comparison keras_deep_learning.py @ 2:38c4f8a98038 draft

"planemo upload for repository https://github.com/bgruening/galaxytools/tree/master/tools/sklearn commit 5b2ac730ec6d3b762faa9034eddd19ad1b347476"

1:c1b0c8232816

import six
import warnings

from ast import literal_eval
from keras.models import Sequential, Model
from galaxy_ml.utils import try_get_attr, get_search_params


def _handle_shape(literal):
    """Eval integer or list/tuple of integers from string

            "constraint_type":
                "MinMaxNorm"
            }
    """
    constraint_type = config['constraint_type']
    if constraint_type == 'None':
        return None

    klass = getattr(keras.constraints, constraint_type)
    options = config.get('constraint_options', {})
    if 'axis' in options:

def _handle_layer_parameters(params):
    """Access to handle all kinds of parameters
    """
    for key, value in six.iteritems(params):
        if value == 'None':
            params[key] = None
            continue

        if type(value) in [int, float, bool]\
                or (type(value) is str and value.isalpha()):
            continue

        if key in ['input_shape', 'noise_shape', 'shape', 'batch_shape',
                   'target_shape', 'dims', 'kernel_size', 'strides',
                   'dilation_rate', 'output_padding', 'cropping', 'size',
                   'padding', 'pool_size', 'axis', 'shared_axes']:
            params[key] = _handle_shape(value)

        elif key.endswith('_regularizer'):
            params[key] = _handle_regularizer(value)

        elif key.endswith('_constraint'):
            params[key] = _handle_constraint(value)

        elif key == 'function':  # No support for lambda/function eval
            params.pop(key)

    layers = config['layers']
    for layer in layers:
        options = layer['layer_selection']
        layer_type = options.pop('layer_type')
        klass = getattr(keras.layers, layer_type)
        other_options = options.pop('layer_options', {})
        options.update(other_options)

        # parameters needs special care
        options = _handle_layer_parameters(options)

        # add input_shape to the first layer only
        if not getattr(model, '_layers') and input_shape is not None:
            options['input_shape'] = input_shape

    for layer in layers:
        options = layer['layer_selection']
        layer_type = options.pop('layer_type')
        klass = getattr(keras.layers, layer_type)
        inbound_nodes = options.pop('inbound_nodes', None)
        other_options = options.pop('layer_options', {})
        options.update(other_options)

        # parameters needs special care
        options = _handle_layer_parameters(options)
        # merge layers
        if 'merging_layers' in options:
            idxs = literal_eval(options.pop('merging_layers'))
            merging_layers = [all_layers[i-1] for i in idxs]
            new_layer = klass(**options)(merging_layers)

    Parameters
    -----------
    config : dictionary, galaxy tool parameters loaded by JSON
    """
    generator_type = config.pop('generator_type')
    klass = try_get_attr('galaxy_ml.preprocessors', generator_type)

    if generator_type == 'GenomicIntervalBatchGenerator':
        config['ref_genome_path'] = 'to_be_determined'
        config['intervals_path'] = 'to_be_determined'

        model = get_functional_model(layers_config)

    json_string = model.to_json()

    with open(outfile, 'w') as f:
        f.write(json_string)


def build_keras_model(inputs, outfile, model_json, infile_weights=None,
                      batch_mode=False, outfile_params=None):
    """ for `keras_model_builder` tool

2:38c4f8a98038

import six
import warnings

from ast import literal_eval
from keras.models import Sequential, Model
from galaxy_ml.utils import try_get_attr, get_search_params, SafeEval


safe_eval = SafeEval()


def _handle_shape(literal):
    """Eval integer or list/tuple of integers from string

            "constraint_type":
                "MinMaxNorm"
            }
    """
    constraint_type = config['constraint_type']
    if constraint_type in ('None', ''):
        return None

    klass = getattr(keras.constraints, constraint_type)
    options = config.get('constraint_options', {})
    if 'axis' in options:

def _handle_layer_parameters(params):
    """Access to handle all kinds of parameters
    """
    for key, value in six.iteritems(params):
        if value in ('None', ''):
            params[key] = None
            continue

        if type(value) in [int, float, bool]\
                or (type(value) is str and value.isalpha()):
            continue

        if key in ['input_shape', 'noise_shape', 'shape', 'batch_shape',
                   'target_shape', 'dims', 'kernel_size', 'strides',
                   'dilation_rate', 'output_padding', 'cropping', 'size',
                   'padding', 'pool_size', 'axis', 'shared_axes'] \
                and isinstance(value, str):
            params[key] = _handle_shape(value)

        elif key.endswith('_regularizer') and isinstance(value, dict):
            params[key] = _handle_regularizer(value)

        elif key.endswith('_constraint') and isinstance(value, dict):
            params[key] = _handle_constraint(value)

        elif key == 'function':  # No support for lambda/function eval
            params.pop(key)

    layers = config['layers']
    for layer in layers:
        options = layer['layer_selection']
        layer_type = options.pop('layer_type')
        klass = getattr(keras.layers, layer_type)
        kwargs = options.pop('kwargs', '')

        # parameters needs special care
        options = _handle_layer_parameters(options)

        if kwargs:
            kwargs = safe_eval('dict(' + kwargs + ')')
            options.update(kwargs)

        # add input_shape to the first layer only
        if not getattr(model, '_layers') and input_shape is not None:
            options['input_shape'] = input_shape

    for layer in layers:
        options = layer['layer_selection']
        layer_type = options.pop('layer_type')
        klass = getattr(keras.layers, layer_type)
        inbound_nodes = options.pop('inbound_nodes', None)
        kwargs = options.pop('kwargs', '')

        # parameters needs special care
        options = _handle_layer_parameters(options)

        if kwargs:
            kwargs = safe_eval('dict(' + kwargs + ')')
            options.update(kwargs)

        # merge layers
        if 'merging_layers' in options:
            idxs = literal_eval(options.pop('merging_layers'))
            merging_layers = [all_layers[i-1] for i in idxs]
            new_layer = klass(**options)(merging_layers)

    Parameters
    -----------
    config : dictionary, galaxy tool parameters loaded by JSON
    """
    generator_type = config.pop('generator_type')
    if generator_type == 'none':
        return None

    klass = try_get_attr('galaxy_ml.preprocessors', generator_type)

    if generator_type == 'GenomicIntervalBatchGenerator':
        config['ref_genome_path'] = 'to_be_determined'
        config['intervals_path'] = 'to_be_determined'

        model = get_functional_model(layers_config)

    json_string = model.to_json()

    with open(outfile, 'w') as f:
        json.dump(json.loads(json_string), f, indent=2)


def build_keras_model(inputs, outfile, model_json, infile_weights=None,
                      batch_mode=False, outfile_params=None):
    """ for `keras_model_builder` tool