def after_run(self):
super(InferenceTask, self).after_run()
# retrieve inference data
inputs = None
visualizations = []
outputs = {}
if self.inference_data_filename is not None:
# the HDF5 database contains:
# - input images, in a dataset "/inputs"
# - all network outputs, in a group "/outputs/"
# - layer activations and weights, if requested, in a group "/layers/"
db = h5py.File(self.inference_data_filename, 'r')
# collect inputs
inputs = db['inputs'][...]
# collect outputs
for output_key, output_data in db['outputs'].items():
output_name = base64.urlsafe_b64decode(str(output_key))
outputs[output_name] = output_data[...]
# collect layer data, if applicable
if 'layers' in db.keys():
for layer_id, layer in db['layers'].items():
visualization = {
'id': int(layer_id),
'name': layer.attrs['name'],
'vis_type': layer.attrs['vis_type'],
'data_stats': {
'shape': layer.attrs['shape'],
'mean': layer.attrs['mean'],
'stddev': layer.attrs['stddev'],
'histogram': [
layer.attrs['histogram_y'].tolist(),
layer.attrs['histogram_x'].tolist(),
layer.attrs['histogram_ticks'].tolist(),
]
}
}
if 'param_count' in layer.attrs:
visualization['param_count'] = layer.attrs['param_count']
if 'layer_type' in layer.attrs:
visualization['layer_type'] = layer.attrs['layer_type']
vis = layer[...]
if vis.shape[0] > 0:
visualization['image_html'] = embed_image_html(vis)
visualizations.append(visualization)
# sort by layer ID (as HDF5 ASCII sorts)
visualizations = sorted(visualizations,key=lambda x:x['id'])
db.close()
self.inference_log.close()
# save inference to data for further use
self.inference_inputs = inputs
self.inference_outputs = outputs
self.inference_layers = visualizations
def after_run(self):
super(InferenceTask, self).after_run()
# retrieve inference data
visualizations = []
outputs = OrderedDict()
if self.inference_data_filename is not None:
# the HDF5 database contains:
# - input images, in a dataset "/inputs"
# - all network outputs, in a group "/outputs/"
# - layer activations and weights, if requested, in a group "/layers/"
db = h5py.File(self.inference_data_filename, 'r')
# collect paths and data
input_ids = db['input_ids'][...]
input_data = db['input_data'][...]
# collect outputs
o = []
for output_key, output_data in db['outputs'].items():
output_name = base64.urlsafe_b64decode(str(output_key))
o.append({
'id': output_data.attrs['id'],
'name': output_name,
'data': output_data[...]
})
# sort outputs by ID
o = sorted(o, key=lambda x: x['id'])
# retain only data (using name as key)
for output in o:
outputs[output['name']] = output['data']
# collect layer data, if applicable
if 'layers' in db.keys():
for layer_id, layer in db['layers'].items():
visualization = {
'id': int(layer_id),
'name': layer.attrs['name'],
'vis_type': layer.attrs['vis_type'],
'data_stats': {
'shape':
layer.attrs['shape'],
'mean':
layer.attrs['mean'],
'stddev':
layer.attrs['stddev'],
'histogram': [
layer.attrs['histogram_y'].tolist(),
layer.attrs['histogram_x'].tolist(),
layer.attrs['histogram_ticks'].tolist(),
]
}
}
if 'param_count' in layer.attrs:
visualization['param_count'] = layer.attrs[
'param_count']
if 'layer_type' in layer.attrs:
visualization['layer_type'] = layer.attrs['layer_type']
vis = layer[...]
if vis.shape[0] > 0:
visualization['image_html'] = embed_image_html(vis)
visualizations.append(visualization)
# sort by layer ID (as HDF5 ASCII sorts)
visualizations = sorted(visualizations, key=lambda x: x['id'])
db.close()
# save inference data for further use
self.inference_inputs = {'ids': input_ids, 'data': input_data}
self.inference_outputs = outputs
self.inference_layers = visualizations
self.inference_log.close()