baseline | Bohrium-玻尔科研空间站

空间站广场

论文

Notebooks

比赛

课程

Apps

我的主页

我的Notebooks

我的论文库

我的足迹

我的工作空间

任务

节点

文件

数据集

镜像

项目

数据库

公开

baseline

AI4S

洋憨憨

发布于 2023-10-25

推荐镜像 :Third-party software:ai4s-cup-0.1

推荐机型 :c12_m46_1 * NVIDIA GPU B

[1]

import torch

import math

import torch.nn as nn

import torch.optim as optim

from torch.nn.utils.rnn import pad_sequence

import torch.nn.functional as F

import random

import os

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

import pickle

/opt/conda/lib/python3.8/site-packages/tqdm/auto.py:22: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

代码

文本

[ ]

with open('/bohr/ai4spulseeis-lr97/v3/train_datasets/train_pulse_1.pkl', 'rb') as fp:

pulse_train_1 = pickle.load(fp, encoding='bytes')

pulse_train_1.keys()

代码

文本

[2]

def setup_seed(seed):

np.random.seed(seed) # Numpy module.

random.seed(seed) # Python random module.

os.environ['PYTHONHASHSEED'] = str(seed)

torch.manual_seed(seed)

if torch.cuda.is_available():

torch.cuda.manual_seed(seed)

torch.cuda.manual_seed_all(seed)

torch.backends.cudnn.benchmark = False

torch.backends.cudnn.deterministic = True

代码

文本

[3]

setup_seed(0)

代码

文本

[5]

class MLPModel(nn.Module):

def __init__(self):

super(MLPModel, self).__init__()

self.fc1 = nn.Linear(99, 128)

self.fc2 = nn.Linear(128, 102)

self.relu = nn.ReLU()

def forward(self, x):

x = x.view(-1, 99)

x = self.relu(self.fc1(x))

x = self.fc2(x)

x = x.view(-1, 51, 2)

return x

代码

文本

[6]

model = MLPModel()

print(model)

MLPModel(
  (fc1): Linear(in_features=99, out_features=128, bias=True)
  (fc2): Linear(in_features=128, out_features=102, bias=True)
  (relu): ReLU()
)

代码

文本

[7]

def input_mlp(datasets, mode):

input_data = []

soc_lst = [f'{i*2}%SOC' for i in range(49)]

for i in datasets:

with open(f'/bohr/ai4spulseeis-lr97/v3/{mode}_datasets/{mode}_pulse_{i}.pkl', 'rb') as fp:

pulse_data = pickle.load(fp, encoding='bytes')

for soc in soc_lst:

Vol = pulse_data[soc]['Voltage']

tensor_vol = torch.tensor(Vol).view(1,99,1)

input_data.append(tensor_vol)

return input_data

代码

文本

[8]

train_baty = [1,2,3,4,5,6]

test_baty = [1,2]

input_train = input_mlp(datasets = train_baty, mode = 'train')

input_test = input_mlp(datasets = test_baty, mode = 'test')

print(input_train[0].shape, len(input_train), len(input_test))

torch.Size([1, 99, 1]) 294 98

代码

文本

[9]

def target_mlp(datasets, mode):

target_data = []

soc_lst = [f'{i*2}%SOC' for i in range(49)]

EIS_list = []

for k in datasets:

with open(f'/bohr/ai4spulseeis-lr97/v3/{mode}_datasets/{mode}_eis_{k}.pkl', 'rb') as fp:

eis_data = pickle.load(fp, encoding='bytes')

for soc in soc_lst:

EIS_tot = [[],[]]

re = eis_data[soc]['Real']

im = eis_data[soc]['Imaginary']

EIS_tot[0] = re

EIS_tot[1] = im

EIS_list.append(EIS_tot)

EIS_list = [np.array(t).squeeze().T for t in EIS_list]

target_data = [torch.tensor(t).float().view(1, 51, 2) for t in EIS_list]

return target_data

代码

文本

[10]

target_train = target_mlp(datasets = train_baty, mode = 'train')

print(target_train[0].shape, len(target_train))

torch.Size([1, 51, 2]) 294

代码

文本

[12]

if torch.cuda.is_available():

model = model.cuda()

for i in range(len(input_train)):

input_train[i] = input_train[i].cuda()

target_train[i] = target_train[i].cuda()

for i in range(len(input_test)):

input_test[i] = input_test[i].cuda()

print("CUDE = True")

