From 6bb39ed7ee39b5d4e8e0e2261337904edf323827 Mon Sep 17 00:00:00 2001
From: Yaman Umuroglu <maltanar@gmail.com>
Date: Wed, 17 Jun 2020 23:21:36 +0100
Subject: [PATCH] [Core] rename validate -> sanitize_quant_values, add comments
plus a small bugfix (break -> continue)
---
src/finn/core/onnx_exec.py | 8 ++++----
src/finn/util/basic.py | 29 +++++++++++++++++++++++------
2 files changed, 27 insertions(+), 10 deletions(-)
diff --git a/src/finn/core/onnx_exec.py b/src/finn/core/onnx_exec.py
index 0bacbc8b0..218df22e0 100644
--- a/src/finn/core/onnx_exec.py
+++ b/src/finn/core/onnx_exec.py
@@ -39,7 +39,7 @@ from finn.core.remote_exec import remote_exec
from finn.core.rtlsim_exec import rtlsim_exec
from finn.custom_op.registry import getCustomOp
import finn.analysis.topology as ta
-from finn.util.basic import validate_quant_values
+from finn.util.basic import sanitize_quant_values
def execute_node(node, context, graph):
@@ -103,7 +103,7 @@ def execute_node(node, context, graph):
raise Exception(
"""Output shapes disagree after node execution:
found %s vs expected %s"""
- % (str(output_list[list_ind].shape), str(context[outp].shape),)
+ % (str(output_list[list_ind].shape), str(context[outp].shape))
)
context[outp] = output_list[list_ind]
@@ -161,11 +161,11 @@ def execute_onnx(model, input_dict, return_full_exec_context=False):
# topologically sorted
for node in graph.node:
# call util function match input values to quantization annotation
- execution_context = validate_quant_values(
+ execution_context = sanitize_quant_values(
model, node.input, execution_context
)
execute_node(node, execution_context, graph)
- execution_context = validate_quant_values(
+ execution_context = sanitize_quant_values(
model, node.output, execution_context
)
elif model_exec_mode == "remote_pynq":
diff --git a/src/finn/util/basic.py b/src/finn/util/basic.py
index c4154aad4..585d6b90a 100644
--- a/src/finn/util/basic.py
+++ b/src/finn/util/basic.py
@@ -265,12 +265,29 @@ def calculate_signed_dot_prod_range(dt_a, dt_b, len):
return (min_prod, max_prod)
-def validate_quant_values(model, node_tensors, execution_context, check_values=False):
+def sanitize_quant_values(model, node_tensors, execution_context, check_values=False):
+ """ Sanitize given list of tensors in execution_context by rounding values
+ that are supposed to be integers (as indicated by their quantization
+ annotation). Will raise an assertion if the amount of rounding is too large.
+ Returns the sanitized execution context.
+
+ If check_values is specified, an extra DataType.allowed() check will be
+ performed on any rounded tensors.
+
+ Background:
+ FINN uses floating point tensors as a carrier data type to represent
+ integers. Floating point arithmetic can introduce rounding errors, e.g.
+ (int_num * float_scale) / float_scale is not always equal to int_num.
+ We use this function to ensure that the values that are supposed to be
+ integers are indeed integers.
+ """
+
for tensor in node_tensors:
dtype = model.get_tensor_datatype(tensor)
- # if FLOAT32 skip to next tensor
+ # floats don't need sanitization, skip to next
+ # introduces less quicker runtime
if dtype == DataType.FLOAT32:
- break
+ continue
current_values = execution_context[tensor]
updated_values = current_values
has_to_be_rounded = False
@@ -283,8 +300,8 @@ def validate_quant_values(model, node_tensors, execution_context, check_values=F
updated_values = np.round(current_values)
warnings.warn(
"The values of tensor {} can't be represented "
- "with the set finn datatype ({}), they will be rounded to match the "
- "finn datatype.".format(tensor, dtype)
+ "with the set FINN datatype ({}), they will be rounded to match the "
+ "FINN datatype.".format(tensor, dtype)
)
# check if rounded values are not too far from original values
max_error = max(np.abs(current_values - updated_values).flatten())
@@ -303,7 +320,7 @@ def validate_quant_values(model, node_tensors, execution_context, check_values=F
execution_context[tensor] = updated_values
else:
raise Exception(
- """Rounding error is too high to match set finn
+ """Rounding error is too high to match set FINN
datatype ({}) for input {}""".format(
dtype, tensor
)
--
GitLab