I am having a weird "ValueError: mean must be 1 dimensional" when I am trying to build a Hierarchical GL-LVM model. Basically I'm trying to reproduce this paper: Hierarchical Gaussian Process Latent Variable Models using GPflow.
Therefore I implemented my own new model as follow:
class myGPLVM(gpflow.models.BayesianModel):
def __init__(self, data, latent_data, x_data_mean, kernel):
super().__init__()
print("GPLVM")
self.kernel0 = kernel[0]
self.kernel1 = kernel[1]
self.mean_function = Zero()
self.likelihood0 = gpflow.likelihoods.Gaussian(1.0)
self.likelihood1 = gpflow.likelihoods.Gaussian(1.0)
# make some parameters
self.data = (gpflow.Parameter(x_data_mean), gpflow.Parameter(latent_data), data)
def hierarchy_ll(self):
x, h, y = self.data
K = self.kernel0(x)
num_data = x.shape[0]
k_diag = tf.linalg.diag_part(K)
s_diag = tf.fill([num_data], self.likelihood0.variance)
ks = tf.linalg.set_diag(K, k_diag + s_diag)
L = tf.linalg.cholesky(ks)
m = self.mean_function(x)
return multivariate_normal(h, m, L)
def log_likelihood(self):
"""
Computes the log likelihood.
.. math::
\log p(Y | \theta).
"""
x, h, y = self.data
K = self.kernel1(h)
num_data = h.shape[0]
k_diag = tf.linalg.diag_part(K)
s_diag = tf.fill([num_data], self.likelihood1.variance)
ks = tf.linalg.set_diag(K, k_diag + s_diag)
L = tf.linalg.cholesky(ks)
m = self.mean_function(h)
# [R,] log-likelihoods for each independent dimension of Y
log_prob = multivariate_normal(y, m, L). # <- trows the error!
log_prob_h = self.hierarchy_ll()
log_likelihood = tf.reduce_sum(log_prob) + tf.reduce_sum(log_prob_h)
return log_likelihood
The model seems to work with a toy example:
from sklearn.datasets.samples_generator import make_blobs
X, y = make_blobs(n_samples=40, centers=3, n_features=12, random_state=2)
Y = tf.convert_to_tensor(X, dtype=default_float())
but fails and trough me the error when I am trying with a bvh file (the one from the paper actually). I also used Lawrence's code to read my bvh from mocap which I modified to fit python3
Anyway, it's been few a days and I am out of ideas. I tried multiple way to force my mean array "m" to be of one dimensional but nothing worked. I also tried with the "three_phase_oil_flow" dataset from the first GPLVM paper which works as well.
Therefore, I would assume that my model is correct, or at least I got some optimisation going on, and would think that perhaps the bvh reader could be the cause. But the data seems all fine to me... Especially I don't understand why when forcing multivariate function like:
m = np.zeros((np.shape(m)[0], 1))
log_prob = multivariate_normal(y, m, L)
or even with the gpflow Zero function
m = Zero(h)
log_prob = multivariate_normal(y, m, L)
it still trows me the error. Any help will be highly appreciated.
edited thanks to: Artem Artemev
The rest of the code if anyone wants to try to reproduce:
https://github.com/michaelStettler/h-GPLVM
error flow:
(venv) MacBookMichael2:stackOverflow michaelstettler$ python3 HGPLVM.py
(199, 96)
shape Y (199, 3, 38)
2020-01-26 17:00:48.104029: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2020-01-26 17:00:48.113609: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7f8dd5ff5410 executing computations on platform Host. Devices:
2020-01-26 17:00:48.113627: I tensorflow/compiler/xla/service/service.cc:175] StreamExecutor device (0): Host, Default Version
shape Y (199, 38)
Number of points: 199 and Number of dimensions: 38
shape x_mean_latent (199, 8)
shape x_mean_init (199, 2)
HGPLVM
gpr_data (199, 2) (199, 8) (199, 38)
2020-01-26 17:00:48.139003: W tensorflow/python/util/util.cc:299] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.