CUDE = True

代码

文本

[13]

criterion = nn.MSELoss()

optimizer = optim.Adam(model.parameters(), lr=1e-4)

scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer, 'min', factor=0.8, patience=40, verbose=True, min_lr=5e-6)

num_epochs = 5000

batch_size = 8

代码

文本

[14]

train_losses = [] # 记录每个epoch的训练损失

for epoch in range(num_epochs):

epoch_loss = 0.0

train_size = len(input_train)

for batch_idx in range(0, train_size, batch_size):

batch_input = torch.cat(input_train[batch_idx:batch_idx+batch_size], dim=0)

batch_target = torch.cat(target_train[batch_idx:batch_idx+batch_size], dim=0)

optimizer.zero_grad()

outputs = model(batch_input)

loss = criterion(outputs, batch_target)

loss.backward()

optimizer.step()

epoch_loss += loss.item() * batch_input.shape[0]

scheduler.step(epoch_loss / (train_size / batch_size))

train_losses.append(epoch_loss / (train_size / batch_size))

if (epoch + 1) % 50 == 0:

mean_mse = epoch_loss / (train_size / batch_size)

rmse = mean_mse ** 0.5

print(f"Epoch [{epoch + 1}/{num_epochs}] | Loss: {mean_mse:.4f} | RMSE: {rmse:.4f}")

