def compile(self, inputs):
xmodel = ch2o.compile_model(self.mc, inputs)
f = tempfile.NamedTemporaryFile(delete=False)
f.write(xmodel.SerializeToString())
f.close()
del xmodel
graph = chainer_compiler_core.load(f.name)
os.unlink(f.name)
self.orig_output_names = graph.output_names()
fwd_graph, bwd_graph = graph.backward_to(graph.input_names())
if self.dump_onnx:
sys.stderr.write('=== vvv forward vvv ===\n' + fwd_graph.dump() +
'\n=== ^^^ forward ^^^ ===\n')
sys.stderr.write('=== vvv backward vvv ===\n' + bwd_graph.dump() +
'\n=== ^^^ backward ^^^ ===\n')
assert graph.input_names() == fwd_graph.input_names()
self.fwd_input_names = fwd_graph.input_names()
self.fwd_output_names = fwd_graph.output_names()
self.bwd_input_names = bwd_graph.input_names()
self.bwd_output_names = bwd_graph.output_names()
# TODO(hamaji): Revive shape inference.
self.fwd = fwd_graph.compile(skip_inference=True)
self.bwd = bwd_graph.compile(skip_inference=True)
self.param_names = self.fwd_input_names[len(inputs):]
self.compiled = True
def create_backprop_test(test_name, model, input_values):
chainer.config.train = True
model.cleargrads()
output_values = model(*map(chainer.variable.Variable, input_values))
test_dir = 'out/backprop_test_pc_%s' % test_name
test_data_set_dir = os.path.join(test_dir, 'test_data_set_0')
os.makedirs(test_data_set_dir, exist_ok=True)
xmodel = ch2o.compile_model(model, input_values)
all_input_tensors = xmodel.graph.input
output_tensors = xmodel.graph.output
if not isinstance(output_values, (list, tuple)):
output_values = (output_values, )
for output_value in output_values:
output_value.grad = np.ones(output_value.shape, output_value.dtype)
output_value.backward()
ch2o.testcasegen.edit_onnx_protobuf(xmodel, model)
initializer_names = set()
for initializer in xmodel.graph.initializer:
initializer_names.add(initializer.name)
input_tensors = []
for input_tensor in all_input_tensors:
if input_tensor.name not in initializer_names:
input_tensors.append(input_tensor)
assert len(input_tensors) == len(input_values)
assert len(output_tensors) == len(output_values)
outputs = []
for tensor, value in zip(output_tensors, output_values):
outputs.append((tensor, value.array))
for name, param in sorted(model.namedparams()):
bp_name = onnx.helper.make_tensor_value_info('grad_out@' + name,
onnx.TensorProto.FLOAT,
())
outputs.append((bp_name, param.grad))
ch2o.testcasegen.dump_test_inputs_outputs(
list(zip(input_tensors, input_values)), outputs, test_data_set_dir)
with open(os.path.join(test_dir, 'model.onnx'), 'wb') as fp:
fp.write(xmodel.SerializeToString())
def compile(self, inputs):
if self.translator == 'ch2o':
xmodel = ch2o.compile_model(self.mc, inputs)
f = tempfile.NamedTemporaryFile(delete=False)
f.write(xmodel.SerializeToString())
f.close()
del xmodel
elif self.translator == 'onnx_chainer':
import onnx_chainer
f = tempfile.NamedTemporaryFile(delete=False)
onnx_chainer.export(self.mc, inputs, filename=f)
f.close()
else:
raise NotImplementedError('Unsupported translator:',
self.translator)
graph = chainer_compiler_core.load(f.name)
os.unlink(f.name)
self.orig_output_names = graph.output_names()
if self.computation_order is None:
fwd_graph, bwd_graph = graph.backward_to(graph.input_names() +
graph.param_names())
else:
fwd_graph, bwd_graph = graph.backward_to_with_order(
self.computation_order)
if self.dump_onnx:
sys.stderr.write('=== vvv forward vvv ===\n' + fwd_graph.dump() +
'\n=== ^^^ forward ^^^ ===\n')
sys.stderr.write('=== vvv backward vvv ===\n' + bwd_graph.dump() +
'\n=== ^^^ backward ^^^ ===\n')
assert graph.input_names() == fwd_graph.input_names()
self.fwd_input_names = fwd_graph.input_names()
self.fwd_output_names = fwd_graph.output_names()
self.bwd_input_names = bwd_graph.input_names()
self.bwd_output_names = bwd_graph.output_names()
# TODO(hamaji): Revive shape inference.
self.fwd = fwd_graph.compile(skip_inference=True)
self.bwd = bwd_graph.compile(skip_inference=True)
self.param_names = fwd_graph.param_names()
self.compiled = True