From e851459f9a4b1ed406f1af0ac6d4a075983bf641 Mon Sep 17 00:00:00 2001
From: ahoni <aurelie.saulq@proton.me>
Date: Sun, 9 Mar 2025 14:36:19 +0100
Subject: [PATCH] add type to quantizer)
---
modneflib/modnef/modnef_torch/model.py | 14 ++++++
.../modnef_neurons/blif_model/blif.py | 38 ++++++----------
.../modnef_neurons/blif_model/rblif.py | 40 +++++++----------
.../modnef_neurons/modnef_torch_neuron.py | 9 ++--
.../modnef_neurons/slif_model/rslif.py | 45 +++++++------------
.../modnef_neurons/slif_model/slif.py | 43 +++++++-----------
.../modnef_neurons/srlif_model/rshiftlif.py | 41 +++++++----------
.../modnef_neurons/srlif_model/shiftlif.py | 39 ++++++----------
.../quantizer/dynamic_scale_quantizer.py | 20 +++++----
.../modnef/quantizer/fixed_point_quantizer.py | 20 +++++----
.../modnef/quantizer/min_max_quantizer.py | 20 +++++----
modneflib/modnef/quantizer/quantizer.py | 31 ++++++-------
modneflib/setup.py | 2 +-
13 files changed, 161 insertions(+), 201 deletions(-)
diff --git a/modneflib/modnef/modnef_torch/model.py b/modneflib/modnef/modnef_torch/model.py
index f35d486..0451bfe 100644
--- a/modneflib/modnef/modnef_torch/model.py
+++ b/modneflib/modnef/modnef_torch/model.py
@@ -91,6 +91,20 @@ class ModNEFModel(nn.Module):
m.set_quant(quant)
return super().train(mode=mode)
+
+ def quantize(self, force_init=False):
+ """
+ Quantize synaptic weight and neuron hyper-parameters
+
+ Parameters
+ ----------
+ force_init = Fasle : bool
+ force quantizer initialization
+ """
+
+ for m in self.modules():
+ if isinstance(m, ModNEFNeuron):
+ m.quantize(force_init=force_init)
def train(self, mode : bool = True, quant : bool = False):
"""
diff --git a/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/blif.py b/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/blif.py
index 78cac10..3c83a40 100644
--- a/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/blif.py
+++ b/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/blif.py
@@ -1,7 +1,7 @@
"""
File name: blif
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, math, snntorch, modnef.archbuilder, modnef_torch_neuron, modnef.quantizer
@@ -56,11 +56,11 @@ class BLIF(Leaky, ModNEFNeuron):
Update membrane voltage
get_builder_module(module_name, output_path=".")
Create ModNEFArchBuilder from internal neuron description
- quantize_weight(dtype=torch.int32)
+ quantize_weight()
Quantize synaptic weight
- quantize_parameters(dtype=torch.int32)
+ quantize_parameters()
Quantize neuron hyper-parameters
- quantize(dtype=torch.int32)
+ quantize(force_init=False)
Quantize synaptic weight and neuron hyper-parameters
hardware_estimation()
Toggle hardware estimation calculation
@@ -305,51 +305,41 @@ class BLIF(Leaky, ModNEFNeuron):
)
return module
- def quantize_weight(self, dtype=torch.int32):
+ def quantize_weight(self):
"""
Quantize synaptic weight
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight)
- self.fc.weight.data = self.quantizer(self.fc.weight.data, True, dtype)
+ self.fc.weight.data = self.quantizer(self.fc.weight.data, True)
- def quantize_parameters(self, dtype=torch.int32):
+ def quantize_parameters(self):
"""
Quantize neuron hyper-parameters
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight)
- self.threshold.data = self.quantizer(self.threshold.data, True, dtype)
- self.beta.data = self.quantizer(self.beta.data, True, dtype)
+ self.threshold.data = self.quantizer(self.threshold.data, True)
+ self.beta.data = self.quantizer(self.beta.data, True)
- def quantize(self, force_init=False, dtype=torch.int32):
+ def quantize(self, force_init=False):
"""
Quantize synaptic weight and neuron hyper-parameters
Parameters
----------
- dtype = torch.int32 : dtype
- type use during quantization
+ force_init = Fasle : bool
+ force quantizer initialization
"""
if force_init or not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight)
- self.quantize_weight(dtype)
- self.quantize_parameters(dtype)
+ self.quantize_weight()
+ self.quantize_parameters()
self.quantization_flag = True
@classmethod
diff --git a/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/rblif.py b/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/rblif.py
index 06b766e..f0ddcdd 100644
--- a/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/rblif.py
+++ b/modneflib/modnef/modnef_torch/modnef_neurons/blif_model/rblif.py
@@ -1,7 +1,7 @@
"""
File name: rblif
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, snntorch, modnef.archbuilder, modnef_torch_neuron, math, modnef.quantizer
@@ -58,11 +58,11 @@ class RBLIF(Leaky, ModNEFNeuron):
Update membrane voltage
get_builder_module(module_name, output_path=".")
