immich-app/ml-models
1
0
Fork 0
mirror of https://github.com/immich-app/ml-models.git synced 2026-09-30 13:22:55 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
60e028dee6134d70f3ede9d77a1b3ceb536c6a7b
ml-models/ci/graphs/ESLAV__PP-OCRv5_mobile/recognition.txt
T
Merttodorangrggithub-actions
 60e028dee6 feat: add ocr to exporter (#61) 
* feat: add ppocr to exporter

* ocr: unpack scrfd heads

* ocr: affine fold

Co-authored-by: todorangrg <todorangrg@gmail.com>

* rknn dynamic inputs

* fp16 outputs

* tiled transpose

* refactor

* target multiple aspect ratios

* update cli

* add 16:9

* better ci fixture integration

* ci bug fixes

* check cancellation

* chore: bless ci fixtures

* tweak runners

* add metadata to rknn binaries

* canvas->dim

* fix fixture trigger

* update onnxscript locals

* chore: bless ci fixtures

* download unconditionally

* simplify fix-fixtures

* use huge runner for server model

* run compilation in subprocess

* try/catch recv

* remove 16/9

* fix detection mac utilization

* chore: bless ci fixtures

* conditional rknn compile

* add rknn config to render

---------

Co-authored-by: todorangrg <todorangrg@gmail.com>
Co-authored-by: github-actions <41898282+github-actions[bot]@users.noreply.github.com>
2026-08-27 05:24:29 -04:00

