Add left-right padding checkbox

chatbot_constructor.py

@@ -30,7 +30,7 @@ def todset(text: str):
 def hash_str(data: str):
     return hashlib.md5(data.encode('utf-8')).hexdigest()
 
-def train(message: str = "", regularization: float = 0.0001, dropout: float = 0.1, learning_rate: float = 0.001, epochs: int = 16, emb_size: int = 128, input_len: int = 16, kernels_count: int = 8, kernel_size: int = 8, data: str = ""):
+def train(message: str = "", regularization: float = 0.0001, dropout: float = 0.1, learning_rate: float = 0.001, epochs: int = 16, emb_size: int = 128, input_len: int = 16, kernels_count: int = 8, kernel_size: int = 8, left_padding: bool = True, data: str = ""):
     data_hash = None
     if "→" not in data or "\n" not in data:
             if data in os.listdir("cache"):
@@ -82,12 +82,18 @@ def train(message: str = "", regularization: float = 0.0001, dropout: float = 0.
         
         X = []
         y = []
-        
-        for key in dset:
-            tokens = tokenizer.texts_to_sequences([key,])[0]
-            X.append(np.array((list(tokens)+[0,]*inp_len)[:inp_len]))
-            y.append(dset[key])
-        
+        if left_padding:
+            for key in dset:
+                tokens = tokenizer.texts_to_sequences([key,])[0]
+                X.append(np.array(([0,]*inp_len+list(tokens))[-inp_len:]))
+                y.append(dset[key])
+
+        else:
+            for key in dset:
+                tokens = tokenizer.texts_to_sequences([key,])[0]
+                X.append(np.array(list(tokens)+[0,]*inp_len)[:inp_len]))
+                y.append(dset[key])
+
         X = np.array(X)
         y = np.array(y)
         
@@ -102,14 +108,15 @@ def train(message: str = "", regularization: float = 0.0001, dropout: float = 0.
 
 if __name__ == "__main__":
     iface = gr.Interface(fn=train, inputs=["text",
-                                           gr.inputs.Slider(0, 0.01, default=0.0001, step=0.00000001, label="Regularization L1"),
-                                           gr.inputs.Slider(0, 0.5, default=0.1, step=0.00000001, label="Dropout"),
-                                           gr.inputs.Slider(0.00000001, 0.01, default=0.001, step=0.00000001, label="Learning rate"),
+                                           gr.inputs.Slider(0, 0.01, default=0.0001, step=1e-8, label="Regularization L1"),
+                                           gr.inputs.Slider(0, 0.5, default=0.1, step=1e-8, label="Dropout"),
+                                           gr.inputs.Slider(1e-8, 0.01, default=0.001, step=1e-8, label="Learning rate"),
                                            gr.inputs.Slider(1, 64, default=32, step=1, label="Epochs"),
                                            gr.inputs.Slider(1, 256, default=100, step=1, label="Embedding size"),
                                            gr.inputs.Slider(1, 128, default=16, step=1, label="Input Length"),
                                            gr.inputs.Slider(1, 128, default=64, step=1, label="Convolution kernel count"),
                                            gr.inputs.Slider(1, 16, default=8, step=1, label="Convolution kernel size"),
+                                           gr.inputs.CheckBox(False, label="Use left padding", info="Mostly used to continue a sequence"),
                                            "text"],
                          outputs="text")
     iface.launch()