Euler Sample Path does not work under TensorFlow 2.0 with Higher Dimension

[jjchan121]

When we specify the SDE with dimension greater than one and simulate under Euler Scheme, we will have the following errors.

~/anaconda/envs/py36dev/lib/python3.6/site-packages/tf_quant_finance/models/ito_process.py in step_fn(i, written_count, current_state, result)
332 dw = random.mv_normal_sample(
333 (num_samples,), mean=wiener_mean, random_type=random_type,
--> 334 seed=seed)
335 dw = dw * sqrt_dt[i]
336 dt_inc = dt[i] * self.drift_fn()(current_time, current_state) # pylint: disable=not-callable

~/anaconda/envs/py36dev/lib/python3.6/site-packages/tf_quant_finance/math/random_ops/multivariate_normal.py in multivariate_normal(sample_shape, mean, covariance_matrix, scale_matrix, random_type, validate_args, seed, dtype, name, **kwargs)
170 name,
171 default_name='multivariate_normal',
--> 172 values=[sample_shape, mean, covariance_matrix, scale_matrix]):
173 if mean is not None:
174 mean = tf.convert_to_tensor(mean, dtype=dtype, name='mean')

TypeError: init() got an unexpected keyword argument 'default_name'

I'm using Tensorflow 2.0

[saxena-ashish-g]

I have pushed a fix for this (we needed to replace tf.name_scope with tf.compat.v1.name_scope). Can you please give it another spin (against the head version and not the pip version) and let me know if it still doesn't work.

[jjchan121]

Even worse, I can't import the library.

----> 1 import tf_quant_finance as tqf

~/.conda/envs/idp36p/lib/python3.6/site-packages/tf_quant_finance/init.py in
66
67 from tf_quant_finance import black_scholes
---> 68 from tf_quant_finance import math
69 from tf_quant_finance import models
70 from tf_quant_finance import rates

~/.conda/envs/idp36p/lib/python3.6/site-packages/tf_quant_finance/math/init.py in
23 from tf_quant_finance.math import pde
24 from tf_quant_finance.math import piecewise
---> 25 from tf_quant_finance.math import random_ops as random
26 from tf_quant_finance.math import root_search
27 from tf_quant_finance.math import segment_ops

~/.conda/envs/idp36p/lib/python3.6/site-packages/tf_quant_finance/math/random_ops/init.py in
19
20 from tf_quant_finance.math.random_ops import halton
---> 21 from tf_quant_finance.math.random_ops import sobol
22 from tf_quant_finance.math.random_ops.multivariate_normal import multivariate_normal as mv_normal_sample
23 from tf_quant_finance.math.random_ops.multivariate_normal import RandomType

~/.conda/envs/idp36p/lib/python3.6/site-packages/tf_quant_finance/math/random_ops/sobol/init.py in
18 from future import print_function
19
---> 20 from tf_quant_finance.math.random_ops.sobol.sobol_impl import sample
21
22 from tensorflow.python.util.all_util import remove_undocumented # pylint: disable=g-direct-tensorflow-import

~/.conda/envs/idp36p/lib/python3.6/site-packages/tf_quant_finance/math/random_ops/sobol/sobol_impl.py in
223
224 (_PRIMITIVE_POLYNOMIAL_COEFFICIENTS,
--> 225 _INITIAL_DIRECTION_NUMBERS) = _load_sobol_data()

~/.conda/envs/idp36p/lib/python3.6/site-packages/tf_quant_finance/math/random_ops/sobol/sobol_impl.py in _load_sobol_data()
207 direction_numbers = np.zeros(shape=(18, 21200), dtype=np.int64)
208 index = 0
--> 209 with open(path) as f:
210 for line in f:
211 # Skip first line (header).

[matejr]

Hi @jjchan121,

we tried to reproduce your last error in an environment similar to yours (ubuntu, anaconda, latest pip package installed within anaconda). However, the "import tf_quant_finance as tqf" statement works in this environment.

Could you please provide more details about your environment? We would need the following information:

• base operating system (e.g., ubuntu)
• anaconda, python, pip and tensorflow versions
• tf-quant-finance version (pip package version)
• how did you install tf-quant-finance package (e.g., using pip install tf-quant-finance or did you build the pip package yourself)?
• if you've build the pip package yourself: how did you build it?

Ideally, we would get a sequence of commands that can completely reproduce the environment in which the "import tf_quant_finance as tqf" statement fails.

As a temporary workaround, you could just comment out the line 224 (line with _load_sobol_data()) in ~/.conda/envs/idp36p/lib/python3.6/site-packages/tf_quant_finance/math/random_ops/sobol/sobol_impl.py

In this case, Sobol number generator won't work, but you can use other types of random numbers (e.g., Halton numbers or pseudorandom numbers).

[cyrilchim]

All should be working now with TF 2.0