Create ModNEFArchBuilder from internal neuron description
- quantize_weight(dtype=torch.int32)
+ quantize_weight()
Quantize synaptic weight
- quantize_parameters(dtype=torch.int32)
+ quantize_parameters()
Quantize neuron hyper-parameters
- quantize(dtype=torch.int32)
+ quantize(force_init=False)
Quantize synaptic weight and neuron hyper-parameters
hardware_estimation()
Toggle hardware estimation calculation
@@ -318,53 +318,43 @@ class RBLIF(Leaky, ModNEFNeuron):
)
return module
- def quantize_weight(self, dtype=torch.int32):
+ def quantize_weight(self):
"""
Quantize synaptic weight
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight, self.reccurent.weight)
- self.fc.weight.data = self.quantizer(self.fc.weight.data, True, dtype)
- self.reccurent.weight.data = self.quantizer(self.reccurent.weight.data, True, dtype)
+ self.fc.weight.data = self.quantizer(self.fc.weight.data, True)
+ self.reccurent.weight.data = self.quantizer(self.reccurent.weight.data, True)
- def quantize_parameters(self, dtype=torch.int32):
+ def quantize_parameters(self):
"""
Quantize neuron hyper-parameters
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight, self.reccurent.weight)
- self.threshold.data = self.quantizer(self.threshold.data, True, dtype)
- self.beta.data = self.quantizer(self.beta.data, True, dtype)
+ self.threshold.data = self.quantizer(self.threshold.data, True)
+ self.beta.data = self.quantizer(self.beta.data, True)
- def quantize(self, force_init=False, dtype=torch.int32):
+ def quantize(self, force_init=False):
"""
Quantize synaptic weight and neuron hyper-parameters
Parameters
----------
- dtype = torch.int32 : dtype
- type use during quantization
+ force_init = Fasle : bool
+ force quantizer initialization
"""
if force_init or not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight, self.reccurent.weight)
- self.quantize_weight(dtype)
- self.quantize_parameters(dtype)
+ self.quantize_weight()
+ self.quantize_parameters()
self.quantization_flag = True
@classmethod
diff --git a/modneflib/modnef/modnef_torch/modnef_neurons/modnef_torch_neuron.py b/modneflib/modnef/modnef_torch/modnef_neurons/modnef_torch_neuron.py
index 544c765..dc49183 100644
--- a/modneflib/modnef/modnef_torch/modnef_neurons/modnef_torch_neuron.py
+++ b/modneflib/modnef/modnef_torch/modnef_neurons/modnef_torch_neuron.py
@@ -79,23 +79,20 @@ class ModNEFNeuron():
raise NotImplementedError()
- def quantize(self, force_init=False, dtype=torch.int32):
+ def quantize(self, force_init=False):
"""
Quantize synaptic weight and neuron hyper-parameters
Parameters
----------
- dtype = torch.int32 : dtype
- type use during quantization
+ force_init = Fasle : bool
+ force quantizer initialization
"""
raise NotImplementedError()
def set_quant(self, mode=False):
self.quantization_flag = mode
-
- if mode:
- self.quantize(False)
def hardware_estimation(self, mode = False):
"""
diff --git a/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/rslif.py b/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/rslif.py
index 0f6415a..c950a76 100644
--- a/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/rslif.py
+++ b/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/rslif.py
@@ -1,7 +1,7 @@
"""
File name: rslif
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, snntorch, modnef.archbuilder, modnef_torch_neuron, math, modnef.quantizer
@@ -60,11 +60,11 @@ class RSLIF(LIF, ModNEFNeuron):
Update membrane voltage
get_builder_module(module_name, output_path=".")