shape m (199, 1)
Traceback (most recent call last):
File "HGPLVM.py", line 131, in <module>
_ = opt.minimize(closure, method="bfgs", variables=model.trainable_variables, options=dict(maxiter=maxiter))
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/gpflow/optimizers/scipy.py", line 60, in minimize
**scipy_kwargs)
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/scipy/optimize/_minimize.py", line 594, in minimize
return _minimize_bfgs(fun, x0, args, jac, callback, **options)
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/scipy/optimize/optimize.py", line 998, in _minimize_bfgs
gfk = myfprime(x0)
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/scipy/optimize/optimize.py", line 327, in function_wrapper
return function(*(wrapper_args + args))
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/scipy/optimize/optimize.py", line 73, in derivative
self(x, *args)
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/scipy/optimize/optimize.py", line 65, in __call__
fg = self.fun(x, *args)
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/gpflow/optimizers/scipy.py", line 72, in _eval
loss, grads = _compute_loss_and_gradients(closure, variables)
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/gpflow/optimizers/scipy.py", line 116, in _compute_loss_and_gradients
loss = loss_cb()
File "HGPLVM.py", line 127, in closure
return - model.log_marginal_likelihood()
File "/Users/michaelstettler/PycharmProjects/GPflow/venv/lib/python3.6/site-packages/gpflow/models/model.py", line 45, in log_marginal_likelihood
return self.log_likelihood(*args, **kwargs) + self.log_prior()
File "HGPLVM.py", line 62, in log_likelihood
log_prob = multivariate_normal(y, m, L)
File "mtrand.pyx", line 3729, in numpy.random.mtrand.RandomState.multivariate_normal
ValueError: mean must be 1 dimensional
I would recommend posting a working MWE code. I have tried to use your code snippets, but it gives me errors.
I don't have issues with multivariate_normal function. If you have localised the issue correctly you can debug TF2.0 more thoroughly and find the place that causes that exception. Here is the code which I'm running:
In [2]: from sklearn.datasets.samples_generator import make_blobs
...: X, y = make_blobs(n_samples=40, centers=3, n_features=12, random_state=2)
In [10]: m = np.zeros((np.shape(y)[0], 1))
In [11]: m.shape
Out[11]: (40, 1)
In [12]: y.shape
Out[12]: (40,)
In [13]: L = np.eye(m.shape[0])
In [15]: gpflow.logdensities.multivariate_normal(y, m, L)
Out[15]:
<tf.Tensor: shape=(40,), dtype=float64, numpy=
array([ -56.75754133, ...])>
Related
For a long time I was able to add data and fit, then plot the curve with data. But recently I get this:
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-6-6f645a2744bc> in <module>
1 poland = prepare_data(europe_data, 'Poland')
----> 2 plot_all(poland, max_y=400000)
3 poland
~/Pulpit/library.py in plot_all(country, max_x, max_y)
43 def plot_all(country, max_x = 1000, max_y = 500000):
44
---> 45 parameters_logistic = scipy.optimize.curve_fit(func_logistic, country['n'], country['all'])[0]
46 parameters_expo = scipy.optimize.curve_fit(func_expo, country['n'], country['all'])[0]
47
/usr/local/lib64/python3.6/site-packages/scipy/optimize/minpack.py in curve_fit(f, xdata, ydata, p0, sigma, absolute_sigma, check_finite, bounds, method, jac, **kwargs)
787 cost = np.sum(infodict['fvec'] ** 2)
788 if ier not in [1, 2, 3, 4]:
--> 789 raise RuntimeError("Optimal parameters not found: " + errmsg)
790 else:
791 # Rename maxfev (leastsq) to max_nfev (least_squares), if specified.
RuntimeError: Optimal parameters not found: Number of calls to function has reached maxfev = 800.
Here are all Python Jupyter Notebook files: https://files.fm/u/zj7cc6ne#sign_up
How to solve this?
scipy.optimize.curve_fit takes a keyword argument p0.
Initial guess for the parameters (length N). If None, then the initial
values will all be 1 (if the number of parameters for the function can
be determined using introspection, otherwise a ValueError is raised).
If the defaults 1 are too far of from the result the algorithm may not converge. Try to put some values that make sense for your problem.
I am struggling to understand the behaviour of dsolve for this simple ODE:
Y''(t) = b*Y'(t) + f(t)
For some reason, dsolve throws an error if I use f(t)=exp(t-a), but for general f(t) or f(t)=exp(a*t) or if I put a value for a, dsolve succeeds. The complete error message:
File "~/.local/lib/python3.7/site-packages/sympy/solvers/ode.py",
line 679, in dsolve
return _helper_simplify(eq, hint, hints, simplify, ics=ics)
File "~/.local/lib/python3.7/site-packages/sympy/solvers/ode.py",
line 704, in _helper_simplify
sols = solvefunc(eq, func, order, match)
File "~/.local/lib/python3.7/site-packages/sympy/solvers/ode.py",
line 5674, in ode_nth_linear_constant_coeff_undetermined_coefficients
return _solve_undetermined_coefficients(eq, func, order, match)
File "~/.local/lib/python3.7/site-packages/sympy/solvers/ode.py",
line 5766, in _solve_undetermined_coefficients
coeffsdict[s[x]] += s['coeff']
KeyError: exp(t)
I am using this code:
from sympy import symbols, Function, dsolve, exp, Eq
a, b, t = symbols('a b t')
Y = Function('Y')(t)
#f = Function('f')(t) # works
#f = exp(a*t) # works
f = exp(t-a) # KeyError: exp(t)
#f = exp(t-2) # works
odeY = Eq( Y.diff(t,t), b*Y.diff(t) + f )
dsolve(odeY,Y)
I am using sympy version 1.5.1 with python3.7
Many thanks!
