Exploring mrdbourke's awesome 🔥 course on Deep Learning using Tensorflow. Beautiful Jupyter notebooks, a well structured installable Python package, issues tracking on a Kanban board, and a deployed Github Page website. Come dive in!
Meanwhile checkout my repositories below 👇
class LearningCurve:
def __init__(self, model):
self.is_fitted = model.history is None
def from_history(self, history):
pass
def from_history_dict(self, history_dict):
passFeatures:
See this Stackoverflow question
I want this for Notebook 01C for transforming y by boxcox and then inverse transforming it back.
Tasks
ClassicImageDatasetDirectoryTasks:
Reproduce BUG:
from src.image import ClassicImageDataDirectory
import numpy as np
import matplotlib.pyplot as plt
data_dir = '../data/pizza_steak/'
imgdir = ClassicImageDataDirectory(data_dir, target_image_size=(224, 224), dtype=np.uint8)
datagen = imgdir.load(4)
batch = next(datagen)
batch.view_random_images(class_names='all', n_each=3)
plt.show()
class ClassificationEvaluator:
def __init__(self, train_preds, test_preds):
pass
Features:
kind: [tp, fp, tn, fn]conf: [low, medium, high]def show_prediction(kind, conf):
passintercept = tf.Variable(polymodel.params[0], dtype=tf.float32)
slope = tf.Variable(polymodel.params[1], dtype=tf.float32)
def predict(intercept, slope, X):
m = PolynomialModel([intercept, slope], 0)
return m(X)
def mse(labels, preds):
return tf.keras.losses.mse(tf.squeeze(labels), tf.squeeze(preds))
def loss_func(intercept, slope, X, y):
predictions = predict(intercept, slope, X)
return mse(y, predictions)
with tf.GradientTape() as tape:
tape.watch(intercept)
tape.watch(slope)
opt = tf.keras.optimizers.Adam(lr=0.05)
for i in range(20):
opt.minimize(lambda : loss_func(intercept, slope, X, y),
var_list=[intercept, slope])Traceback:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-20-3ef69543ef29> in <module>
19 opt = tf.keras.optimizers.Adam(learning_rate=0.01)
20 for i in range(20):
---> 21 opt.minimize(lambda : loss_func(intercept, slope, X, y), var_list=[intercept, slope])
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\optimizer_v2\optimizer_v2.py in minimize(self, loss, var_list, grad_loss, name, tape)
496 grads_and_vars = self._compute_gradients(
497 loss, var_list=var_list, grad_loss=grad_loss, tape=tape)
--> 498 return self.apply_gradients(grads_and_vars, name=name)
499
500 def _compute_gradients(self, loss, var_list, grad_loss=None, tape=None):
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\optimizer_v2\optimizer_v2.py in apply_gradients(self, grads_and_vars, name, experimental_aggregate_gradients)
596 RuntimeError: If called in a cross-replica context.
597 """
--> 598 grads_and_vars = optimizer_utils.filter_empty_gradients(grads_and_vars)
599 var_list = [v for (_, v) in grads_and_vars]
600
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\optimizer_v2\utils.py in filter_empty_gradients(grads_and_vars)
77 if not filtered:
78 raise ValueError("No gradients provided for any variable: %s." %
---> 79 ([v.name for _, v in grads_and_vars],))
80 if vars_with_empty_grads:
81 logging.warning(
ValueError: No gradients provided for any variable: ['Variable:0', 'Variable:0'].
def from_countdict(self, train_dict, test_dict=None, classes=None):
passWould be possible to abstract away almost all logic from default __init__ to reuse self.from_countdict
Tasks:
data_dir = '../data/pizza_steak/'
imgdir = ClassicImageDataDirectory(data_dir)>>> imgdir.lab2nm
{0: 'steak', 1: 'pizza'}This should be sorted in alphabetical order i.e. {0: 'pizza', 1: 'steak'}
Something like this:
def plot_learning_curve(model_or_history, extra_metric=None, include_validation=True):
if isinstance(model_or_history, dict):
...
elif isinstance(model_or_history, (tf.keras.models.Sequential, tf.keras.models.Model)):
...Tasks:
Generally, what I want to define the shuffling in such a way that each batch is a good mix and represents the population distribution
For e.g., if we have 10k images -> 5k pizza, 5k steak. So each batch of 32 should ideally have 16 pizza and 16 steak. But should all pizza be pepporini and all steak be same kind of steak. In real world we would not have labels beyond class labels, but we can cluster the images and sample based on the cluster.
