230523

README.md

@@ -1,8 +1,23 @@
 # pytorch study
 
-## ENV
+## BASE ENV
 ```shell
 conda create -n pt python=3.10 -y
 
 conda activate pt
 ```
+
+## MAC
+```shell
+# 安装 pytorch v1.12版本已经正式支持了用于mac m1芯片gpu加速的mps后端
+conda install pytorch::pytorch torchvision torchaudio -c pytorch -y
+
+pip install numpy
+pip install pandas
+pip install matplotlib
+```
+
+## gpt4free
+```
+pip install -U g4f[all]
+```

autograd.py

@@ -0,0 +1,15 @@
+import torch
+from torch import autograd
+
+device = torch.device('mps')
+
+x = torch.tensor(1.)
+a = torch.tensor(2., requires_grad=True)
+b = torch.tensor(2., requires_grad=True)
+c = torch.tensor(3., requires_grad=True)
+
+y = a ** 2 * x + b * x + c ** 3
+
+print('before:', a.grad, b.grad, c.grad)
+grads = autograd.grad(y, [a, b, c])
+print('after:', grads[0], grads[1], grads[2])

gpt.py

@@ -0,0 +1,11 @@
+from g4f.client import Client
+
+content = "张量在机器学习中的主要用途"
+
+
+client = Client()
+response = client.chat.completions.create(
+    model="gpt-4o",
+    messages=[{"role": "user", "content": content}],
+)
+print(response.choices[0].message.content)

linear regression/1.py

@@ -0,0 +1,56 @@
+import numpy as np
+import torch
+
+
+def compute_error_for_line_given_points(b, w, points):
+    totalError = 0
+    N = float(len(points))
+    for i in range(len(points)):
+        x = points[i][0]
+        y = points[i][1]
+        totalError += (y - (w * x + b)) ** 2
+    return totalError / N
+
+
+def step_gradient(b_current, w_current, points, learningRate):
+    b_gradient = torch.tensor(0.0, device=points.device, dtype=torch.float32)
+    w_gradient = torch.tensor(0.0, device=points.device, dtype=torch.float32)
+    N = float(len(points))
+    for i in range(len(points)):
+        x = points[i][0]
+        y = points[i][1]
+        b_gradient += -(2 / N) * (y - (w_current * x + b_current) + b_current)
+        w_gradient += -(2 / N) * x * (y - (w_current * x + b_current + b_current))
+    new_b = b_current - (learningRate * b_gradient)
+    new_w = w_current - (learningRate * w_gradient)
+    return [new_b, new_w]
+
+
+def gradient_descent_runner(points, starting_b, starting_w, learningRate, num_iterations):
+    b = torch.tensor(starting_b, device=points.device, dtype=torch.float32)
+    w = torch.tensor(starting_w, device=points.device, dtype=torch.float32)
+    for i in range(num_iterations):
+        b, w = step_gradient(b, w, points, learningRate)
+    return [b, w]
+
+
+def run():
+    # 修改为生成数据的文件路径
+    points_np = np.genfromtxt("data1.csv", delimiter=',').astype(np.float32)
+    points = torch.tensor(points_np, device='mps')
+    learning_rate = 0.0001  # 使用较小的学习率
+    initial_b = 0.0
+    initial_w = 0.0
+    num_iterations = 1000
+    print("Starting gradient descent at b={0},w={1},error={2}".format(initial_b, initial_w,
+                                                                      compute_error_for_line_given_points(initial_b,
+                                                                                                          initial_w,
+                                                                                                          points)))
+    print("running...")
+    [b, w] = gradient_descent_runner(points, initial_b, initial_w, learning_rate, num_iterations)
+    print("After gradient descent at b={0},w={1},error={2}".format(b.item(), w.item(),
+                                                                   compute_error_for_line_given_points(b, w, points)))
+
+
+if __name__ == '__main__':
+    run()