I am trying to build a neural network model with one hidden layer (1024 nodes). The hidden layer is nothing but a relu unit. I am also processing the input data in batches of 128.
The inputs are images of size 28 * 28. In the following code I get the error in line
_, c = sess.run([optimizer, loss], feed_dict={x: batch_x, y: batch_y})
Error: TypeError: Cannot interpret feed_dict key as Tensor: Tensor Tensor("Placeholder_64:0", shape=(128, 784), dtype=float32) is not an element of this graph.
Here is the code I have written
#Initialize
batch_size = 128
layer1_input = 28 * 28
hidden_layer1 = 1024
num_labels = 10
num_steps = 3001
#Create neural network model
def create_model(inp, w, b):
layer1 = tf.add(tf.matmul(inp, w['w1']), b['b1'])
layer1 = tf.nn.relu(layer1)
layer2 = tf.matmul(layer1, w['w2']) + b['b2']
return layer2
#Initialize variables
x = tf.placeholder(tf.float32, shape=(batch_size, layer1_input))
y = tf.placeholder(tf.float32, shape=(batch_size, num_labels))
w = {
'w1': tf.Variable(tf.random_normal([layer1_input, hidden_layer1])),
'w2': tf.Variable(tf.random_normal([hidden_layer1, num_labels]))
}
b = {
'b1': tf.Variable(tf.zeros([hidden_layer1])),
'b2': tf.Variable(tf.zeros([num_labels]))
}
init = tf.initialize_all_variables()
train_prediction = tf.nn.softmax(model)
tf_valid_dataset = tf.constant(valid_dataset)
tf_test_dataset = tf.constant(test_dataset)
model = create_model(x, w, b)
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(model, y))
optimizer = tf.train.GradientDescentOptimizer(0.5).minimize(loss)
#Process
with tf.Session(graph=graph1) as sess:
tf.initialize_all_variables().run()
total_batch = int(train_dataset.shape[0] / batch_size)
for epoch in range(num_steps):
loss = 0
for i in range(total_batch):
batch_x, batch_y = train_dataset[epoch * batch_size:(epoch+1) * batch_size, :], train_labels[epoch * batch_size:(epoch+1) * batch_size,:]
_, c = sess.run([optimizer, loss], feed_dict={x: batch_x, y: batch_y})
loss = loss + c
loss = loss / total_batch
if epoch % 500 == 0:
print ("Epoch :", epoch, ". cost = {:.9f}".format(avg_cost))
print("Minibatch accuracy: %.1f%%" % accuracy(predictions, batch_labels))
valid_prediction = tf.run(tf_valid_dataset, {x: tf_valid_dataset})
print("Validation accuracy: %.1f%%" % accuracy(valid_prediction.eval(), valid_labels))
test_prediction = tf.run(tf_test_dataset, {x: tf_test_dataset})
print("TEST accuracy: %.1f%%" % accuracy(test_prediction.eval(), test_labels))
This worked for me
from keras import backend as K
and after predicting my data i inserted this part of code
then i had again loaded the model.
K.clear_session()
i faced this problem in production server,
but in my pc it was running fine
...........
from keras import backend as K
#Before prediction
K.clear_session()
#After prediction
K.clear_session()
Variable x is not in the same graph as model, try to define all of these in the same graph scope. For example,
# define a graph
graph1 = tf.Graph()
with graph1.as_default():
# placeholder
x = tf.placeholder(...)
y = tf.placeholder(...)
# create model
model = create(x, w, b)
with tf.Session(graph=graph1) as sess:
# initialize all the variables
sess.run(init)
# then feed_dict
# ......
If you use django server, just runserver with --nothreading
for example:
python manage.py runserver --nothreading
I had the same issue with flask. adding --without-threads flag to flask run or threaded=False to app.run() fixed it
In my case, I was using loop while calling in CNN multiple times, I fixed my problem by doing the following:
# Declare this as global:
global graph
graph = tf.get_default_graph()
# Then just before you call in your model, use this
with graph.as_default():
# call you models here
Note: In my case too, the app ran fine for the first time and then gave the error above. Using the above fix solved the problem.
