def load_model(MODEL_TYPE):
curr_model = None
if MODEL_TYPE == "SVM":
print("LOADING SVM...")
curr_model = load("svm.joblib")
elif MODEL_TYPE == "LR":
print("LOADING LR...")
lr = LogReg(74) #(env.matches.shape[1])
lr.load_weights("weights/weights-improvement-100-0.31.hdf5")
curr_model = lr
elif MODEL_TYPE == "DT":
print("LOADING DT...")
curr_model = load("dt.joblib")
elif MODEL_TYPE == "GB":
print("LOADING GB...")
curr_model = load("gb.joblib")
elif MODEL_TYPE == "RF":
print("LOADING RF...")
curr_model = load("rfc.joblib")
elif MODEL_TYPE == "NB":
print("LOADING NB...")
curr_model = load("nb.joblib")
elif MODEL_TYPE == "AB":
print("LOADING AB...")
curr_model = load("ab.joblib")
elif MODEL_TYPE == "DQN":
print("LOADING DQN...")
BetNet = DQNAgent(75)
BetNet.load("weights/betnet-weights-dqn.h5")
curr_model = BetNet
else:
print("LOADING NN...")
BetNet = Network(74) #(env.matches.shape[1])
BetNet.load_weights(
'weights/Adadelta/test9_400_Best/weights-improvement-400-0.48.hdf5'
) #PCA("weights/Adadelta/test13_100iter_reglast2/weights-improvement-100-0.52.hdf5") # Most recent weights
curr_model = BetNet
return curr_model
elif MODEL_TYPE == "NB":
print("LOADING NB...")
curr_model = load("nb.joblib")
elif MODEL_TYPE == "AB":
print("LOADING AB...")
curr_model = load("ab.joblib")
elif MODEL_TYPE == "DQN":
print("LOADING DQN...")
BetNet = DQNAgent(75)
BetNet.load("weights/betnet-weights-dqn.h5")
curr_model = BetNet
else:
print("LOADING NN...")
BetNet = Network(env.matches.shape[1])
BetNet.load_weights(
"weights/Adadelta/test13_100iter_reglast2/weights-improvement-100-0.52.hdf5"
) # Most recent weights
curr_model = BetNet
###############################################################################
#GETS THE PREDICTION VEC GIVEN MODEL
def generatePrediction(mt, curr_model, to_process):
prediction = None
if mt == "SVM" or mt == "DT" or mt == "GB" or mt == "NB" or mt == "RF" or mt == "AB":
temp_pred = curr_model.predict(np.asarray(to_process))
hardmax = np.zeros((1, 3))
hardmax[0][temp_pred[0]] = 1
prediction = hardmax[0]
elif mt == "LR":
# If you want, you can experiment with the parameters or use a different policy. Another popular one
# is Boltzmann-style exploration:
# policy = BoltzmannQPolicy(tau=1.)
# Feel free to give it a try!
dqn = DQNAgent(model=model, nb_actions=nb_actions, policy=policy, window_length=WINDOW_LENGTH, memory=memory,
processor=processor, nb_steps_warmup=50000, gamma=.99, delta_range=(-1., 1.),
target_model_update=10000, train_interval=4)
dqn.compile(Adam(lr=.00025), metrics=['mae'])
if args.mode == 'train':
# Okay, now it's time to learn something! We capture the interrupt exception so that training
# can be prematurely aborted. Notice that you can the built-in Keras callbacks!
weights_filename = 'dqn_{}_weights.h5f'.format(args.env_name)
checkpoint_weights_filename = 'dqn_' + args.env_name + '_weights_{step}.h5f'
log_filename = 'dqn_{}_log.json'.format(args.env_name)
callbacks = [ModelIntervalCheckpoint(checkpoint_weights_filename, interval=250000)]
callbacks += [FileLogger(log_filename, interval=100)]
dqn.fit(env, callbacks=callbacks, nb_steps=1750000, log_interval=10000)
# After training is done, we save the final weights one more time.
dqn.save_weights(weights_filename, overwrite=True)
# Finally, evaluate our algorithm for 10 episodes.
dqn.test(env, nb_episodes=10, visualize=False)
elif args.mode == 'test':
weights_filename = 'dqn_{}_weights.h5f'.format(args.env_name)
if args.weights:
weights_filename = args.weights
dqn.load_weights(weights_filename)
dqn.test(env, nb_episodes=10, visualize=True)