Epoch [50/5000] | Loss: 42.0119 | RMSE: 6.4817
Epoch 00062: reducing learning rate of group 0 to 8.0000e-05.
Epoch [100/5000] | Loss: 41.9762 | RMSE: 6.4789
Epoch 00103: reducing learning rate of group 0 to 6.4000e-05.
Epoch 00146: reducing learning rate of group 0 to 5.1200e-05.
Epoch [150/5000] | Loss: 41.7950 | RMSE: 6.4649
Epoch [200/5000] | Loss: 41.7837 | RMSE: 6.4640
Epoch [250/5000] | Loss: 41.7694 | RMSE: 6.4629
Epoch [300/5000] | Loss: 41.7548 | RMSE: 6.4618
Epoch [350/5000] | Loss: 41.7400 | RMSE: 6.4607
Epoch [400/5000] | Loss: 41.7252 | RMSE: 6.4595
Epoch [450/5000] | Loss: 41.7105 | RMSE: 6.4584
Epoch [500/5000] | Loss: 41.6957 | RMSE: 6.4572
Epoch [550/5000] | Loss: 41.6809 | RMSE: 6.4561
Epoch [600/5000] | Loss: 41.6661 | RMSE: 6.4549
Epoch [650/5000] | Loss: 41.6513 | RMSE: 6.4538
Epoch [700/5000] | Loss: 41.6365 | RMSE: 6.4526
Epoch [750/5000] | Loss: 41.6217 | RMSE: 6.4515
Epoch [800/5000] | Loss: 41.6069 | RMSE: 6.4503
Epoch [850/5000] | Loss: 41.5921 | RMSE: 6.4492
Epoch [900/5000] | Loss: 41.5772 | RMSE: 6.4480
Epoch [950/5000] | Loss: 41.5624 | RMSE: 6.4469
Epoch [1000/5000] | Loss: 41.5475 | RMSE: 6.4457
Epoch [1050/5000] | Loss: 41.5326 | RMSE: 6.4446
Epoch [1100/5000] | Loss: 41.5178 | RMSE: 6.4434
Epoch [1150/5000] | Loss: 41.5029 | RMSE: 6.4423
Epoch [1200/5000] | Loss: 41.4880 | RMSE: 6.4411
Epoch [1250/5000] | Loss: 41.4730 | RMSE: 6.4400
Epoch [1300/5000] | Loss: 41.4581 | RMSE: 6.4388
Epoch [1350/5000] | Loss: 41.4431 | RMSE: 6.4376
Epoch [1400/5000] | Loss: 41.4282 | RMSE: 6.4365
Epoch [1450/5000] | Loss: 41.4132 | RMSE: 6.4353
Epoch [1500/5000] | Loss: 41.3981 | RMSE: 6.4341
Epoch [1550/5000] | Loss: 41.3831 | RMSE: 6.4330
Epoch [1600/5000] | Loss: 41.3681 | RMSE: 6.4318
Epoch [1650/5000] | Loss: 41.3530 | RMSE: 6.4306
Epoch [1700/5000] | Loss: 41.3379 | RMSE: 6.4295
Epoch [1750/5000] | Loss: 41.3228 | RMSE: 6.4283
Epoch [1800/5000] | Loss: 41.3076 | RMSE: 6.4271
Epoch [1850/5000] | Loss: 41.2925 | RMSE: 6.4259
Epoch [1900/5000] | Loss: 41.2773 | RMSE: 6.4247
Epoch [1950/5000] | Loss: 41.2621 | RMSE: 6.4236
Epoch [2000/5000] | Loss: 41.2468 | RMSE: 6.4224
Epoch [2050/5000] | Loss: 41.2315 | RMSE: 6.4212
Epoch [2100/5000] | Loss: 41.2162 | RMSE: 6.4200
Epoch [2150/5000] | Loss: 41.2009 | RMSE: 6.4188
Epoch [2200/5000] | Loss: 41.1855 | RMSE: 6.4176
Epoch [2250/5000] | Loss: 41.1701 | RMSE: 6.4164
Epoch [2300/5000] | Loss: 41.1547 | RMSE: 6.4152
Epoch [2350/5000] | Loss: 41.1393 | RMSE: 6.4140
Epoch [2400/5000] | Loss: 41.1238 | RMSE: 6.4128
Epoch [2450/5000] | Loss: 41.1082 | RMSE: 6.4116
Epoch [2500/5000] | Loss: 41.0927 | RMSE: 6.4104
Epoch [2550/5000] | Loss: 41.0771 | RMSE: 6.4091
Epoch [2600/5000] | Loss: 41.0614 | RMSE: 6.4079
Epoch [2650/5000] | Loss: 41.0458 | RMSE: 6.4067
Epoch [2700/5000] | Loss: 41.0300 | RMSE: 6.4055
Epoch [2750/5000] | Loss: 41.0143 | RMSE: 6.4042
Epoch [2800/5000] | Loss: 40.9985 | RMSE: 6.4030
Epoch [2850/5000] | Loss: 40.9826 | RMSE: 6.4018
Epoch [2900/5000] | Loss: 40.9667 | RMSE: 6.4005
Epoch [2950/5000] | Loss: 40.9508 | RMSE: 6.3993
Epoch [3000/5000] | Loss: 40.9348 | RMSE: 6.3980
Epoch [3050/5000] | Loss: 40.9188 | RMSE: 6.3968
Epoch [3100/5000] | Loss: 40.9027 | RMSE: 6.3955
Epoch [3150/5000] | Loss: 40.8866 | RMSE: 6.3943
Epoch [3200/5000] | Loss: 40.8704 | RMSE: 6.3930
Epoch [3250/5000] | Loss: 40.8541 | RMSE: 6.3917
Epoch [3300/5000] | Loss: 40.8379 | RMSE: 6.3905
Epoch [3350/5000] | Loss: 40.8215 | RMSE: 6.3892
Epoch [3400/5000] | Loss: 40.8051 | RMSE: 6.3879
Epoch [3450/5000] | Loss: 40.7886 | RMSE: 6.3866
Epoch [3500/5000] | Loss: 40.7721 | RMSE: 6.3853
Epoch [3550/5000] | Loss: 40.7555 | RMSE: 6.3840
Epoch [3600/5000] | Loss: 40.7389 | RMSE: 6.3827
Epoch [3650/5000] | Loss: 40.7222 | RMSE: 6.3814
Epoch [3700/5000] | Loss: 40.7054 | RMSE: 6.3801
Epoch [3750/5000] | Loss: 40.6886 | RMSE: 6.3788
Epoch [3800/5000] | Loss: 40.6717 | RMSE: 6.3774
Epoch [3850/5000] | Loss: 40.6547 | RMSE: 6.3761
Epoch [3900/5000] | Loss: 40.6376 | RMSE: 6.3748
Epoch [3950/5000] | Loss: 40.6205 | RMSE: 6.3734
Epoch [4000/5000] | Loss: 40.6033 | RMSE: 6.3721
Epoch [4050/5000] | Loss: 40.5860 | RMSE: 6.3707
Epoch [4100/5000] | Loss: 40.5687 | RMSE: 6.3694
Epoch [4150/5000] | Loss: 40.5512 | RMSE: 6.3680
Epoch [4200/5000] | Loss: 40.5337 | RMSE: 6.3666
Epoch [4250/5000] | Loss: 40.5161 | RMSE: 6.3652
Epoch [4300/5000] | Loss: 40.4984 | RMSE: 6.3638
Epoch [4350/5000] | Loss: 40.4806 | RMSE: 6.3624
Epoch [4400/5000] | Loss: 40.4627 | RMSE: 6.3610
Epoch [4450/5000] | Loss: 40.4448 | RMSE: 6.3596
Epoch [4500/5000] | Loss: 40.4267 | RMSE: 6.3582
Epoch [4550/5000] | Loss: 40.4085 | RMSE: 6.3568
Epoch [4600/5000] | Loss: 40.3903 | RMSE: 6.3553
Epoch [4650/5000] | Loss: 40.3719 | RMSE: 6.3539
Epoch [4700/5000] | Loss: 40.3534 | RMSE: 6.3524
Epoch [4750/5000] | Loss: 40.3348 | RMSE: 6.3510
Epoch [4800/5000] | Loss: 40.3161 | RMSE: 6.3495
Epoch [4850/5000] | Loss: 40.2973 | RMSE: 6.3480
Epoch [4900/5000] | Loss: 40.2784 | RMSE: 6.3465
Epoch [4950/5000] | Loss: 40.2593 | RMSE: 6.3450
Epoch [5000/5000] | Loss: 40.2401 | RMSE: 6.3435

