aligning with branch 'deep', throw away changes pulled from 'main'

examples/ffm/generate.py

@@ -1,123 +1,76 @@
-##########################################################################################
-# A script for generation of synthetic data, suitable for sanity-checking a given binary #
-##########################################################################################
+import random, pathlib
+DATASETS_DIRECTORY = pathlib.Path("datasets")
+TRAIN_EXAMPLES = 100000
+EVAL_EXAMPLES = 10000
+NUM_ANIMALS = 5
+NUM_FOODS = 5
+BLOCK_BEYOND = 3
 
-import random
-import pathlib
-import os
-import argparse
+Aval = []
+Bval = []
 
+random.seed(1)
 
-def get_score(a, b):
-
+def get_score(a,b):
     if (a[0] == "Herbivore" and b[0] == "Plant"):
         return 1
     elif a[0] == "Carnivore" and b[0] == "Meat":
         return 1
     else:
         return -1
-
-
+
 def render_example(a, b):
+    score = get_score(a, b);
+    return " ".join([str(score), u"|A", a[0] + u"-" + str(a[1]), u"|B", b[0] + u"-" + str(b[1])]) + "\n"
 
-    score = get_score(a, b)
-    return " ".join([
-        str(score), u"|A", a[0] + u"-" + str(a[1]), u"|B",
-        b[0] + u"-" + str(b[1])
-    ]) + "\n"
-
-
-def generate_synthetic_dataset():
-
-    DATASETS_DIRECTORY.mkdir(exist_ok=True)
-
-    f = open(os.path.join(DATASETS_DIRECTORY, "vw_namespace_map.csv"), "w")
-    f.write("A,animal\n")
-    f.write("B,food\n")
 
-    i = 0
-    f = open(os.path.join(DATASETS_DIRECTORY, "train.vw"), "w")
-    while i < TRAIN_EXAMPLES:
-        animal_type = random.choices(['Herbivore', 'Carnivore'])[0]
-        food_type = random.choices(['Plant', 'Meat'])[0]
-        missone = random.randint(0, 1)
-        if missone:
-            person = random.randint(0, NUM_ANIMALS)
-            movie = random.randint(0, BLOCK_BEYOND)
-        else:
-            person = random.randint(0, BLOCK_BEYOND)
-            movie = random.randint(0, NUM_FOODS)
-        f.write(render_example((animal_type, person), (food_type, movie)))
-        i += 1
+DATASETS_DIRECTORY.mkdir(exist_ok=True);
 
-    i = 0
-    # this has the same distribution as for train...
-    f = open(os.path.join(DATASETS_DIRECTORY, "test-easy.vw"), "w")
-    while i < EVAL_EXAMPLES:
-        animal_type = random.choices(['Herbivore', 'Carnivore'])[0]
-        food_type = random.choices(['Plant', 'Meat'])[0]
-        missone = random.randint(0, 1)
-        if missone:
-            person = random.randint(0, NUM_ANIMALS)
-            movie = random.randint(0, BLOCK_BEYOND)
-        else:
-            person = random.randint(0, BLOCK_BEYOND)
-            movie = random.randint(0, NUM_FOODS)
-        f.write(render_example((animal_type, person), (food_type, movie)))
-        i += 1
+f = open(DATASETS_DIRECTORY / "vw_namespace_map.csv", "w");
+f.write("A,animal\n")
+f.write("B,food\n")
 
-    # now we will test for completely unseen combos
-    f = open(os.path.join(DATASETS_DIRECTORY, "test-hard.vw"), "w")
-    i = 0
-    while i < EVAL_EXAMPLES:
-        animal_type = random.choices(['Herbivore', 'Carnivore'])[0]
-        food_type = random.choices(['Plant', 'Meat'])[0]
-        person = random.randint(BLOCK_BEYOND + 1, NUM_ANIMALS)
-        movie = random.randint(BLOCK_BEYOND + 1, NUM_FOODS)
-
-        f.write(render_example((animal_type, person), (food_type, movie)))
-        i += 1
-
-
-if __name__ == "__main__":
-
-    parser = argparse.ArgumentParser(
-        description="FW synthetic data generator.")
-    parser.add_argument("--num_train_examples",
-                        type=int,
-                        default=1000000,
-                        help="How many instances to generate?")
-    parser.add_argument("--num_eval_examples",
-                        type=int,
-                        default=10000,
-                        help="How many instances to evaluate?")
-    parser.add_argument("--num_animals",
-                        type=int,
-                        default=5,
-                        help="How many possible animals are there?")
-    parser.add_argument("--num_foods",
-                        type=int,
-                        default=5,
-                        help="Number of possible foods for the animals?")
-    parser.add_argument("--block_beyond",
-                        type=int,
-                        default=3,
-                        help="block_beyond parameter.")
-    parser.add_argument("--random_seed",
-                        type=int,
-                        default=1,
-                        help="Random seed for the generation.")
+i = 0 
+f = open(DATASETS_DIRECTORY / "train.vw", "w")
+while i < TRAIN_EXAMPLES:
+    animal_type = random.choices(['Herbivore', 'Carnivore'])[0]
+    food_type = random.choices(['Plant', 'Meat'])[0]
+    missone = random.randint(0,1)
+    if missone:
+        person = random.randint(0, NUM_ANIMALS)
+        movie = random.randint(0, BLOCK_BEYOND)
+    else:
+        person = random.randint(0, BLOCK_BEYOND)
+        movie = random.randint(0, NUM_FOODS)
+    f.write(render_example((animal_type, person), (food_type, movie)))
+    i+=1 
+
+i = 0
+# this has the same distribution as for train...
+f = open(DATASETS_DIRECTORY /"test-easy.vw", "w")
+while i < EVAL_EXAMPLES:
+    animal_type = random.choices(['Herbivore', 'Carnivore'])[0]
+    food_type = random.choices(['Plant', 'Meat'])[0]
+    missone = random.randint(0,1)
+    if missone:
+        person = random.randint(0, NUM_ANIMALS)
+        movie = random.randint(0, BLOCK_BEYOND)
+    else:
+        person = random.randint(0, BLOCK_BEYOND)
+        movie = random.randint(0, NUM_FOODS)
+    f.write(render_example((animal_type, person), (food_type, movie)))
+    i+=1 
 
-    args = parser.parse_args()
+# now we will test for completely unseen combos
+f = open(DATASETS_DIRECTORY /"test-hard.vw", "w")
+i = 0
+while i < EVAL_EXAMPLES:
+    animal_type = random.choices(['Herbivore', 'Carnivore'])[0]
+    food_type = random.choices(['Plant', 'Meat'])[0]
+    person = random.randint(BLOCK_BEYOND+1, NUM_ANIMALS)
+    movie = random.randint(BLOCK_BEYOND+1, NUM_FOODS)
 
-    DATASETS_DIRECTORY = pathlib.Path("datasets")
-    TRAIN_EXAMPLES = args.num_train_examples
-    EVAL_EXAMPLES = args.num_eval_examples
-    NUM_ANIMALS = args.num_animals
-    NUM_FOODS = args.num_foods
-    BLOCK_BEYOND = args.block_beyond
-    Aval, Bval = [], []
+    f.write(render_example((animal_type, person), (food_type, movie)))
+    i+=1 
 
-    random.seed(args.random_seed)
 
-    generate_synthetic_dataset()