You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The documents of MCT describe countermeasures for accuracy degradation based on the causes identified by the XQuant Extension Tool.
18
+
This document describes countermeasures for accuracy degradation based on causes identified by the XQuant Extension Tool.
19
19
20
20
Overall Process Flow
21
21
============================
@@ -27,19 +27,19 @@ The overall process follows the steps below:
27
27
1. Input the float model, quantized model, and quantization log.
28
28
2. Detect layers that have large difference between float and quantized models.
29
29
3. Judge degradation causes on the detected layers.
30
-
4. Based on the judge results, individual countermeasure procedures or general improvement measures are proposed from the troubleshooting manual.
31
-
5. Recommend general troubleshoots additionally when accuracy does not improve after steps 1-4.
30
+
4. [Judgeable Troubleshoots] Based on the judge results, individual countermeasure procedures from the troubleshooting manual.
31
+
5. [General Troubleshoots] When accuracy does not improve after steps 1-4, general improvement measures from the troubleshooting manual.
32
32
33
33
Please refer to the `Troubleshooting Manual <https://sonysemiconductorsolutions.github.io/mct-model-optimization/docs_troubleshoot/index.html>`_ for the Judgeable Troubleshoots and General Troubleshoots in detail.
34
34
35
35
How to Run
36
36
===============
37
37
38
-
This XQuant Extension Tool was created based on xquant, as shown in the link below.
39
-
In addition to the conventional xquant functions, it judges degradation causes and links to a troubleshooting manual that provides appropriate countermeasures for each cause of degradation.
38
+
This XQuant Extension Tool was created based on XQuant, as shown in the link below.
39
+
In addition to the conventional XQuant functions, it judges degradation causes and links to the Troubleshooting Manual that provides appropriate countermeasures for each cause of degradation.
40
40
It can suggest more specific countermeasures than conventional tools and provides manuals that are easy to understand even for users who are not familiar with quantization.
41
41
42
-
Please replace xquant_report_pytorch_experimental in `the XQuant tutorial <https://github.com/SonySemiconductorSolutions/mct-model-optimization/tree/main/tutorials/notebooks/mct_features_notebooks/pytorch/example_pytorch_xquant.ipynb>`_ with xquant_report_troubleshoot_pytorch_experimental.
42
+
Please replace *xquant_report_pytorch_experimental* in `the XQuant tutorial <https://github.com/SonySemiconductorSolutions/mct-model-optimization/tree/main/tutorials/notebooks/mct_features_notebooks/pytorch/example_pytorch_xquant.ipynb>`_ with *xquant_report_troubleshoot_pytorch_experimental*.
43
43
44
44
.. code-block:: python
45
45
@@ -154,14 +154,10 @@ The initial threshold value of 0.1 is set, and layers exceeding this threshold a
154
154
* **Red dashed line**: Threshold for accuracy degradation as set in XQuantConfig
155
155
* **Red circle**: Layers judged to have degraded accuracy
156
156
157
-
.. note::
158
-
159
-
You can define new calculation methods. For details, see the `API Document <https://sonysemiconductorsolutions.github.io/mct-model-optimization/api/api_docs/classes/XQuantConfig.html#ug-xquantconfig>`_.
@@ -12,7 +12,7 @@ The Model Compression Toolkit (MCT) offers numerous functionalities to compress
12
12
13
13
Outlined below are a series of steps aimed at recovering lost accuracy resulting from compression with MCT. Some steps may be applicable to your model, while others may not.
14
14
15
-
For ovewview of Xquant Extension Tool, see `About XQuant Extension Tool <https://sonysemiconductorsolutions.github.io/mct-model-optimization/guidelines/XQuant_Extension_Tool.html>`_ [1].
15
+
For overview of XQuant Extension Tool, see `About XQuant Extension Tool <https://sonysemiconductorsolutions.github.io/mct-model-optimization/guidelines/XQuant_Extension_Tool.html>`_ [1].
<spanid="ug-index"></span><h1>TroubleShooting Manual (MCT XQuant Extension Tool)<aclass="headerlink" href="#troubleshooting-manual-mct-xquant-extension-tool" title="Link to this heading">¶</a></h1>
43
43
<sectionid="overview">
44
44
<h2>Overview<aclass="headerlink" href="#overview" title="Link to this heading">¶</a></h2>
45
45
<p>The Model Compression Toolkit (MCT) offers numerous functionalities to compress neural networks with minimal accuracy lost. However, in some cases, the compressed model may experience a significant decrease in accuracy. Fear not, as this lost accuracy can often be reclaimed by adjusting the quantization configuration or setup.</p>
46
46
<p>Outlined below are a series of steps aimed at recovering lost accuracy resulting from compression with MCT. Some steps may be applicable to your model, while others may not.</p>
47
-
<p>For ovewview of Xquant Extension Tool, see <aclass="reference external" href="https://sonysemiconductorsolutions.github.io/mct-model-optimization/guidelines/XQuant_Extension_Tool.html">About XQuant Extension Tool</a> [1].</p>
47
+
<p>For overview of XQuant Extension Tool, see <aclass="reference external" href="https://sonysemiconductorsolutions.github.io/mct-model-optimization/guidelines/XQuant_Extension_Tool.html">About XQuant Extension Tool</a> [1].</p>
0 commit comments