add: data generator

Board/__init__.py

@@ -15,6 +15,10 @@ class Board:
             self.lives.add((x, y))
 
     def evaluate(self):
+        new_lives = self.try_evaluate()
+        self.lives = new_lives
+
+    def try_evaluate(self) -> set[(int, int)]:
         new_lives = set()
         for (x, y) in itertools.product(range(self.width), range(self.height)):
             neighbor_count = 0
@@ -33,5 +37,18 @@ class Board:
                 new_lives.add((x, y))
             if (x, y) not in self.lives and neighbor_count == 3:
                 new_lives.add((x, y))
-        self.lives = new_lives
+        return new_lives
 
+    def __str__(self):
+        res = ""
+        for y in range(self.height):
+            for x in range(self.width):
+                if (x, y) in self.lives:
+                    res += "o"
+                else:
+                    res += " "
+            res += "\n"
+        return res
+
+    def __repr__(self):
+        return self.__str__()

DataGenerator/CircleGenerator.py

@@ -0,0 +1,56 @@
+import random
+
+from Board import Board
+from DataGenerator import DataGenerator
+
+def _circle_generator(cx, cy, px, py):
+    yield cx + px, cy + py
+    yield cx - px, cy + py
+    yield cx + px, cy - py
+    yield cx - px, cy - py
+    yield cx + py, cy + px
+    yield cx - py, cy + px
+    yield cx + py, cy - px
+    yield cx - py, cy - px
+
+def circle_generator(cx, cy, radius):
+    x = 0
+    y = radius
+    d = 1 - radius
+    for tx, ty in _circle_generator(cx, cy, x, y):
+        yield tx, ty
+    while x < y:
+        x += 1
+        if d < 0:
+            d += 2 * x + 1
+        else:
+            y -= 1
+            d += 2 * (x - y) + 1
+        for tx, ty in _circle_generator(cx, cy, x, y):
+            yield tx, ty
+
+class CircleGenerator(DataGenerator):
+    def generate(self, amount: int) -> list[Board]:
+        res = []
+        generated = set()
+        for i in range(amount):
+            success = False
+            while not success:
+                success = True
+                b = Board(self.width, self.height, self.qx, self.qy)
+                cx, cy, radius = self.random_config()
+                while (cx, cy, radius) in generated:
+                    cx, cy, radius = self.random_config()
+                for tx, ty in circle_generator(cx, cy, radius):
+                    b.lives.add((tx, ty))
+                if len(b.try_evaluate()) == 0:
+                    success = False
+                    continue
+                res.append(b)
+        return res
+
+    def random_config(self):
+        cx = random.randint(min(5, self.width), self.width - 1)
+        cy = random.randint(min(5, self.width), self.height - 1)
+        radius = random.randint(2, min(self.width, self.height))
+        return cx, cy, radius

DataGenerator/__init__.py

@@ -0,0 +1,17 @@
+from abc import ABC, abstractmethod
+
+from Board import Board
+
+
+class DataGenerator:
+    def __init__(self, width, height, qx, qy):
+        self.width = width
+        self.height = height
+        self.qx = qx
+        self.qy = qy
+
+    @abstractmethod
+    def generate(self, amount: int) -> list[Board]:
+        pass
+
+

NeuralSolver/NeuralSolver.py

@@ -1,7 +1,7 @@
 import numpy as np
 import tensorflow as tf
 from tensorflow import keras
-
+from tensorflow.keras import regularizers
 from Board import Board
 import random
 import os
@@ -29,8 +29,10 @@ class NeuralSolver:
 
     def _build_reverse_model(self):
         inputs = keras.Input(shape=(self.Width, self.Height, 1), name="FinalState")
-        hidden = keras.layers.Conv2D(32, 3, padding="same", activation="relu")(inputs)
+        hidden = keras.layers.Conv2D(32, 3, padding="same", activation="relu",kernel_regularizer=regularizers.l2(1e-4))(inputs)
-        hidden = keras.layers.Conv2D(32, 3, padding="same", activation="relu")(hidden)
+        hidden = keras.layers.BatchNormalization()(hidden)
+        hidden = keras.layers.Conv2D(32, 5, padding="same", activation="relu",kernel_regularizer=regularizers.l2(1e-4))(hidden)
+        hidden = keras.layers.Conv2D(32, 3, padding="same", activation="relu",kernel_regularizer=regularizers.l2(1e-4))(hidden)
         outputs = keras.layers.Conv2D(1, 1, padding="same", activation="sigmoid")(hidden)
         self.ReverseModel = keras.Model(inputs, outputs, name="ReverseModel")
 
@@ -102,7 +104,7 @@ class NeuralSolver:
 
 
     def spec(self):
-        return f"{self.Width}x{self.Height}x{self.QuotientX}x{self.QuotientY}"
+        return f"{self.Width}x{self.Height}x{self.QuotientX}x{self.QuotientY}.weights"
 
 
     def save_model(self):

app.py

@@ -56,10 +56,13 @@ def train(request: TrainRequest, background_tasks: BackgroundTasks):
 def try_load(request: TrainRequest, background_tasks: BackgroundTasks):
     global solver
     global status
+    if status != "initialized":
+        return {"Status": "model not initialized"}
     if solver.load_model():
         status = "ready"
         return {"Status": "ok"}
-    background_tasks.add_task(lambda : task(request))
+    else:
+        background_tasks.add_task(lambda : task(request))
 
 
 class BoardRequest(BaseModel):
@@ -115,7 +118,6 @@ def load():
     return {"Status": "model loading failed"}
 
 
-
 @app.get("/status")
 def get_status():
     global status