Commit ·
3d77f0b
1
Parent(s): df61939
Add a16w8, the QNN 16-bit-activation / 8-bit-weight precision
Browse files- precisions.json +1 -0
precisions.json
CHANGED
|
@@ -5,6 +5,7 @@
|
|
| 5 |
"4w",
|
| 6 |
"8da4w",
|
| 7 |
"8da8w",
|
|
|
|
| 8 |
"a8w8",
|
| 9 |
"int4",
|
| 10 |
"int8",
|
|
|
|
| 5 |
"4w",
|
| 6 |
"8da4w",
|
| 7 |
"8da8w",
|
| 8 |
+
"a16w8",
|
| 9 |
"a8w8",
|
| 10 |
"int4",
|
| 11 |
"int8",
|