github/concrete
1
1
Fork 0
mirror of https://github.com/zama-ai/concrete.git synced 2026-02-09 03:55:04 -05:00
Code Issues Packages Projects Releases Wiki Activity

feat: remove transpose from layers

Browse Source
This commit is contained in:
jfrery
2021-11-17 12:19:40 +01:00
committed by jfrery
parent b9c52947f6
commit c978107124
3 changed files with 13 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
concrete/quantization/quantized_layers.py
View File

@@ -29,7 +29,7 @@ class QuantizedLinear:
self.n_bits = n_bits
if self.q_bias is None:
self.q_bias = QuantizedArray(n_bits, numpy.zeros(self.q_weights.values.shape[:-1]))
self.q_bias = QuantizedArray(n_bits, numpy.zeros(self.q_weights.values.shape[-1]))
self.q_out = None
def calibrate(self, x: numpy.ndarray):
@@ -39,7 +39,7 @@ class QuantizedLinear:
x (numpy.ndarray): Inputs.
"""
assert self.q_bias is not None
self.q_out = QuantizedArray(self.n_bits, x @ self.q_weights.values.T + self.q_bias.values)
self.q_out = QuantizedArray(self.n_bits, (x @ self.q_weights.values) + self.q_bias.values)
def __call__(self, q_input: QuantizedArray) -> QuantizedArray:
"""Process the forward pass of the quantized linear layer.
@@ -68,13 +68,11 @@ class QuantizedLinear:
p = self.q_weights.qvalues.shape[0]
# Core matmul operation in full intergers with a shape change (INTEGERS)
matmul = q_input.qvalues @ self.q_weights.qvalues.T
matmul = q_input.qvalues @ self.q_weights.qvalues
# Sum operation in full integers resulting in large integers (INTEGERS)
sum_input = self.q_weights.zero_point * numpy.sum(q_input.qvalues, axis=1, keepdims=True)
sum_weights = q_input.zero_point * numpy.sum(
self.q_weights.qvalues.T, axis=0, keepdims=True
)
sum_weights = q_input.zero_point * numpy.sum(self.q_weights.qvalues, axis=0, keepdims=True)
# Quantization scales and zero points (FLOATS involved)
# This is going to be compiled with a PBS (along with the following activation function)

12
concrete/torch/numpy_module.py
View File

@@ -12,10 +12,10 @@ class NumpyModule:
"""Initialize our numpy module.
Current constraint: All objects used in the forward have to be defined in the
__init__() of torch.nn.Module and follow the exact same order.
(i.e. each linear layer must have one variable defined in the
right order). This constraint will disappear when
TorchScript is in place. (issue #818)
__init__() of torch.nn.Module and follow the exact same order.
(i.e. each linear layer must have one variable defined in the
right order). This constraint will disappear when
TorchScript is in place. (issue #818)
Args:
torch_model (nn.Module): A fully trained, torch model alond with its parameters.
@@ -43,7 +43,7 @@ class NumpyModule:
for name, weights in self.torch_model.state_dict().items():
params = weights.detach().numpy()
self.numpy_module_dict[name] = params
self.numpy_module_dict[name] = params.T if "weight" in name else params
def __call__(self, x: numpy.ndarray):
"""Return the function to be compiled by concretefhe.numpy."""
@@ -64,7 +64,7 @@ class NumpyModule:
if isinstance(layer, nn.Linear):
# Apply a matmul product and add the bias.
x = (
x @ self.numpy_module_dict[f"{name}.weight"].T
x @ self.numpy_module_dict[f"{name}.weight"]
+ self.numpy_module_dict[f"{name}.bias"]
)
elif isinstance(layer, nn.Sigmoid):

6
tests/quantization/test_quantized_layers.py
View File

@@ -30,11 +30,11 @@ def test_quantized_linear(n_examples, n_features, n_neurons, n_bits):
inputs = numpy.random.uniform(size=(n_examples, n_features))
q_inputs = QuantizedArray(n_bits, inputs)
# shape of weights: (n_examples, n_features, n_neurons)
weights = numpy.random.uniform(size=(n_neurons, n_features))
# shape of weights: (n_neurons, n_features)
weights = numpy.random.uniform(size=(n_features, n_neurons))
q_weights = QuantizedArray(n_bits, weights)
bias = numpy.random.uniform(size=(n_neurons))
bias = numpy.random.uniform(size=(1, n_neurons))
q_bias = QuantizedArray(n_bits, bias)
# Define our QuantizedLinear layer