Create ModNEFArchBuilder from internal neuron description
- quantize_weight(dtype=torch.int32)
+ quantize_weight()
Quantize synaptic weight
- quantize_parameters(dtype=torch.int32)
+ quantize_parameters(d)
Quantize neuron hyper-parameters
- quantize(dtype=torch.int32)
+ quantize(force_init=False)
Quantize synaptic weight and neuron hyper-parameters
hardware_estimation()
Toggle hardware estimation calculation
@@ -336,56 +336,45 @@ class RSLIF(LIF, ModNEFNeuron):
)
return module
- def quantize_weight(self, dtype=torch.int32):
+ def quantize_weight(self):
"""
Quantize synaptic weight
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight, rec_weight=self.reccurent.weight)
- self.fc.weight.data = self.quantizer(self.fc.weight.data, True, dtype)
- self.reccurent.weight.data = self.quantizer(self.reccurent.weight.data, True, dtype)
+ self.fc.weight.data = self.quantizer(self.fc.weight.data, True)
+ self.reccurent.weight.data = self.quantizer(self.reccurent.weight.data, True)
- def quantize_parameters(self, dtype=torch.int32):
+ def quantize_parameters(self):
"""
Quantize neuron hyper-parameters
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight, rec_weight=self.reccurent.weight)
- self.v_leak.data = self.quantizer(self.v_leak.data, True, dtype)
- self.v_min.data = self.quantizer(self.v_min.data, True, dtype)
- self.v_rest.data = self.quantizer(self.v_rest.data, True, dtype)
- self.threshold.data = self.quantizer(self.threshold, True, dtype)
+ self.v_leak.data = self.quantizer(self.v_leak.data, True)
+ self.v_min.data = self.quantizer(self.v_min.data, True)
+ self.v_rest.data = self.quantizer(self.v_rest.data, True)
+ self.threshold.data = self.quantizer(self.threshold, True)
- def quantize(self, force_init=False, dtype=torch.int32):
+ def quantize(self, force_init=False):
"""
Quantize synaptic weight and neuron hyper-parameters
Parameters
----------
- dtype = torch.int32 : dtype
- type use during quantization
+ force_init = Fasle : bool
+ force quantizer initialization
"""
if force_init or not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight, self.reccurent.weight)
- self.quantize_weight(dtype)
- self.quantize_parameters(dtype)
- self.quantization_flag = True
+ self.quantize_weight()
+ self.quantize_parameters()
@classmethod
def detach_hidden(cls):
diff --git a/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/slif.py b/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/slif.py
index 0bd22d6..315a07d 100644
--- a/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/slif.py
+++ b/modneflib/modnef/modnef_torch/modnef_neurons/slif_model/slif.py
@@ -1,7 +1,7 @@
"""
File name: slif
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, snntorch, modnef.archbuilder, modnef_torch_neuron, math, modnef.quantizer
@@ -60,11 +60,11 @@ class SLIF(LIF, ModNEFNeuron):
Update membrane voltage
get_builder_module(module_name, output_path=".")