Hope that helps.
The error message TypeError: Cannot interpret feed_dict key as Tensor: Tensor Tensor("...", dtype=dtype) is not an element of this graph can also arise in case you run a session outside of the scope of its with statement. Consider:
with tf.Session() as sess:
sess.run(logits, feed_dict=feed_dict)
sess.run(logits, feed_dict=feed_dict)
If logits and feed_dict are defined properly, the first sess.run command will execute normally, but the second will raise the mentioned error.
You can also experience this while working on notebooks hosted on online learning platforms like Coursera. So, implementing following code could help get over with the issue.
Implement this at the topmost block of Notebook file:
from keras import backend as K
K.clear_session()
Similar to #javan-peymanfard and #hmadali-shafiee, I ran into this issue when loading the model in an API. I was using FastAPI with uvicorn. To fix the issue I just set the API function definitions to async similar to this:
#app.post('/endpoint_name')
async def endpoint_function():
# Do stuff here, including possibly (re)loading the model
Using the MNIST tutorial of Tensorflow, I try to make a convolutional network for face recognition with the "Database of Faces".
The images size are 112x92, I use 3 more convolutional layer to reduce it to 6 x 5 as adviced here
I'm very new at convolutional network and most of my layer declaration is made by analogy to the Tensorflow MNIST tutorial, it may be a bit clumsy, so feel free to advice me on this.
x_image = tf.reshape(x, [-1, 112, 92, 1])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)
W_conv3 = weight_variable([5, 5, 64, 128])
b_conv3 = bias_variable([128])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
h_pool3 = max_pool_2x2(h_conv3)
W_conv4 = weight_variable([5, 5, 128, 256])
b_conv4 = bias_variable([256])
h_conv4 = tf.nn.relu(conv2d(h_pool3, W_conv4) + b_conv4)
h_pool4 = max_pool_2x2(h_conv4)
W_conv5 = weight_variable([5, 5, 256, 512])
b_conv5 = bias_variable([512])
h_conv5 = tf.nn.relu(conv2d(h_pool4, W_conv5) + b_conv5)
h_pool5 = max_pool_2x2(h_conv5)
W_fc1 = weight_variable([6 * 5 * 512, 1024])
b_fc1 = bias_variable([1024])
h_pool5_flat = tf.reshape(h_pool5, [-1, 6 * 5 * 512])
h_fc1 = tf.nn.relu(tf.matmul(h_pool5_flat, W_fc1) + b_fc1)
keep_prob = tf.placeholder("float")
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
print orlfaces.train.num_classes # 40
W_fc2 = weight_variable([1024, orlfaces.train.num_classes])
b_fc2 = bias_variable([orlfaces.train.num_classes])
y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)
My problem appear when the session run the "correct_prediction" op which is
tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
At least I think given the error message:
W tensorflow/core/common_runtime/executor.cc:1027] 0x19369d0 Compute status: Invalid argument: Incompatible shapes: [8] vs. [20]
[[Node: Equal = Equal[T=DT_INT64, _device="/job:localhost/replica:0/task:0/cpu:0"](ArgMax, ArgMax_1)]]
Traceback (most recent call last):
File "./convolutional.py", line 133, in <module>
train_accuracy = accuracy.eval(feed_dict = {x: batch[0], y_: batch[1], keep_prob: 1.0})
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 405, in eval
return _eval_using_default_session(self, feed_dict, self.graph, session)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2728, in _eval_using_default_session
return session.run(tensors, feed_dict)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 345, in run
results = self._do_run(target_list, unique_fetch_targets, feed_dict_string)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 419, in _do_run
e.code)
tensorflow.python.framework.errors.InvalidArgumentError: Incompatible shapes: [8] vs. [20]
[[Node: Equal = Equal[T=DT_INT64, _device="/job:localhost/replica:0/task:0/cpu:0"](ArgMax, ArgMax_1)]]
Caused by op u'Equal', defined at:
File "./convolutional.py", line 125, in <module>
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gen_math_ops.py", line 328, in equal
return _op_def_lib.apply_op("Equal", x=x, y=y, name=name)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/op_def_library.py", line 633, in apply_op
op_def=op_def)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 1710, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 988, in __init__
self._traceback = _extract_stack()
It looks like the y_conv output a matrix of shape 8 x batch_size instead of number_of_class x batch_size
If I change the batch size from 20 to 10, the error message stay the same but instead [8] vs. [20] I get [4] vs. [10]. So from that I conclude that the problem may come from the y_conv declaration (last line of the code above).
