def test_shared_cache(data_setup):
train, _ = data_setup
shared_cache = {}
main_chain = SharedChain(chain_first(), shared_cache)
other_chain = SharedChain(chain_first(), shared_cache)
other_chain.fit(train)
# test cache is shared
assert isinstance(main_chain.root_node.cache, SharedCache)
# test cache is actual
assert main_chain.root_node.cache.actual_cached_state is not None
saved_model = main_chain.root_node.cache.actual_cached_state
main_chain.root_node.cache.clear()
# test cache is still actual despite the clearing of local cache
assert main_chain.root_node.cache.actual_cached_state is not None
shared_cache.clear()
# test cache is not actual after clearing shared cache
assert main_chain.root_node.cache.actual_cached_state is None
main_chain.root_node.cache.append(saved_model)
# test cache is actual after manual appending of model
assert main_chain.root_node.cache.actual_cached_state is not None
assert shared_cache[main_chain.root_node.descriptive_id] == saved_model
def test_chain_sharing_and_unsharing(data_setup):
chain = chain_first()
assert all([isinstance(node.cache, FittedModelCache) for node in chain.nodes])
chain = SharedChain(chain, {})
assert all([isinstance(node.cache, SharedCache) for node in chain.nodes])
chain = chain.unshare()
assert all([isinstance(node.cache, FittedModelCache) for node in chain.nodes])
assert isinstance(chain, Chain)
def metric_for_nodes(self, metric_function, train_data: InputData,
test_data: InputData, is_chain_shared: bool,
chain: Chain) -> float:
validate(chain)
if is_chain_shared:
chain = SharedChain(base_chain=chain,
shared_cache=self.shared_cache)
chain.fit(input_data=train_data)
return metric_function(chain, test_data)
def metric_for_nodes(self, metric_function, train_data: InputData,
test_data: InputData, is_chain_shared: bool,
chain: Chain) -> float:
try:
validate(chain)
if is_chain_shared:
chain = SharedChain(base_chain=chain,
shared_cache=self.shared_cache)
chain.fit(input_data=train_data)
return metric_function(chain, test_data)
except Exception as ex:
print(
f'Error in chain assessment during composition: {ex}. Continue.'
)
return max_int_value
def test_multi_chain_caching_with_shared_cache(data_setup):
train, _ = data_setup
shared_cache = {}
main_chain = SharedChain(base_chain=chain_second(), shared_cache=shared_cache)
other_chain = SharedChain(base_chain=chain_first(), shared_cache=shared_cache)
# fit other_chain that contains the parts identical to main_chain
other_chain.fit(input_data=train)
nodes_with_non_actual_cache = [main_chain.root_node, main_chain.root_node.nodes_from[0]] + \
[_ for _ in main_chain.root_node.nodes_from[0].nodes_from]
nodes_with_actual_cache = [node for node in main_chain.nodes if node not in nodes_with_non_actual_cache]
# check that using of SharedChain make identical of the main_chain fitted,
# despite the main_chain.fit() was not called
assert all([node.cache.actual_cached_state for node in nodes_with_actual_cache])
# the non-identical parts are still not fitted
assert not any([node.cache.actual_cached_state for node in nodes_with_non_actual_cache])
# check the same case with another chains
shared_cache = {}
main_chain = SharedChain(base_chain=chain_fourth(), shared_cache=shared_cache)
prev_chain_first = SharedChain(base_chain=chain_third(), shared_cache=shared_cache)
prev_chain_second = SharedChain(base_chain=chain_fifth(), shared_cache=shared_cache)
prev_chain_first.fit(input_data=train)
prev_chain_second.fit(input_data=train)
nodes_with_non_actual_cache = [main_chain.root_node, main_chain.root_node.nodes_from[1]]
nodes_with_actual_cache = [child for child in main_chain.root_node.nodes_from[0].nodes_from]
assert not any([node.cache.actual_cached_state for node in nodes_with_non_actual_cache])
assert all([node.cache.actual_cached_state for node in nodes_with_actual_cache])