linear regression/data1.csv

@@ -0,0 +1,100 @@
+0.0,2.4360562173289115
+0.10101010101010101,2.4288710820592065
+0.20202020202020202,3.677943201375977
+0.30303030303030304,3.5029234515863217
+0.40404040404040403,4.007715980839878
+0.5050505050505051,3.95999321461469
+0.6060606060606061,3.220853916066527
+0.7070707070707071,3.2211460206798623
+0.8080808080808081,4.2516957270374505
+0.9090909090909091,4.311826715084292
+1.0101010101010102,4.153966583608258
+1.1111111111111112,4.224290328721461
+1.2121212121212122,4.551324602105953
+1.3131313131313131,5.157200101408408
+1.4141414141414141,5.199011258508288
+1.5151515151515151,5.248911218901843
+1.6161616161616161,5.789628423512512
+1.7171717171717171,5.126592322934872
+1.8181818181818181,4.546631494636344
+1.9191919191919191,5.7260434379514065
+2.0202020202020203,5.607446671816119
+2.121212121212121,5.401744626671172
+2.2222222222222223,5.568078510495838
+2.323232323232323,6.136817713051054
+2.4242424242424243,5.399802896696589
+2.525252525252525,6.7465591899811415
+2.6262626262626263,6.510002771256968
+2.727272727272727,6.194107987238278
+2.8282828282828283,6.280445605022811
+2.929292929292929,6.413289184504817
+3.0303030303030303,8.178951965980268
+3.131313131313131,7.438933818741419
+3.2323232323232323,8.161193108124682
+3.3333333333333335,6.466487953447159
+3.4343434343434343,7.6815673373443385
+3.5353535353535355,7.412509123916619
+3.6363636363636362,7.712231039046388
+3.7373737373737375,7.512155302443977
+3.8383838383838382,8.169468174953455
+3.9393939393939394,8.201406255891817
+4.040404040404041,9.413915839209679
+4.141414141414141,7.2131607261403
+4.242424242424242,8.244196707034996
+4.343434343434343,8.059400613529792
+4.444444444444445,9.127093042087843
+4.545454545454545,8.232456814994503
+4.646464646464646,9.026988954051767
+4.747474747474747,8.936405824368308
+4.848484848484849,8.838334259675397
+4.94949494949495,9.717080564295035
+5.05050505050505,9.635892324495916
+5.151515151515151,10.802758752616178
+5.252525252525253,9.889268431487773
+5.353535353535354,9.262021983987134
+5.454545454545454,9.905732041295009
+5.555555555555555,9.697006564677089
+5.656565656565657,10.435437946557755
+5.757575757575758,10.257651825530608
+5.858585858585858,11.394734709569004
+5.959595959595959,10.872621683473387
+6.0606060606060606,10.750944058491058
+6.161616161616162,11.375400587831757
+6.262626262626262,11.834436555701465
+6.363636363636363,11.536088544119654
+6.4646464646464645,11.261555999325722
+6.565656565656566,12.529961808490153
+6.666666666666667,12.19345219105891
+6.767676767676767,11.950653180245155
+6.8686868686868685,12.176773142948385
+6.96969696969697,12.055083206520518
+7.070707070707071,13.498633194384489
+7.171717171717171,12.542518727882712
+7.2727272727272725,13.318372269865769
+7.373737373737374,12.542630883166513
+7.474747474747475,12.93490122675753
+7.575757575757575,14.4040220344926
+7.6767676767676765,13.314367294113964
+7.777777777777778,14.061236496574551
+7.878787878787879,12.686346979737731
+7.979797979797979,14.024375221983842
+8.080808080808081,13.7042096336008
+8.181818181818182,13.342730021126272
+8.282828282828282,14.136548357864573
+8.383838383838384,14.619569834949138
+8.484848484848484,14.01453898823226
+8.585858585858587,15.154877807203663
+8.686868686868687,14.081910297898048
+8.787878787878787,14.474564310016353
+8.88888888888889,14.966525346412723
+8.98989898989899,15.526107019435932
+9.09090909090909,14.352357736719853
+9.191919191919192,15.843742065895144
+9.292929292929292,15.787083172159111
+9.393939393939394,15.211828607109144
+9.494949494949495,15.845176532492374
+9.595959595959595,15.622518083688107
+9.696969696969697,15.589081237426006
+9.797979797979798,15.511248085690712
+9.8989898989899,16.27050774059862
+10.0,16.1105549896166

linear regression/np_genPoints.py

@@ -0,0 +1,30 @@
+import numpy as np
+import csv
+
+# 定义回归方程参数
+w = 1.35
+b = 2.89
+
+# 生成x值范围
+x_min = 0
+x_max = 10
+
+# 生成100个在x轴附近的点
+x = np.linspace(x_min, x_max, 100)
+
+# 根据回归方程计算y值
+y = w * x + b
+
+# 添加一些噪声，使数据更真实
+y += np.random.normal(scale=0.5, size=y.shape)
+
+# 将x和y合并成一个二维数组
+data = np.column_stack((x, y))
+
+# 将数据保存到CSV文件
+with open('data1.csv', 'w', newline='') as csvfile:
+    writer = csv.writer(csvfile)
+    # 写入表头
+    # writer.writerow(['x', 'y'])
+    # 写入数据
+    writer.writerows(data)

linear regression/plt_print.py

@@ -0,0 +1,22 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
+# 原始数据
+points = np.genfromtxt("data1.csv", delimiter=',')
+x = points[:, 0]
+y = points[:, 1]
+
+# 拟合直线
+x_range = np.linspace(min(x), max(x), 100)
+y_pred = 1.6455038785934448 * x_range + 1.827562689781189
+
+# 绘图
+plt.figure(figsize=(8, 6))
+plt.scatter(x, y, color='blue', label='Original data')
+plt.plot(x_range, y_pred, color='red', label='Fitted line')
+plt.xlabel('X')
+plt.ylabel('Y')
+plt.title('Fitting a line to random data')
+plt.legend()
+plt.grid(True)
+plt.savefig('print1.png')

requirements.txt

@@ -0,0 +1,7 @@
+pytorch::pytorch
+torchvision
+torchaudio
+pandas
+matplotlib
+numpy
+g4f

test/macTest.py

@@ -0,0 +1,6 @@
+import torch
+
+print(torch.backends.mps.is_available())
+print(torch.backends.mps.is_built())
+
+print(torch.device("mps"))

test/performance.py

@@ -0,0 +1,29 @@
+import torch
+import time
+
+print(torch.__version__)
+print(torch.backends.mps.is_available())
+print(torch.cuda.is_available())
+
+a = torch.randn(10000,1000)
+b = torch.randn(1000,2000)
+
+t0 = time.time()
+c = torch.matmul(a, b)
+t1 = time.time()
+print(a.device,t1-t0,c.norm(2))
+
+device = torch.device('mps')
+
+a = a.to(device)
+b = b.to(device)
+
+t0 = time.time()
+c = torch.matmul(a, b)
+t1 = time.time()
+print(a.device,t1-t0,c.norm(2))
+
+t0 = time.time()
+c = torch.matmul(a, b)
+t1 = time.time()
+print(a.device,t1-t0,c.norm(2))