Create ModNEFArchBuilder from internal neuron description
- quantize_weight(dtype=torch.int32)
+ quantize_weight()
Quantize synaptic weight
- quantize_parameters(dtype=torch.int32)
+ quantize_parameters()
Quantize neuron hyper-parameters
- quantize(dtype=torch.int32)
+ quantize(force_init=False)
Quantize synaptic weight and neuron hyper-parameters
hardware_estimation()
Toggle hardware estimation calculation
@@ -327,55 +327,44 @@ class SLIF(LIF, ModNEFNeuron):
)
return module
- def quantize_weight(self, dtype=torch.int32):
+ def quantize_weight(self):
"""
Quantize synaptic weight
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight)
- self.fc.weight.data = self.quantizer(self.fc.weight.data, True, dtype)
+ self.fc.weight.data = self.quantizer(self.fc.weight.data, True)
- def quantize_parameters(self, dtype=torch.int32):
+ def quantize_parameters(self):
"""
Quantize neuron hyper-parameters
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight)
- self.v_leak.data = self.quantizer(self.v_leak.data, True, dtype)
- self.v_min.data = self.quantizer(self.v_min.data, True, dtype)
- self.v_rest.data = self.quantizer(self.v_rest.data, True, dtype)
- self.threshold.data = self.quantizer(self.threshold, True, dtype)
+ self.v_leak.data = self.quantizer(self.v_leak.data, True)
+ self.v_min.data = self.quantizer(self.v_min.data, True)
+ self.v_rest.data = self.quantizer(self.v_rest.data, True)
+ self.threshold.data = self.quantizer(self.threshold, True)
- def quantize(self, force_init=False, dtype=torch.int32):
+ def quantize(self, force_init=False):
"""
Quantize synaptic weight and neuron hyper-parameters
Parameters
----------
- dtype = torch.int32 : dtype
- type use during quantization
+ force_init = Fasle : bool
+ force quantizer initialization
"""
if force_init or not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight)
- self.quantize_weight(dtype)
- self.quantize_parameters(dtype)
- self.quantization_flag = True
+ self.quantize_weight()
+ self.quantize_parameters()
@classmethod
diff --git a/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/rshiftlif.py b/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/rshiftlif.py
index 692a5f1..71194ac 100644
--- a/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/rshiftlif.py
+++ b/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/rshiftlif.py
@@ -1,7 +1,7 @@
"""
File name: rsrlif
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, snntorch, modnef.archbuilder, modnef_torch_neuron, math, modnef.quantizer
@@ -59,11 +59,11 @@ class RShiftLIF(LIF, ModNEFNeuron):
Update membrane voltage
get_builder_module(module_name, output_path=".")
Create ModNEFArchBuilder from internal neuron description
- quantize_weight(dtype=torch.int32)
+ quantize_weight()
Quantize synaptic weight
- quantize_parameters(dtype=torch.int32)
+ quantize_parameters()
Quantize neuron hyper-parameters
- quantize(dtype=torch.int32)
+ quantize(force_init=False)
Quantize synaptic weight and neuron hyper-parameters
hardware_estimation()
Toggle hardware estimation calculation
@@ -336,54 +336,43 @@ class RShiftLIF(LIF, ModNEFNeuron):
)
return module
- def quantize_weight(self, dtype=torch.int32):
+ def quantize_weight(self):
"""
Quantize synaptic weight
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight, rec_weight=self.reccurent.weight)
- self.fc.weight.data = self.quantizer(self.fc.weight.data, True, dtype)
- self.reccurent.weight.data = self.quantizer(self.reccurent.weight.data, True, dtype)
+ self.fc.weight.data = self.quantizer(self.fc.weight.data, True)
+ self.reccurent.weight.data = self.quantizer(self.reccurent.weight.data, True)