└── src/
├── __init__.py
├── evaluate
├── image
├── models
├── preprocessing
├── text
├── tfplay
├── utils
└── visualize
Notebooks to refactor:
This is more like a terminology thing, but keep the following convention:
class_label: The numeric label we assign to each classclass_name: The actual class name in wordsPredict medical cost based on sex, age, bmi, smoker status, region.
inputs = tf.keras.layers.Input(shape=(224, 224, 3), name="input_layer")
data_augmentation = tf.keras.Sequential([
preprocessing.RandomFlip("horizontal"),
preprocessing.RandomRotation(0.2),
preprocessing.RandomZoom(0.2),
preprocessing.RandomHeight(0.2),
preprocessing.RandomWidth(0.2),
], name ="data_augmentation")
data_augmentation(inputs)Traceback:
---------------------------------------------------------------------------
NotImplementedError Traceback (most recent call last)
<ipython-input-94-7923874bb0d7> in <module>
10 ], name ="data_augmentation")
11
---> 12 data_augmentation(inputs)
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\engine\base_layer.py in __call__(self, *args, **kwargs)
950 if _in_functional_construction_mode(self, inputs, args, kwargs, input_list):
951 return self._functional_construction_call(inputs, args, kwargs,
--> 952 input_list)
953
954 # Maintains info about the `Layer.call` stack.
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\engine\base_layer.py in _functional_construction_call(self, inputs, args, kwargs, input_list)
1089 # Check input assumptions set after layer building, e.g. input shape.
1090 outputs = self._keras_tensor_symbolic_call(
-> 1091 inputs, input_masks, args, kwargs)
1092
1093 if outputs is None:
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\engine\base_layer.py in _keras_tensor_symbolic_call(self, inputs, input_masks, args, kwargs)
820 return nest.map_structure(keras_tensor.KerasTensor, output_signature)
821 else:
--> 822 return self._infer_output_signature(inputs, args, kwargs, input_masks)
823
824 def _infer_output_signature(self, inputs, args, kwargs, input_masks):
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\engine\base_layer.py in _infer_output_signature(self, inputs, args, kwargs, input_masks)
861 # TODO(kaftan): do we maybe_build here, or have we already done it?
862 self._maybe_build(inputs)
--> 863 outputs = call_fn(inputs, *args, **kwargs)
864
865 self._handle_activity_regularization(inputs, outputs)
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\engine\sequential.py in call(self, inputs, training, mask)
387 kwargs['training'] = training
388
--> 389 outputs = layer(inputs, **kwargs)
390
391 if len(nest.flatten(outputs)) != 1:
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\engine\base_layer.py in __call__(self, *args, **kwargs)
1010 with autocast_variable.enable_auto_cast_variables(
1011 self._compute_dtype_object):
-> 1012 outputs = call_fn(inputs, *args, **kwargs)
1013
1014 if self._activity_regularizer:
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\layers\preprocessing\image_preprocessing.py in call(self, inputs, training)
865
866 output = control_flow_util.smart_cond(training, random_rotated_inputs,
--> 867 lambda: inputs)
868 output.set_shape(inputs.shape)
869 return output
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\utils\control_flow_util.py in smart_cond(pred, true_fn, false_fn, name)
113 pred, true_fn=true_fn, false_fn=false_fn, name=name)
114 return smart_module.smart_cond(
--> 115 pred, true_fn=true_fn, false_fn=false_fn, name=name)
116
117
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\framework\smart_cond.py in smart_cond(pred, true_fn, false_fn, name)
52 if pred_value is not None:
53 if pred_value:
---> 54 return true_fn()
55 else:
56 return false_fn()
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\layers\preprocessing\image_preprocessing.py in random_rotated_inputs()
859 return transform(
860 inputs,
--> 861 get_rotation_matrix(angles, img_hd, img_wd),
862 fill_mode=self.fill_mode,
863 fill_value=self.fill_value,
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\keras\layers\preprocessing\image_preprocessing.py in get_rotation_matrix(angles, image_height, image_width, name)
755 math_ops.cos(angles)[:, None],
756 y_offset[:, None],
--> 757 array_ops.zeros((num_angles, 2), dtypes.float32),
758 ],
759 axis=1)
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\util\dispatch.py in wrapper(*args, **kwargs)
199 """Call target, and fall back on dispatchers if there is a TypeError."""