代码

文本

[15]

def evaluation():

model.eval()

outputs_eval = []

for batch_idx in range(0, len(input_test), batch_size):

batch_input = torch.cat(input_test[batch_idx:batch_idx+batch_size], dim=0)

with torch.no_grad():

output = model(batch_input)

outputs_eval.append(output)

return outputs_eval

代码

文本

[16]

outputs_eval = evaluation()

print('outputs:', len(outputs_eval), outputs_eval[0].shape)

outputs: 13 torch.Size([8, 51, 2])

代码

文本

[17]

predict_outputs = torch.cat(outputs_eval, dim=0).view(1, len(input_test)*51, 2).tolist()

predict_data = np.array(predict_outputs).squeeze().T

print(predict_data.shape)

(2, 4998)

代码

文本

[18]

results_data = {}

results_data['test_data_number'] = [1 for i in range(49*51)] + [2 for i in range(49*51)]

results_data['SOC(%)'] = []

for i in range(49):

results_data['SOC(%)'] += [i*2 for j in range(51)]

for i in range(49):

results_data['SOC(%)'] += [i*2 for j in range(51)]

results_data['EIS_real'] = predict_data[0].tolist()

results_data['EIS_imaginary'] = predict_data[1].tolist()

data_submit = pd.DataFrame(results_data)

data_submit.to_csv('submission.csv', index=False, header=True)

print(data_submit)

      test_data_number  SOC(%)   EIS_real  EIS_imaginary
0                    1       0  13.229534      -7.091629
1                    1       0  13.180953      -5.373521
2                    1       0  13.199190      -3.968995
3                    1       0  13.245510      -2.822267
4                    1       0  13.367637      -1.843068
...                ...     ...        ...            ...
4993                 2      96  37.379169       2.062883
4994                 2      96  37.594353       2.261335
4995                 2      96  37.823223       2.526676
4996                 2      96  38.096107       2.826188
4997                 2      96  38.392311       3.200425

[4998 rows x 4 columns]

代码

文本

AI4S

点个赞吧