ThiagoLira

from tqdm import tqdm

pbar = tqdm(range(50))
for epoch in pbar:  # loop over the dataset multiple times
    
    running_loss = 0.0
    # sample a bunch of timesteps
    Ts = np.random.randint(1,num_diffusion_timesteps, size=training_steps_per_epoch)
    for _, t in enumerate(Ts):
        # produce corrupted sample

ThiagoLira

def denoise_with_mu(denoise_model, x_t, t, list_alpha, list_alpha_bar, DATA_SIZE, device):
    """
    Denoising function considering the denoising models tries to model the posterior mean
    """
    alpha_t = list_alpha[t]
    beta_t = 1 - alpha_t
    alpha_bar_t = list_alpha_bar[t]
    
    mu_theta = denoise_model(x_t,t)
    

ThiagoLira

def q_sample(x_start, t, list_bar_alphas, device):
    """
    Diffuse the data (t == 0 means diffused for 1 step)
    """
    alpha_bar_t = list_bar_alphas[t]
    
    mean = alpha_bar_t*x_start
    cov = torch.eye(x_start.shape[0]).to(device)
    cov = cov*(1-alpha_bar_t)

ThiagoLira

// Update State
// hook position
bobberBarPos = Helper.Reflection.GetField<float>(bar, "bobberBarPos").GetValue();
// fish position
bobberPosition = Helper.Reflection.GetField<float>(bar, "bobberPosition").GetValue();
// hook speed
bobberBarSpeed = Helper.Reflection.GetField<float>(bar, "bobberBarSpeed").GetValue();
// amount of green bar filled
distanceFromCatching = Helper.Reflection.GetField<float>(bar, "distanceFromCatching").GetValue();

ThiagoLira

replayMemory[updateCounter,0] = OldState[0];
replayMemory[updateCounter,1] = OldState[1];
replayMemory[updateCounter,2] = OldState[2];
replayMemory[updateCounter,3] = NewState[0];
replayMemory[updateCounter,4] = NewState[1];
replayMemory[updateCounter,5] = NewState[2];
replayMemory[updateCounter,6] = reward;
replayMemory[updateCounter,7] = actionBuffer? 1 : 0;

ThiagoLira

[HarmonyPatch(typeof(Game1), "isOneOfTheseKeysDown")]
class IsButtonDownHack
{
    // ...
    // some important stuff
    // ...
    
    // change function return value to true
    // makes the game think a mouse left button click ocurred
    __result = true;

ThiagoLira

public int Update(double[] currentState)
{

    Tensor<double> input = new DenseTensor<double>(new[] {3});

    input[0] = currentState[0];
    input[1] = currentState[1];
    input[2] = currentState[2];

    // Setup inputs and outputs

ThiagoLira

@app.route('/',methods=['GET']) 
def generate():
 # Since Flask forks the python process to answer for requests
 # we need to do this to avoid errors with tensorflow
 tf.reset_default_graph()
 
 
 # Start tf session and load model into memory
 sess = gpt2.start_tf_sess(threads=1)
 gpt2.load_gpt2(sess)

ThiagoLira

; Equal?

;numbers
(test (v*s-v (interpS '(equal? 3 2))) (numV 0))
(test (v*s-v (interpS '(equal? (+ 3 2) 5))) (numV 1))
;closures
(test (v*s-v (interpS '(equal? (lambda x (car x)) (lambda y (car y))))) (numV 0))
(test (v*s-v (interpS '(equal? (lambda x (car x)) (lambda x (car x))))) (numV 1))

;boxes