- def quantize_parameters(self, dtype=torch.int32):
+ def quantize_parameters(self):
"""
Quantize neuron hyper-parameters
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight, rec_weight=self.reccurent.weight)
- self.threshold.data = self.quantizer(self.threshold.data, True, dtype)
- self.beta.data = self.quantizer(self.beta.data, True, dtype)
+ self.threshold.data = self.quantizer(self.threshold.data, True)
+ self.beta.data = self.quantizer(self.beta.data, True)
- def quantize(self, force_init=False, dtype=torch.int32):
+ def quantize(self, force_init=False):
"""
Quantize synaptic weight and neuron hyper-parameters
Parameters
----------
- dtype = torch.int32 : dtype
- type use during quantization
+ force_init = Fasle : bool
+ force quantizer initialization
"""
if force_init or not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight, self.reccurent.weight)
- self.quantize_weight(dtype)
- self.quantize_parameters(dtype)
- self.quantization_flag = True
+ self.quantize_weight()
+ self.quantize_parameters()
@classmethod
def detach_hidden(cls):
diff --git a/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/shiftlif.py b/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/shiftlif.py
index af97a70..adcd3d4 100644
--- a/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/shiftlif.py
+++ b/modneflib/modnef/modnef_torch/modnef_neurons/srlif_model/shiftlif.py
@@ -1,7 +1,7 @@
"""
File name: srlif
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, snntorch, modnef.archbuilder, modnef_torch_neuron, math, modnef.quantizer
@@ -56,11 +56,11 @@ class ShiftLIF(LIF, ModNEFNeuron):
Update membrane voltage
get_builder_module(module_name, output_path=".")
Create ModNEFArchBuilder from internal neuron description
- quantize_weight(dtype=torch.int32)
+ quantize_weight()
Quantize synaptic weight
- quantize_parameters(dtype=torch.int32)
+ quantize_parameters()
Quantize neuron hyper-parameters
- quantize(dtype=torch.int32)
+ quantize(force_init=False)
Quantize synaptic weight and neuron hyper-parameters
hardware_estimation()
Toggle hardware estimation calculation
@@ -321,53 +321,42 @@ class ShiftLIF(LIF, ModNEFNeuron):
)
return module
- def quantize_weight(self, dtype=torch.int32):
+ def quantize_weight(self):
"""
Quantize synaptic weight
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight)
- self.fc.weight.data = self.quantizer(self.fc.weight.data, True, dtype)
+ self.fc.weight.data = self.quantizer(self.fc.weight.data, True)
- def quantize_parameters(self, dtype=torch.int32):
+ def quantize_parameters(self):
"""
Quantize neuron hyper-parameters
-
- Parameters
- ----------
- dtype = torch.int32 : dtype
- type use during quantization
"""
if not self.quantizer.is_initialize:
self.quantizer.init_from_weight(weight=self.fc.weight)
- self.threshold.data = self.quantizer(self.threshold.data, True, dtype)
- self.beta.data = self.quantizer(self.beta.data, True, dtype)
+ self.threshold.data = self.quantizer(self.threshold.data, True)
+ self.beta.data = self.quantizer(self.beta.data, True)
- def quantize(self, force_init=False, dtype=torch.int32):
+ def quantize(self, force_init=False):
"""
Quantize synaptic weight and neuron hyper-parameters
Parameters
----------
- dtype = torch.int32 : dtype
- type use during quantization
+ force_init = Fasle : bool
+ force quantizer initialization
"""
if force_init or not self.quantizer.is_initialize:
self.quantizer.init_from_weight(self.fc.weight)
- self.quantize_weight(dtype)
- self.quantize_parameters(dtype)
- self.quantization_flag = True
+ self.quantize_weight()
