def _gpt2_eval_the_model(model, eval_dataloader, config):
device = config["device"]
model.eval()
eval_loss, eval_accuracy = 0, 0
nb_eval_steps, nb_eval_examples = 0, 0
for batch in tqdm(eval_dataloader, desc="Evaluating"):
batch = tuple(t.to(device) for t in batch)
input_ids, mc_token_ids, lm_labels, mc_labels = batch
with torch.no_grad():
_, mc_loss = model(input_ids, mc_token_ids, lm_labels, mc_labels)
_, mc_logits, pre = model(input_ids, mc_token_ids)
mc_logits = mc_logits.detach().cpu().numpy()
mc_labels = mc_labels.to('cpu').numpy()
tmp_eval_accuracy = func.accuracy(mc_logits, mc_labels)
eval_loss += mc_loss.mean().item()
eval_accuracy += tmp_eval_accuracy
nb_eval_examples += input_ids.size(0)
nb_eval_steps += 1
eval_loss = eval_loss / nb_eval_steps
eval_accuracy = eval_accuracy / nb_eval_examples
return eval_loss, eval_accuracy
trainLength = int(len(allData) * 0.6)
testLength = len(allData) - trainLength
trainData = []
for i in range(trainLength):
trainData.append(random.choice(allData))
[xTrainData, yTrainData] = func.parseInfo(trainData)
w_temp2[t] = linearRegression.LinearRegression(
xTrainData, yTrainData)
##---- Get the error norm and score on the training data -----##
TrainToolsDict_temp[t] = func.getPredictedScoreError(
xTrainData, yTrainData, w_temp2[t])
testData = [item for item in allData if item not in trainData]
[xTestData, yTestData] = func.parseInfo(testData)
accuracy_lst.append(func.accuracy(w_temp2[t], xTestData,
yTestData))
#func.reportAccuracy(tool, accuracy_lin[tool])
## --- Get the mean of acuuracies --- ##
accuracy_lin[tool] = sum(accuracy_lst) / len(accuracy_lst) * 100
func.reportAccuracy(tool, accuracy_lin[tool])
## --- Get the mean of weights --- ##
w_temp2_array = [w_temp2[i] for i in w_temp2.keys()]
w_temp2_t = numpy.transpose(w_temp2_array)
w[tool] = [
sum(w_temp2_t[i]) / len(w_temp2_t[i])
for i in range(0, len(w_temp2_t))
]
if __name__ == "__main__":
# 载入数据，生成散点图
x, y = creatdata()
x = np.c_[np.ones(x.shape[0]), x]
plt.scatter(x[:, 1], x[:, 2], c=y)
plt.show()
y = one_hot_code(y)
# 划分数据集
x_train, x_val, y_train, y_val = train_test_split(x, y, test_size=0.2)
# 使用分类器
cls = softmax_cls()
y_val = np.argmax(y_val, axis=1)
w, err = cls.fit(x_train, y_train, learning_rate=0.1, epochs=1000, L2=0.1)
# 预测
y_pred = cls.predict(x_val)
y_pred = np.argmax(y_pred, axis=1)
print(accuracy(y_val, y_pred))
compare = (y_val == y_pred) * 1
plt.scatter(x_val[:, 1], x_val[:, 2], c=y_val)
plt.show()
plt.scatter(x_val[:, 1], x_val[:, 2], c=compare)
plt.legend(["1", "0"])
plt.show()
plt.plot(err)
plt.show()
temp = np.c_[y_val, y_pred]
pass