Bienqq
/
car-detection-bayes
2
1
Fork 0
Code Issues Pull Requests Releases Wiki Activity

updates

This commit is contained in:
Glenn Jocher 2020-01-10 23:28:54 -08:00
parent ba265d91b2
commit fc0748f876
1 changed files with 5 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
train.py
View File

@ -451,24 +451,22 @@ if __name__ == '__main__':
if parent == 'single' or len(x) == 1:
x = x[fitness(x).argmax()]
elif parent == 'weighted': # weighted combination
n = min(10, x.shape[0]) # number to merge
n = min(10, len(x)) # number to merge
x = x[np.argsort(-fitness(x))][:n] # top n mutations
w = fitness(x) - fitness(x).min() # weights
x = (x[:n] * w.reshape(n, 1)).sum(0) / w.sum() # new parent
for i, k in enumerate(hyp.keys()):
hyp[k] = x[i + 7]
x = (x * w.reshape(n, 1)).sum(0) / w.sum() # new parent
# Mutate
np.random.seed(int(time.time()))
s = np.random.random() * 0.2 # sigma
g = [1, 1, 1, 1, 1, 1, 1, 0, .1, 1, 1, 1, 1, 1, 1, 1, 1, 1] # gains
g = (np.random.randn(len(g)) * np.array(g) * s + 1) ** 2.0 # plt.hist(x.ravel(), 300)
for i, k in enumerate(hyp.keys()):
x = (np.random.randn() * s * g[i] + 1) ** 2.0 # plt.hist(x.ravel(), 300)
hyp[k] *= float(x) # vary by sigmas
hyp[k] = x[i + 7] * g[i] # mutate parent
# Clip to limits
keys = ['lr0', 'iou_t', 'momentum', 'weight_decay', 'hsv_s', 'hsv_v', 'translate', 'scale', 'fl_gamma']
limits = [(1e-5, 1e-2), (0.00, 0.70), (0.60, 0.99), (0, 0.001), (0, .9), (0, .9), (0, .9), (0, .9), (0, 3)]
limits = [(1e-5, 1e-2), (0.00, 0.70), (0.60, 0.98), (0, 0.001), (0, .9), (0, .9), (0, .9), (0, .9), (0, 3)]
for k, v in zip(keys, limits):
hyp[k] = np.clip(hyp[k], v[0], v[1])