Upload 2 files

app.py

@@ -0,0 +1,181 @@
+# import sklearn
+import gradio as gr
+# import joblib
+import pandas as pd
+import numpy as np
+import lightgbm as lgb
+from sklearn.model_selection import train_test_split
+from PIL import Image
+# import datasets
+
+# pipe = joblib.load("./model.pkl")
+
+title = "RegMix"
+description = "TBD."
+
+df = pd.read_csv('data.csv')
+headers = df.columns.tolist()
+
+inputs = [gr.Dataframe(headers=headers, row_count = (8, "dynamic"), datatype='number', col_count=(4,"fixed"), label="Dataset", interactive=1)]
+outputs = [gr.ScatterPlot(), gr.Image(), gr.Dataframe(row_count = (2, "dynamic"), col_count=(2, "fixed"), datatype='number', label="Results", headers=["True Loss", "Pred Loss"])]
+
+def infer(inputs):
+    df = pd.DataFrame(inputs, columns=headers)
+    
+    X_columns = df.columns[0:-1]
+    y_column = df.columns[-1]
+
+    df_train, df_val = train_test_split(df, test_size=0.125, random_state=42)
+
+    hyper_params = {
+        'task': 'train',
+        'boosting_type': 'gbdt',
+        'objective': 'regression',
+        'metric': ['l1','l2'],
+        "num_iterations": 1000, 
+        'seed': 42,
+        'learning_rate': 1e-2,
+    }
+
+    target = df_train[y_column]
+    eval_target = df_val[y_column]
+        
+    np.random.seed(42)
+
+    gbm = lgb.LGBMRegressor(**hyper_params)
+
+    reg = gbm.fit(df_train[X_columns].values, target,
+        eval_set=[(df_val[X_columns].values, eval_target)],
+        eval_metric='l2',
+    callbacks=[
+        lgb.early_stopping(stopping_rounds=3),
+    ]
+        )
+    
+    predictions = reg.predict(df_val[X_columns].values)
+    df_val['Prediction'] = predictions
+
+    ####
+    import matplotlib.pyplot as plt
+    plt.rcParams["font.family"] = "Times New Roman" # !!!!
+    plt.rcParams.update({'font.size': 24})
+    plt.rcParams.update({'axes.labelpad': 20})
+
+    from matplotlib import cm
+    from matplotlib.ticker import LinearLocator
+
+    fig, ax = plt.subplots(figsize=(12, 12), layout='compressed', subplot_kw={"projection": "3d"})
+
+    stride = 0.025
+    X = np.arange(0, 1+stride, stride)
+    Y = np.arange(0, 1+stride, stride)
+
+    X, Y = np.meshgrid(X, Y)
+    Z = []
+    for (x,y) in zip(X.reshape(-1), Y.reshape(-1)):
+        if (x+y)>1:
+            Z.append(np.inf)
+        else:
+            Z.append(
+                reg.predict(np.asarray([x, y, 1-x-y]).reshape(1, -1)
+                                        )[0])
+    Z = np.asarray(Z).reshape(len(np.arange(0, 1+stride, stride)), len(np.arange(0, 1+stride, stride)))
+
+    # Plot the surface.
+    surf = ax.plot_surface(X, Y, Z, 
+                        edgecolor='white', 
+                        lw=0.5, rstride=2, cstride=2,
+                    alpha=0.85,
+                        cmap='coolwarm', 
+                        vmin=min(Z[Z!=np.inf]),
+                        vmax=max(Z[Z!=np.inf]),
+                        # linewidth=8, 
+                        antialiased=False, )
+
+    ax.zaxis.set_major_locator(LinearLocator(10))
+    ax.zaxis.set_major_formatter('{x:.02f}')
+
+    ax.view_init(elev=25, azim=45, roll=0) #####
+
+    ax.contourf(X, Y, Z, zdir='z', 
+                    offset=np.min(Z)-0.35, 
+                    cmap=cm.coolwarm)
+    
+    from matplotlib.patches import Circle
+    from mpl_toolkits.mplot3d import art3d
+
+    def add_point(ax, x, y, z, fc = None, ec = None, radius = 0.005):
+        xy_len, z_len = ax.get_figure().get_size_inches()
+        axis_length = [x[1] - x[0] for x in [ax.get_xbound(), ax.get_ybound(), ax.get_zbound()]]
+        axis_rotation =  {'z': ((x, y, z), axis_length[1]/axis_length[0]),
