def test_chain_right_branch(model1, model2, model3):
# Previously we 'flattened' these nested calls. We might opt to do so
# again, especially for the operators.
merge1 = chain(model1, model2)
merge2 = chain(merge1, model3)
assert len(merge1.layers) == 2
assert len(merge2.layers) == 2
def test_chain(ops):
data = numpy.asarray([[1, 2, 3, 4]], dtype="f")
model = chain(Linear(1), Dropout(), Linear(1))
model.ops = ops
model.initialize(data, data)
Y, backprop = model(data, is_train=True)
backprop(Y)
# Layers with and without nO/nI
model = chain(Linear(1), Dropout(), Linear(1, 1))
model.initialize(data, data)
# Setting dim on model
model = chain(Linear(1), Dropout(), Linear(1))
model.set_dim("nO", 1)
model.initialize(data, None)
model = chain(Linear(1, 1), Dropout(), Linear(1, 1))
model.set_dim("nI", 1)
model.initialize(None, data)
# Not enough arguments
with pytest.raises(TypeError):
chain(Linear())
with pytest.raises(TypeError):
chain()
def test_chain_two(model1, model2):
model = chain(model1, model2)
assert len(model.layers) == 2
def test_chain_one(model1):
with pytest.raises(TypeError):
chain(model1)
def test_chain_zero():
with pytest.raises(TypeError):
chain()
def test_chain_three(model1, model2, model3):
model = chain(model1, model2, model3)
assert len(model.layers) == 3
from typing import List
from thinc.layers import Relu, Softmax, chain, reduce_max, concatenate
from thinc.model import Model
# Define Custom X/Y types
MyModelX = List[List[float]]
MyModelY = List[List[float]]
model: Model[MyModelX, MyModelY] = chain(
Relu(12),
Relu(12, dropout=0.2),
Softmax(),
)
# ERROR: incompatible type "bool", expected "List[List[float]]"
model(False)
# ERROR: List item 0 has incompatible type "str"; expected "float"
model.begin_update([["0"]])
# ERROR: incompatible type "bool", expected "List[List[float]]"
model.predict(True)
# This example should be run with mypy. This is an example of type-level checking
# for network validity.
#
# We first define an invalid network.
# It's invalid because reduce_max expects Array3d as input, while Relu produces
# Array2d as output. chain has type-logic to verify input and output types
# line up.
#
# You should see the error an error,
# examples/howto/type_chain.py:10: error: Cannot infer type argument 2 of "chain"
bad_model = chain(Relu(10), reduce_max(), Softmax())
