context_turns=num_context_turns, *args, **kwargs)
-
         if self.load_path is not None:
             self.load()