Skip to content

Commit 6a9640a

Browse files
xiaolil1xiaolil1
committed
Add img
1 parent 1b3cf01 commit 6a9640a

File tree

2 files changed

+2
-31
lines changed

2 files changed

+2
-31
lines changed

‎_static/img/pt2e_quant_xpu_inductor.png‎

124 KB
Loading

‎prototype_source/pt2e_quant_xpu_inductor.rst‎

Lines changed: 2 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -33,37 +33,8 @@ Besides oneDNN kernels, triton would be responsible for generating kernels on ou
3333

3434
The high-level architecture of this flow could look like this:
3535

36-
::
37-
38-
float_model(Python) Example Input
39-
\ /
40-
\ /
41-
—--------------------------------------------------------
42-
| export |
43-
—--------------------------------------------------------
44-
|
45-
FX Graph in ATen
46-
| XPUInductorQuantizer
47-
| /
48-
—--------------------------------------------------------
49-
| prepare_pt2e |
50-
| | |
51-
| Calibrate/Train |
52-
| | |
53-
| convert_pt2e |
54-
—--------------------------------------------------------
55-
|
56-
Quantized Model
57-
|
58-
—--------------------------------------------------------
59-
| Lower into Inductor |
60-
—--------------------------------------------------------
61-
|
62-
Inductor
63-
|
64-
—--------------------------------------------------------
65-
| oneDNN Kernels ATen Ops Triton Kernels |
66-
—--------------------------------------------------------
36+
.. image:: /_static/img/pt2e_quant_xpu_inductor.png
37+
:align: center
6738

6839
Post Training Quantization
6940
----------------------------

0 commit comments

Comments
 (0)