The loss function, optimizer, training, etc declarations is the same as in the MNIST tutorial:
cross_entropy = -tf.reduce_sum(y_ * tf.log(y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
sess.run((tf.initialize_all_variables()))
for i in xrange(1000):
batch = orlfaces.train.next_batch(20)
if i % 100 == 0:
train_accuracy = accuracy.eval(feed_dict = {x: batch[0], y_: batch[1], keep_prob: 1.0})
print "Step %d, training accuracy %g" % (i, train_accuracy)
train_step.run(feed_dict = {x: batch[0], y_: batch[1], keep_prob: 0.5})
print "Test accuracy %g" % accuracy.eval(feed_dict = {x: orlfaces.test.images, y_: orlfaces.test.labels, keep_prob: 1.0})
Thanks for reading, have a good day
Well, after a lot debugging, I found that my issue was due to a bad instantiation of the labels. Instead of creating arrays full of zeros and replace one value by one, I created them with random value! Stupid mistake. In case someone wondering what I did wrong there and how I fix it here is the change I made.
Anyway during all the debugging I made, to find this mistake, I found some useful information to debug this kind of problem:
For the cross entropy declaration, the tensorflow's MNIST tutorial use a formula that can lead to NaN value
This formula is
cross_entropy = -tf.reduce_sum(y_ * tf.log(y_conv))
Instead of this, I found two ways to declare it in a safer fashion:
cross_entropy = -tf.reduce_sum(y_ * tf.log(tf.clip_by_value(y_conv, 1e-10, 1.0)))
or also:
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logit, y_))
As mrry says. printing the shape of the tensors can help to detect shape anomaly.
To get the shape of a tensor just call his get_shape() method like this:
print "W shape:", W.get_shape()
user1111929 in this question use a debug print that help me assert where the problem come from.
After importing this data file from Matlab with scipy.io.loadmat, things appeared to work fine until we tried to calculate the conditioning number of one of the matrixes within.
Here's the minimum amount of code that reproduces for us:
import scipy
import numpy
stuff = scipy.io.loadmat("dati-esercizio1.mat")
numpy.linalg.cond(stuff["A"])
Here's the extended stacktrace courtesy of iPython:
In [3]: numpy.linalg.cond(A)
---------------------------------------------------------------------------
LapackError Traceback (most recent call last)
/snip/<ipython-input-3-15d9ef00a605> in <module>()
----> 1 numpy.linalg.cond(A)
/snip/python2.7/site-packages/numpy/linalg/linalg.py in cond(x, p)
1409 x = asarray(x) # in case we have a matrix
1410 if p is None:
-> 1411 s = svd(x,compute_uv=False)
1412 return s[0]/s[-1]
1413 else:
/snip/python2.7/site-packages/numpy/linalg/linalg.py in svd(a, full_matrices, compute_uv)
1313 work = zeros((lwork,), t)
1314 results = lapack_routine(option, m, n, a, m, s, u, m, vt, nvt,
-> 1315 work, -1, iwork, 0)
1316 lwork = int(work[0])
1317 work = zeros((lwork,), t)
LapackError: Parameter a has non-native byte order in lapack_lite.dgesdd
All obvious ideas (like flattening and reshaping the matrix or recreating the matrix from scratch reassigning it element by element) failed. How can I want to massage the data, then, in order to make it more agreeable with numpy?
It's a bug, fixed some time ago: https://github.com/numpy/numpy/pull/235
Workaround:
np.linalg.cond(stuff['A'].newbyteorder('='))
This works for me:
In [33]: stuff = loadmat('dati-esercizio1.mat')
In [34]: a = stuff['A']
In [35]: try: np.linalg.cond(a)
....: except: print "Fail!"
Fail!
In [36]: b = np.array(a, dtype='>d')
In [37]: np.linalg.cond(b)
Out[37]: 62493201976.673141
In [38]: np.all(a == b) # Verify they hold the same data.
Out[38]: True
Apparently it's something wrong with the byte order (endianness?) of each number in the resulting ndarray and not just with the ndarray object itself.
Something like this but more elegant should do the trick:
n, m = A.shape()
B = numpy.empty_like(A)
for i in xrange(n):
for j in xrange(m):
B[i,j] = float(A[i,j])
del A
B = A
print numpy.linalg.cond(A) # 62493210091.354507
(For some reason an in-place replacement still gives that error - so there's something wrong with the byte order of the whole object, too.)