200 try:
--> 201 return target(*args, **kwargs)
202 except (TypeError, ValueError):
203 # Note: convert_to_eager_tensor currently raises a ValueError, not a
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\ops\array_ops.py in wrapped(*args, **kwargs)
2817
2818 def wrapped(*args, **kwargs):
-> 2819 tensor = fun(*args, **kwargs)
2820 tensor._is_zeros_tensor = True
2821 return tensor
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\ops\array_ops.py in zeros(shape, dtype, name)
2866 # Create a constant if it won't be very big. Otherwise create a fill
2867 # op to prevent serialized GraphDefs from becoming too large.
-> 2868 output = _constant_if_small(zero, shape, dtype, name)
2869 if output is not None:
2870 return output
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\ops\array_ops.py in _constant_if_small(value, shape, dtype, name)
2802 def _constant_if_small(value, shape, dtype, name):
2803 try:
-> 2804 if np.prod(shape) < 1000:
2805 return constant(value, shape=shape, dtype=dtype, name=name)
2806 except TypeError:
<__array_function__ internals> in prod(*args, **kwargs)
~\anaconda3\envs\ds-py37\lib\site-packages\numpy\core\fromnumeric.py in prod(a, axis, dtype, out, keepdims, initial, where)
3029 """
3030 return _wrapreduction(a, np.multiply, 'prod', axis, dtype, out,
-> 3031 keepdims=keepdims, initial=initial, where=where)
3032
3033
~\anaconda3\envs\ds-py37\lib\site-packages\numpy\core\fromnumeric.py in _wrapreduction(obj, ufunc, method, axis, dtype, out, **kwargs)
85 return reduction(axis=axis, out=out, **passkwargs)
86
---> 87 return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
88
89
~\anaconda3\envs\ds-py37\lib\site-packages\tensorflow\python\framework\ops.py in __array__(self)
853 "Cannot convert a symbolic Tensor ({}) to a numpy array."
854 " This error may indicate that you're trying to pass a Tensor to"
--> 855 " a NumPy call, which is not supported".format(self.name))
856
857 def __len__(self):
NotImplementedError: Cannot convert a symbolic Tensor (data_augmentation/random_rotation_10/rotation_matrix/strided_slice:0) to a numpy array. This error may indicate that you're trying to pass a Tensor to a NumPy call, which is not supported
Tasks:
From Notebook 01C:
fig, axn = plt.subplots(2, 3, figsize=(16, 8))
columns = {'cont': ['age', 'bmi'],
'cat': ['sex', 'children', 'smoker', 'region']}
axn_flat = axn.flatten()
i = 0
for vartype, varlist in columns.items():
for var in varlist:
ax = axn_flat[i]
if vartype == 'cont':
df[var].plot(kind='hist', ax=ax)
else:
d = df[var].value_counts()
d.plot(kind='bar', ax=ax)
ax.set_title(var, fontdict=dict(weight='bold', size=10))
i += 1
imgdir = ClassicImageDataDirectory(data_dir, target_image_size=IMAGE_DIM, dtype=np.uint8)
batch = imgdir.load(128)
imgen = next(batch)
imgen.view_random_images(class_names=np.random.choice(imgen.class_names, size=10, replace=False))Traceback:
--------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-25-88675e61ccb9> in <module>
1 batch = imgdir.load(128)
2 imgen = next(batch)
----> 3 imgen.view_random_images(class_names=np.random.choice(imgen.class_names, size=10, replace=False))
c:\users\bhati\google drive\projects\tensorflow-tutorial-daniel-bourke\src\image\dataset.py in view_random_images(self, class_names, n_each, subset, cmap)
63
64
---> 65 if class_names == 'any':