+ self.quantize_parameters()
@classmethod
def detach_hidden(cls):
diff --git a/modneflib/modnef/quantizer/dynamic_scale_quantizer.py b/modneflib/modnef/quantizer/dynamic_scale_quantizer.py
index c6e95d4..4f4b096 100644
--- a/modneflib/modnef/quantizer/dynamic_scale_quantizer.py
+++ b/modneflib/modnef/quantizer/dynamic_scale_quantizer.py
@@ -1,7 +1,7 @@
"""
File name: dynamic_scale_quantizer
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, modnef.quantizer.Quantizer
@@ -23,6 +23,8 @@ class DynamicScaleFactorQuantizer(Quantizer):
set to true if quantizer is signed, false if not
is_initializer : bool
set to true when quantizer parameters are initialize
+ dtype = torch.int32 : torch.dtype
+ type use during quantization
scale_factor : float
scale factor
@@ -33,14 +35,15 @@ class DynamicScaleFactorQuantizer(Quantizer):
generate quantizer from dictionnary description
init_from_weight(weight, rec_weight)
initialize quantizer parameters from synaptic weight
- __call__(data, unscale, dtype)
+ __call__(data, unscale)
Call quantization function, if unscale=true, return dequantization data
"""
def __init__(self,
bitwidth,
signed=None,
- is_initialize=False
+ is_initialize=False,
+ dtype=torch.int32
):
"""
Construct class
@@ -54,12 +57,15 @@ class DynamicScaleFactorQuantizer(Quantizer):
If None, determine during initialization
is_initialize = False : bool
set to true when quantizer parameters are initialize
+ dtype = troch.int32 : torch.dtype
+ type use during quatization
"""
super().__init__(
bitwidth=bitwidth,
signed=signed,
- is_initialize=is_initialize
+ is_initialize=is_initialize,
+ dtype=dtype
)
self.scale_factor = 0
@@ -124,7 +130,7 @@ class DynamicScaleFactorQuantizer(Quantizer):
#self.scale_factor = torch.max(torch.abs(weight).max(), torch.abs(weight).max())/2**(self.bitwidth-1)
- def _quant(self, data, unscale, dtype=torch.int32) -> torch.Tensor:
+ def _quant(self, data, unscale) -> torch.Tensor:
"""
Apply quantization
@@ -134,8 +140,6 @@ class DynamicScaleFactorQuantizer(Quantizer):
data to quantize
unscale = False : bool
If true, apply quantization and then, unquantize data to simulate quantization
- dtype=torch.int32 : dtype
- data type
Returns
-------
@@ -146,7 +150,7 @@ class DynamicScaleFactorQuantizer(Quantizer):
born_max = 2**(self.bitwidth-int(self.signed))-1
#scaled = torch.clamp(data/self.scale_factor, min=born_min, max=born_max).to(dtype)
- scaled = torch.round(data/self.scale_factor).to(dtype)
+ scaled = torch.round(data/self.scale_factor).to(self.dtype)
if unscale:
return scaled*self.scale_factor
diff --git a/modneflib/modnef/quantizer/fixed_point_quantizer.py b/modneflib/modnef/quantizer/fixed_point_quantizer.py
index afb36ac..564fceb 100644
--- a/modneflib/modnef/quantizer/fixed_point_quantizer.py
+++ b/modneflib/modnef/quantizer/fixed_point_quantizer.py
@@ -1,7 +1,7 @@
"""
File name: fixed_point_quantizer
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, math, modnef.quantizer.Quantizer
@@ -24,6 +24,8 @@ class FixedPointQuantizer(Quantizer):
set to true if quantizer is signed, false if not
is_initializer : bool
set to true when quantizer parameters are initialize
+ dtype = torch.int32 : torch.dtype
+ type use during quantization
fixed_point : int
fixed point position
scale_factor : int
@@ -36,7 +38,7 @@ class FixedPointQuantizer(Quantizer):
generate quantizer from dictionnary description
init_from_weight(weight, rec_weight)
initialize quantizer parameters from synaptic weight
- __call__(data, unscale, dtype)
+ __call__(data, unscale)
Call quantization function, if unscale=true, return dequantization data
"""