+                            'y': ((x, z, y), axis_length[2]/axis_length[0]*xy_len/z_len),
+                            'x': ((y, z, x), axis_length[2]/axis_length[1]*xy_len/z_len)}
+        for a, ((x0, y0, z0), ratio) in axis_rotation.items():
+            p = Circle((x0, y0), radius, lw=1.5,
+                        # width = radius, height = radius*ratio, 
+                        fc=fc,
+                        ec=ec)
+            ax.add_patch(p)
+            art3d.pathpatch_2d_to_3d(p, z=z0, zdir=a)
+
+    
+    add_point(ax, X.reshape(-1)[np.argmin(Z)], Y.reshape(-1)[np.argmin(Z)], np.min(Z), 
+            fc='Red', 
+            ec='Red', radius=0.015)
+
+    add_point(ax, X.reshape(-1)[np.argmin(Z)], Y.reshape(-1)[np.argmin(Z)], np.min(Z)-0.35, 
+            fc='Red', 
+            ec='Red', radius=0.015)
+
+
+    ax.set_xlabel('Github (%)', fontdict={
+        'size':24
+    })
+    ax.set_ylabel('Hacker News (%)', fontdict={
+        'size':24
+    })
+
+    ax.set_xticks(np.arange(0, 1, 0.2), [str(np.round(num, 1)) for num in np.arange(0, 100, 20)], )
+    ax.set_yticks(np.arange(0, 1, 0.2), [str(np.round(num, 1)) for num in np.arange(0, 100, 20)], )
+
+    ax.set_zticks(np.arange(np.min(Z), np.max(Z[Z!=np.inf]), 0.2), [str(np.round(num, 1)) for num in np.arange(np.min(Z), np.max(Z[Z!=np.inf]), 0.2)], )
+
+    ax.zaxis.labelpad=1
+
+    ax.set_zlim(np.min(Z)-0.35, max(Z[Z!=np.inf])+0.01)
+    ax.set_xlim(0, 1)
+    ax.set_ylim(0, 1)
+    ax.set_box_aspect(aspect=None, zoom=0.775)
+
+    ax.zaxis._axinfo['juggled'] = (1,2,2)
+
+    # Add a color bar which maps values to colors.
+    cbar = fig.colorbar(surf, 
+                shrink=0.5, 
+                aspect=25, pad=0.01
+                )
+    cbar.ax.set_ylabel('Prediction', fontdict={
+        'size':32
+    }, 
+                    # rotation=270, 
+                    # labelpad=-90
+                    )
+
+
+    filename = "tmp.png"
+    plt.savefig(filename, bbox_inches='tight', pad_inches=0.1)
+    ####
+    return [gr.ScatterPlot(
+            value=df_val,
+            x="Prediction",
+            y="Target",
+            title="Scatter",
+            tooltip=["Prediction", "Target"],
+            x_lim=[min(min(predictions), min(df_val[y_column]))-0.25, max(max(predictions), max(df_val[y_column]))+0.25],
+            y_lim=[min(min(predictions), min(df_val[y_column]))-0.25, max(max(predictions), max(df_val[y_column]))+0.25]
+        ), 
+        gr.Image(Image.open('tmp.png')),
+        df_val[['Target', 'Prediction']], ]
+
+gr.Interface(infer, inputs = inputs, outputs = outputs, title = title,
+                description = description, examples=[df], cache_examples=False, allow_flagging='never').launch(debug=False)

data.csv

@@ -0,0 +1,129 @@
+Github,Hacker News,Philpapers,Target
+0.616,0.227,0.157,5.306806564331055
+0.025,0.131,0.844,6.012040138244629
+0.183,0.28,0.538,5.527338981628418
+0.032,0.548,0.421,5.821677207946777
+0.002,0.347,0.651,6.326714515686035
+0.083,0.827,0.09,5.659005641937256
+0.011,0.478,0.511,6.063345909118652
+0.699,0.002,0.3,5.410695552825928
+0.363,0.389,0.249,5.3828229904174805
+0.542,0.257,0.201,5.31920337677002
+0.262,0.311,0.427,5.465338230133057
+0.677,0.251,0.072,5.3108978271484375
+0.705,0.203,0.092,5.305531024932861
+0.68,0.208,0.112,5.305020809173584
+0.109,0.22,0.671,5.645120620727539
+0.307,0.074,0.619,5.510026931762695
+0.198,0.69,0.112,5.476624488830566
+0.197,0.401,0.402,5.487608432769775
+0.056,0.338,0.606,5.778319358825684
+0.34,0.256,0.405,5.406895637512207
+0.534,0.3,0.167,5.321551322937012
+0.022,0.0,0.977,6.284101486206055
+0.732,0.228,0.04,5.291826248168945
+0.18,0.1,0.72,5.596135139465332
+0.279,0.539,0.182,5.439793586730957
+0.228,0.442,0.33,5.461205959320068
