franespiga

rm -rf lambda_deployment
mkdir - p lambda_deployment/var/task/models

cp requirements.txt lambda_deployment/requirements.txt
cp lambda_dockerfile lambda_deployment/Dockerfile

aws_account_id=$AWS_ACCOUNT
aws_region="us-east-1"

# Store model

franespiga

FROM public.ecr.aws/lambda/python:3.8
ADD . /
WORKDIR /
ENV AWS_ACCESS_KEY_ID=$AWS_ACCESS_KEY_ID
ENV AWS_SECRET_ACCESS_KEY=$AWS_SECRET_ACCESS_KEY
ENV AWS_SESSION_TOKEN=$AWS_SESSION_TOKEN
RUN pip install -r requirements.txt
CMD ["handler.handler"]   

franespiga

import requests
import json
url = "http://localhost:9000/2015-03-31/functions/function/invocations"
headers = {'Content-type': 'application/json'} 
q = requests.post(url, headers = headers, 
                  data = json.dumps({"body": json.dumps({"image":grayscale_array.tolist(), 'source':'payload_array', 'preprocessing_type':'bw'})}))

franespiga

class environment:
    def __init__(self, probabilities):
        """
        Arguments:
        probabilities: `list` or `np.array` with the probabilities of success of each action.
        """
        self.probabilities = probabilities
    def step(self, action):
        """
        Environment step returning the reward of each action

franespiga

env = environment([0.5, 0.75, 0.25])

n_steps = 10000
print("Average reward:\naction 0:{}\naction 1:{}\naction 2:{}\n".format(
    np.mean([env.step(0) for i in range(n_steps)]),
    np.mean([env.step(1) for i in range(n_steps)]),
    np.mean([env.step(2) for i in range(n_steps)])
))

franespiga

def argmax(q_values):
    """
    Takes in a list of q_values and returns the index
    of the item with the highest value. Breaks ties randomly.
    returns: int - the index of the highest value in q_values
    """
    top = float("-inf")
    ties = []
    
    for i in range(len(q_values)):

franespiga

num_runs = 200                      # Number of simulations
num_steps = 100000                  # Number of visits (timesteps)
probabilities = [0.5, 0.75, 0.25]
epsilon = 0.15
all_averages = []
actions = []
for i in tqdm(range(num_runs)):
    env = environment(probabilities) # env initialization
    agent = EpsilonGreedyAgent(action_size = 3, epsilon = epsilon) # agent initialization

franespiga

epsilons = [0.0, 0.01, 0.1, 0.4, 0.5]

num_runs = 200
num_steps = 50000

per_epsilon = {}
per_epsilon_action = {}

for epsilon in epsilons:
    all_averages = []

franespiga

num_runs = 200                      # Número de veces que repetimos el experimento
num_steps = 50000                  # Número de visitas (pasos) por experimento
probabilities = [0.5, 0.75, 0.25]

epsilons = [0.0, 0.01, 0.1, 0.4, 0.5]

per_epsilon_priorkw = {}
for epsilon in epsilons:
    all_averages = []
    for i in tqdm(range(num_runs)):

franespiga

n_samples = 10000

# features -------------------------------------------------------
design = np.random.randint(0,10,n_samples)/10
consumption = np.random.randint(0,10,n_samples)/10
reliability = np.random.randint(0,10,n_samples)/10
innovation = np.random.randint(0,10,n_samples)/10
price = np.random.randint(0,10,n_samples)/10
X = np.array([design, consumption, reliability, innovation, price]).transpose()