@@ -44,7 +46,8 @@ class FixedPointQuantizer(Quantizer):
bitwidth,
fixed_point=-1,
signed=None,
- is_initialize=False
+ is_initialize=False,
+ dtype=torch.int32
):
"""
Construct class
@@ -61,6 +64,8 @@ class FixedPointQuantizer(Quantizer):
If None, determine during initialization
is_initialize = False : bool
set to true when quantizer parameters are initialize
+ dtype = torch.int32 : torch.dtype
+ type use during conversion
"""
if bitwidth==-1 and fixed_point==-1:
@@ -69,7 +74,8 @@ class FixedPointQuantizer(Quantizer):
super().__init__(
bitwidth=bitwidth,
signed=signed,
- is_initialize=is_initialize
+ is_initialize=is_initialize,
+ dtype=dtype
)
self.fixed_point = fixed_point
@@ -141,7 +147,7 @@ class FixedPointQuantizer(Quantizer):
self.scale_factor = 2**self.fixed_point
- def _quant(self, data, unscale, dtype) -> torch.Tensor:
+ def _quant(self, data, unscale) -> torch.Tensor:
"""
Apply quantization
@@ -151,15 +157,13 @@ class FixedPointQuantizer(Quantizer):
data to quantize
unscale = False : bool
If true, apply quantization and then, unquantize data to simulate quantization
- dtype=torch.int32 : dtype
- data type
Returns
-------
Tensor
"""
- scaled = torch.round(data*self.scale_factor).to(dtype)
+ scaled = torch.round(data*self.scale_factor).to(self.dtype)
if unscale:
return (scaled.to(torch.float32))/self.scale_factor
diff --git a/modneflib/modnef/quantizer/min_max_quantizer.py b/modneflib/modnef/quantizer/min_max_quantizer.py
index 28d1cf3..a700f6b 100644
--- a/modneflib/modnef/quantizer/min_max_quantizer.py
+++ b/modneflib/modnef/quantizer/min_max_quantizer.py
@@ -1,7 +1,7 @@
"""
File name: quantizer
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, modnef.quantizer.Quantizer
@@ -23,6 +23,8 @@ class MinMaxQuantizer(Quantizer):
set to true if quantizer is signed, false if not
is_initializer : bool
set to true when quantizer parameters are initialize
+ dtype : torch.dtype
+ type use during quantization
x_min : float
minimal value of weight
x_max : float
@@ -39,14 +41,15 @@ class MinMaxQuantizer(Quantizer):
generate quantizer from dictionnary description
init_from_weight(weight, rec_weight)
initialize quantizer parameters from synaptic weight
- __call__(data, unscale, dtype)
+ __call__(data, unscale)
Call quantization function, if unscale=true, return dequantization data
"""
def __init__(self,
bitwidth,
signed=None,
- is_initialize=False
+ is_initialize=False,
+ dtype=torch.int32
):
"""
Construct class
@@ -60,12 +63,15 @@ class MinMaxQuantizer(Quantizer):
If None, determine during initialization
is_initialize = False : bool
set to true when quantizer parameters are initialize
+ dtype = torch.int32 : toch.dtype
+ type use during quantization
"""
super().__init__(
bitwidth=bitwidth,
signed=signed,
- is_initialize=is_initialize
+ is_initialize=is_initialize,
+ dtype=dtype
)
self.x_min = 0
@@ -128,7 +134,7 @@ class MinMaxQuantizer(Quantizer):
self.b_max = 2**(self.bitwidth-int(self.signed))-1
self.b_min = -int(self.signed)*self.b_max
- def _quant(self, data, unscale, dtype) -> torch.Tensor:
+ def _quant(self, data, unscale) -> torch.Tensor:
"""
Apply quantization
@@ -138,15 +144,13 @@ class MinMaxQuantizer(Quantizer):
data to quantize
unscale = False : bool
If true, apply quantization and then, unquantize data to simulate quantization
- dtype=torch.int32 : dtype
- data type
Returns
-------
Tensor
"""
- scaled = ((data-self.x_min)/(self.x_max-self.x_min)*(self.b_max-self.b_min)+self.b_min).to(dtype)
+ scaled = ((data-self.x_min)/(self.x_max-self.x_min)*(self.b_max-self.b_min)+self.b_min).to(self.dtype)
if unscale:
return (scaled-self.b_min)/(self.b_max-self.b_min)*(self.x_max-self.x_min)+self.x_min
diff --git a/modneflib/modnef/quantizer/quantizer.py b/modneflib/modnef/quantizer/quantizer.py
index 94fc8f5..5a37433 100644
--- a/modneflib/modnef/quantizer/quantizer.py
+++ b/modneflib/modnef/quantizer/quantizer.py
@@ -1,7 +1,7 @@
"""
File name: quantizer
Author: Aurélie Saulquin
-Version: 1.0.0
+Version: 1.1.0
License: GPL-3.0-or-later
Contact: aurelie.saulquin@univ-lille.fr
Dependencies: torch, numpy
@@ -22,7 +22,9 @@ class Quantizer():
signed : bool
set to true if quantizer is signed, false if not
is_initializer : bool
- set to true when quantizer parameters are initialize
+ set to true when quantizer parameters are initialize*
+ dtype : torch.dtype
+ type use during quantization
Methods
-------
@@ -31,14 +33,15 @@ class Quantizer():
generate quantizer from dictionnary description
init_from_weight(weight, rec_weight)
initialize quantizer parameters from synaptic weight
- __call__(data, unscale, dtype)
+ __call__(data, unscale)
Call quantization function, if unscale=true, return dequantization data
"""
def __init__(self,
bitwidth,
signed=None,
- is_initialize=False
+ is_initialize=False,
+ dtype=torch.int32
):
"""
Construct class
@@ -52,12 +55,14 @@ class Quantizer():
If None, determine during initialization
is_initialize = False : bool
set to true when quantizer parameters are initialize
+ dtype = torch.int32 : torch.dtype
+ type use during quantization
"""
self.bitwidth = bitwidth
self.is_initialize = is_initialize
self.signed = signed
- pass
+ self.dtype=dtype
@classmethod
def from_dict(cls, dict):
@@ -86,7 +91,7 @@ class Quantizer():
raise NotImplementedError()
- def __call__(self, data, unscale=False, dtype=torch.int16):
+ def __call__(self, data, unscale=False):
"""
Call quantization function
@@ -96,8 +101,6 @@ class Quantizer():
data to quantize
unscale = False : bool
If true, apply quantization and then, unquantize data to simulate quantization
- dtype=torch.int32 : dtype
- data type
Returns
-------
@@ -105,17 +108,17 @@ class Quantizer():
"""
if isinstance(data, (int, float)):
- return self._quant(data=torch.tensor(data), unscale=unscale, dtype=dtype).item()
+ return self._quant(data=torch.tensor(data), unscale=unscale).item()
elif isinstance(data, list):
- return self._quant(data=torch.tensor(data), unscale=unscale, dtype=dtype).tolist()
+ return self._quant(data=torch.tensor(data), unscale=unscale).tolist()
elif isinstance(data, np.ndarray):
- return self._quant(data=torch.tensor(data), unscale=unscale, dtype=dtype).numpy()
+ return self._quant(data=torch.tensor(data), unscale=unscale).numpy()
elif torch.is_tensor(data):
- return self._quant(data=data, unscale=unscale, dtype=dtype).detach()
+ return self._quant(data=data, unscale=unscale).detach()
else:
raise TypeError("Unsupported data type")
- def _quant(self, data, unscale, dtype) -> torch.Tensor:
+ def _quant(self, data, unscale) -> torch.Tensor:
"""
Apply quantization
@@ -125,8 +128,6 @@ class Quantizer():
data to quantize
unscale = False : bool
If true, apply quantization and then, unquantize data to simulate quantization
- dtype=torch.int32 : dtype
- data type
Returns
-------
diff --git a/modneflib/setup.py b/modneflib/setup.py
index 191c203..6f56208 100644
--- a/modneflib/setup.py
+++ b/modneflib/setup.py
@@ -3,7 +3,7 @@ from setuptools import find_packages, setup
setup(
name = "modnef",
packages=find_packages(),
- version = "=2.0.0",
+ version = "2.0.0",
description="ModNEF python librairy",
author="Aurelie Saulquin",
install_requires=["networkx", "matplotlib", "pyyaml", "torch", "snntorch"],
--
GitLab