66 class_names = [np.random.choice(self.class_names)]
67 elif class_names == 'all':
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()
Reimplement the google's tensorflow playground
def add_index_to_dataframe(df, cols):
passcols to an existing index in dataframeTasks:
Refer Notebook 01B
# Fit the model
>>> history = tfmodel.fit(X, ys['linear'], epochs=10)Output:
Epoch 1/10
32/32 [==============================] - 0s 781us/step - loss: 1457.4397 - mse: 1457.4397
Epoch 2/10
32/32 [==============================] - 0s 785us/step - loss: 40.0271 - mse: 40.0271
Epoch 3/10
32/32 [==============================] - 0s 725us/step - loss: 25.4607 - mse: 25.4607
Epoch 4/10
32/32 [==============================] - 0s 635us/step - loss: 25.9577 - mse: 25.9577
Epoch 5/10
32/32 [==============================] - 0s 646us/step - loss: 25.4081 - mse: 25.4081
Epoch 6/10
32/32 [==============================] - 0s 732us/step - loss: 24.5854 - mse: 24.5854
Epoch 7/10
32/32 [==============================] - 0s 685us/step - loss: 27.7506 - mse: 27.7506
Epoch 8/10
32/32 [==============================] - 0s 637us/step - loss: 25.3395 - mse: 25.3395
Epoch 9/10
32/32 [==============================] - 0s 611us/step - loss: 24.8364 - mse: 24.8364
Epoch 10/10
32/32 [==============================] - 0s 757us/step - loss: 25.0970 - mse: 25.0970
>>> history.history['mse'][788.6161499023438,
38.7478141784668,
25.893997192382812,
25.751623153686523,
25.781696319580078,
25.781963348388672,
25.767667770385742,
25.736785888671875,
25.817962646484375,
25.85219955444336]
Create a linear model data sampler to generate linear model data for
Take cues for this example in Notebook 01A:
xmin, xmax = -10, 40 # range
m = 1000 # no of data points
X = np.expand_dims(np.linspace(xmin, xmax, m), axis=-1) # You can randomly sample as well
# Create the labels aka the outputs (dependent variable)
polymodels = {}
polymodels['linear'] = PolynomialModel([2, 3], 5)
polymodels['quadratic'] = PolynomialModel([1, 2, -0.1], 10)
# Now sample some y
ys = {}
ys['linear'] = polymodels['linear'](X, add_noise=True)
ys['quadratic'] = polymodels['quadratic'](X, add_noise=True)
Tasks:
class LabelAnalyzer:
def __init__(self, train_labels, test_labels=None, class_labels=None):
self.train = self._get_count_dict(train_labels)
self.__has_test_data = test_labels is not None
if self.__has_test_data:
self.test = self._get_count_dict(test_labels)
self._make_count_df() Assign attributes self.class_labels and self.labelmap in these cases:
From notebook 01C
pipe = Pipeline([
('preprocess', preprocess),
('scale', StandardScaler()),
('reg', LinearRegression())
])
# NOTE: Increasing tolerance to make Lasso converge (https://stackoverflow.com/questions/20681864/lasso-on-sklearn-does-not-converge)
param_grid = [{'reg': [LinearRegression()]},
{'reg': [Ridge(), Lasso()], 'reg__alpha': [0.01, 0.1, 1, 10, 100]}]
grid_search = GridSearchCV(pipe, param_grid, scoring='neg_mean_squared_error')
grid_search.fit(data_train, label_train)Output warning:
C:\Users\bhati\anaconda3\envs\ds-py37\lib\site-packages\sklearn\linear_model\_coordinate_descent.py:532: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 892757429.305973, tolerance: 10875092.86663517
positive)
Tasks:
def create_tensorboard_callback(experiment, task=None, parent_dir=None):
...
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.