+0.278,0.271,0.45,5.446168899536133
+0.184,0.511,0.305,5.51557731628418
+0.008,0.542,0.45,6.084725379943848
+0.659,0.121,0.22,5.333937644958496
+0.134,0.041,0.825,5.712595462799072
+0.046,0.002,0.952,6.052066326141357
+0.022,0.236,0.742,6.000083923339844
+0.573,0.044,0.382,5.3919196128845215
+0.57,0.181,0.249,5.336789608001709
+0.244,0.424,0.333,5.454492568969727
+0.464,0.031,0.505,5.44106388092041
+0.349,0.046,0.605,5.503855705261231
+0.435,0.019,0.545,5.472304344177246
+0.011,0.571,0.418,6.020427703857422
+0.083,0.794,0.123,5.637621879577637
+0.433,0.125,0.442,5.397417545318604
+0.032,0.457,0.512,5.846551895141602
+0.248,0.128,0.624,5.5152506828308105
+0.159,0.747,0.094,5.518942832946777
+0.03,0.322,0.648,5.870717525482178
+0.389,0.248,0.363,5.399688720703125
+0.487,0.234,0.279,5.344883918762207
+0.385,0.363,0.252,5.373892784118652
+0.793,0.029,0.178,5.366541862487793
+0.62,0.38,0.0,5.323075294494629
+0.024,0.635,0.34,5.89354133605957
+0.848,0.152,0.0,5.330050468444824
+0.082,0.257,0.661,5.695749759674072
+0.111,0.747,0.142,5.571730136871338
+0.997,0.001,0.002,5.432588577270508
+0.484,0.064,0.452,5.41372537612915
+0.257,0.023,0.72,5.593489646911621
+0.908,0.064,0.028,5.33869743347168
+0.407,0.575,0.018,5.356371879577637
+0.716,0.209,0.074,5.299798488616943
+0.499,0.467,0.034,5.316855430603027
+0.463,0.09,0.447,5.408260822296143
+0.347,0.164,0.49,5.455391883850098
+0.22,0.31,0.47,5.478835105895996
+0.085,0.899,0.017,5.63015079498291
+0.831,0.042,0.126,5.347104549407959
+0.083,0.845,0.072,5.637035846710205
+0.009,0.352,0.639,6.105093955993652
+0.373,0.177,0.45,5.426303386688232
+0.0,1.0,0.0,6.3570756912231445
+0.001,0.0,0.999,7.201324462890625
+0.577,0.032,0.391,5.384527683258057
+0.699,0.248,0.053,5.30719518661499
+0.131,0.379,0.491,5.566975593566895
+0.042,0.865,0.093,5.747323036193848
+0.009,0.773,0.218,6.052563190460205
+0.593,0.198,0.209,5.319425582885742
+0.335,0.063,0.602,5.493361473083496
+0.508,0.2,0.292,5.362349033355713
+0.001,0.073,0.926,6.6276702880859375
+0.472,0.164,0.364,5.364439487457275
+0.021,0.415,0.563,5.923987865447998
+0.995,0.0,0.005,5.450720310211182
+0.613,0.221,0.166,5.330704689025879
+0.238,0.668,0.093,5.449023246765137
+0.521,0.08,0.399,5.382278919219971
+0.102,0.138,0.76,5.70582389831543
+0.627,0.02,0.353,5.393834590911865
+0.027,0.955,0.018,5.853610515594482
+0.215,0.713,0.071,5.452286243438721
+0.265,0.092,0.643,5.527949333190918
+0.178,0.002,0.82,5.708561897277832
+0.028,0.029,0.943,6.10589599609375
+0.002,0.305,0.693,6.344947338104248
+0.608,0.358,0.034,5.328068733215332
+0.579,0.226,0.195,5.31940221786499
+0.171,0.04,0.789,5.646611213684082
+0.056,0.483,0.461,5.769045352935791
+0.175,0.358,0.467,5.5205817222595215
+0.223,0.713,0.065,5.465519428253174
+0.359,0.095,0.546,5.459996223449707
+0.051,0.672,0.276,5.736810684204102
+0.727,0.198,0.075,5.289270401000977
+0.019,0.203,0.778,6.054460525512695
+0.12,0.877,0.003,5.571919918060303
+0.771,0.026,0.203,5.370471477508545
+0.642,0.091,0.267,5.33230447769165
+0.209,0.089,0.702,5.567383766174316
+0.603,0.036,0.361,5.384149551391602
+0.185,0.31,0.504,5.524736404418945
+0.489,0.328,0.183,5.332746982574463
+0.014,0.245,0.742,6.066995620727539
+0.75,0.241,0.009,5.288382530212402
+0.527,0.352,0.121,5.32275915145874
+0.291,0.301,0.408,5.430193901062012
+0.046,0.755,0.199,5.744025707244873
+0.031,0.949,0.02,5.835524559020996
+0.252,0.015,0.733,5.597743988037109
+0.524,0.004,0.471,5.446127414703369
+0.0,0.619,0.381,6.416290760040283
+0.08,0.903,0.017,5.660269260406494
+0.0,0.87,0.13,6.306185245513916
+0.209,0.205,0.586,5.519876003265381
+0.057,0.533,0.41,5.761989116668701
+0.307,0.597,0.096,5.392508506774902
+0.008,0.98,0.011,6.026959896087647
+0.865,0.039,0.096,5.351530075073242