1057 lines
39 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
<
ir_version: 10,
opset_import: ["" : 20],
metadata_props: ["character": "!
\"
#
$
%
&
'
(
)
*
+
,
-
.
/
0
1
2
3
4
5
6
7
8
9
:
;
<
=
>
?
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W
X
Y
Z
[
]
_
`
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
q
r
s
t
u
v
w
x
y
z
©
‥
{
}
\\
|
@
^
~
÷
∕
∙
⋅
·
±
∓
∩
∪
□
←
↔
⇒
⇐
⇔
∀
∃
∄
∴
∵
∝
∞
⊥
∟
∠
∡
∢
′
″
∥
⊾
⊿
∂
∫
∬
∭
∮
∯
∰
∑
∏
√
∛
∜
∱
∲
∳
∶
∷
∼
®
℉
Ω
℧
Å
⌀
ℏ
⅀
⍺
⍵
¢
€
£
¥
₿
Ⅰ
Ⅱ
Ⅲ
Ⅳ
Ⅴ
Ⅵ
Ⅶ
Ⅷ
Ⅸ
Ⅹ
Ⅺ
Ⅻ
ⅰ
ⅱ
ⅲ
ⅳ
ⅴ
ⅵ
ⅶ
ⅷ
ⅸ
ⅹ
ⅺ
ⅻ
➀
➁
➂
➃
➄
➅
➆
➇
➈
➉
➊
➋
➌
➍
➎
➏
➐
➑
➒
➓
❶
❷
❸
❹
❺
❻
❼
❽
❾
❿
①
②
③
④
⑤
⑥
⑦
⑧
⑨
⑩
●
▶
𝑢
︽
–
﹥
𝜓
•
∋
ƒ
०
⬆
Ạ
◀
▫
︾
À
Á
Â
Ã
Ä
Å
Æ
Ç
È
É
Ê
Ë
Ì
Í
Î
Ï
Ð
Ñ
Ò
Ó
Ô
Õ
Ö
Ø
Ù
Ú
Û
Ü
Ý
Þ
à
á
â
ã
ä
å
æ
ç
è
é
ê
ë
ì
í
î
ï
ð
ñ
ò
ó
ô
õ
ö
ø
ù
ú
û
ü
ý
þ
ÿ
¡
¤
¦
§
¨
ª
«
¬
¯
°
²
³
´
µ
¸
¹
º
»
¼
½
¾
¿
×
‐
‑
‒
—
―
‖
‗
‘
’
‚
‛
“
”
„
‟
†
‡
‣
․
…
‧
‰
‴
‵
‶
‷
‸
‹
›
※
‼
‽
‾
₤
₡
₹
−
∖
∗
≈
≠
≡
≤
≥
⊂
⊃
↑
→
↓
↕
™
Ω
℮
∆
✓
✗
✘
▪
◼
✔
✕
☑
☒
№
₽
₴
Α
α
Β
β
Γ
γ
Δ
δ
Ε
ε
Ζ
ζ
Η
η
Θ
θ
Ι
ι
Κ
κ
Λ
λ
Μ
μ
Ν
ν
Ξ
ξ
Ο
ο
Π
π
Ρ
ρ
Σ
σ
ς
Τ
τ
Υ
υ
Φ
φ
Χ
χ
Ψ
ψ
ω
А
Б
В
Г
Ґ
Д
Е
Ё
Є
Ж
З
И
І
Ї
Й
К
Л
М
Н
О
П
Р
С
Т
У
Ў
Ф
Х
Ц
Ч
Ш
Щ
Ъ
Ы
Ь
Э
Ю
Я
а
б
в
г
ґ
д
е
ё
є
ж
з
и
і
ї
й
к
л
м
н
о
п
р
с
т
у
ў
ф
х
ц
ч
ш
щ
ъ
ы
ь
э
ю
я
"]
>
"PaddlePaddle Graph in PIR mode" (uint8[batch,48,width,3] image) => (int32[batch,seq] ctc_indices, float[batch,seq] ctc_confidence)
<
float[batch,240,12,unk__11] "Add.105"
float[batch,240,12,unk__11] "Add.111"
float[batch,240,12,unk__11] "Add.117"
float[batch,240,6,unk__11] "Add.123"
float[batch,240,6,unk__11] "Add.125"
float[batch,60,1,1] "Add.127"
float[batch,240,1,1] "Add.129"
float[batch,480,6,unk__11] "Add.133"
float[batch,480,6,unk__11] "Add.139"
float[batch,480,6,unk__11] "Add.141"
float[batch,120,1,1] "Add.143"
float[batch,480,1,1] "Add.145"
float[batch,480,6,unk__11] "Add.149"
float[batch,32,24,unk__0] "Add.15"
float[batch,480,3,unk__11] "Add.155"
float[batch,480,3,unk__11] "Add.161"
float[batch,480,3,unk__11] "Add.167"
float[batch,480,3,unk__11] "Add.173"
float[batch,seq,120] "Add.181"
float[batch,seq,360] "Add.183"
float[batch,8,seq,15] "Add.183_k"
float[batch,8,seq,15] "Add.183_q"
float[batch,seq,24,15] "Add.183_qkv"
float[batch,24,seq,15] "Add.183_qkv_heads"
float[batch,8,seq,15] "Add.183_v"
float[batch,seq,120] "Add.185"
float[batch,seq,120] "Add.187"
float[batch,seq,120] "Add.191"
float[batch,seq,240] "Add.193"
float[batch,seq,120] "Add.195"
float[batch,seq,120] "Add.197"
float[batch,seq,120] "Add.201"
float[batch,seq,360] "Add.203"
float[batch,8,seq,15] "Add.203_k"
float[batch,8,seq,15] "Add.203_q"
float[batch,seq,24,15] "Add.203_qkv"
float[batch,24,seq,15] "Add.203_qkv_heads"
float[batch,8,seq,15] "Add.203_v"
float[batch,seq,120] "Add.205"
float[batch,seq,120] "Add.207"
float[batch,64,24,unk__0] "Add.21"
float[batch,seq,120] "Add.211"
float[batch,seq,240] "Add.213"
float[batch,seq,120] "Add.215"
float[batch,seq,120] "Add.217"
float[batch,seq,120] "Add.221"
float[batch,seq,519] "Add.223"
float[batch,64,24,unk__0] "Add.27"
float[batch,16,24,unk__0] "Add.3"
float[batch,64,24,unk__0] "Add.33"
float[batch,64,12,unk__0] "Add.39"
float[batch,128,12,unk__0] "Add.45"
float[batch,128,12,unk__0] "Add.51"
float[batch,128,12,unk__0] "Add.57"
float[batch,128,12,unk__11] "Add.63"
float[batch,240,12,unk__11] "Add.69"
float[batch,240,12,unk__11] "Add.75"
float[batch,240,12,unk__11] "Add.81"
float[batch,240,12,unk__11] "Add.87"
float[batch,32,24,unk__0] "Add.9"
float[batch,240,12,unk__11] "Add.93"
float[batch,240,12,unk__11] "Add.99"
float[batch,960,1,seq] "Concat.4"
float[batch,240,6,unk__11] "Mul.83"
float[batch,240,6,unk__11] "Mul.85"
float[batch,480,6,unk__11] "Mul.93"
float[batch,480,6,unk__11] "Mul.95"
float[batch,60,1,seq] "Sigmoid.1"
float[batch,60,1,seq] "Sigmoid.11"
float[batch,120,1,seq] "Sigmoid.13"
float[batch,120,1,seq] "Sigmoid.3"
float[batch,seq,240] "Sigmoid.5"
float[batch,seq,240] "Sigmoid.7"
float[batch,480,1,seq] "Sigmoid.9"
float[batch,16,24,unk__0] "p2o.pd_op.batch_norm_.0.0"
float[batch,60,1,seq] "p2o.pd_op.batch_norm_.1.0"
float[batch,120,1,seq] "p2o.pd_op.batch_norm_.2.0"
float[batch,480,1,seq] "p2o.pd_op.batch_norm_.3.0"
float[batch,60,1,seq] "p2o.pd_op.batch_norm_.4.0"
float[batch,120,1,seq] "p2o.pd_op.batch_norm_.5.0"
float[batch,120,seq] "p2o.pd_op.flatten.0.0"
float[batch,240,1,1] "p2o.pd_op.hardsigmoid.0.0"
float[batch,480,1,1] "p2o.pd_op.hardsigmoid.1.0"
float[batch,16,24,unk__0] "p2o.pd_op.hardswish.0.0"
float[batch,32,24,unk__0] "p2o.pd_op.hardswish.1.0"
float[batch,128,12,unk__11] "p2o.pd_op.hardswish.10.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.11.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.12.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.13.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.14.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.15.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.16.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.17.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.18.0"
float[batch,240,12,unk__11] "p2o.pd_op.hardswish.19.0"
float[batch,32,24,unk__0] "p2o.pd_op.hardswish.2.0"
float[batch,240,6,unk__11] "p2o.pd_op.hardswish.20.0"
float[batch,480,6,unk__11] "p2o.pd_op.hardswish.21.0"
float[batch,480,6,unk__11] "p2o.pd_op.hardswish.22.0"
float[batch,480,6,unk__11] "p2o.pd_op.hardswish.23.0"
float[batch,480,3,unk__11] "p2o.pd_op.hardswish.24.0"
float[batch,480,3,unk__11] "p2o.pd_op.hardswish.25.0"
float[batch,480,3,unk__11] "p2o.pd_op.hardswish.26.0"
float[batch,480,3,unk__11] "p2o.pd_op.hardswish.27.0"
float[batch,64,24,unk__0] "p2o.pd_op.hardswish.3.0"
float[batch,64,24,unk__0] "p2o.pd_op.hardswish.4.0"
float[batch,64,24,unk__0] "p2o.pd_op.hardswish.5.0"
float[batch,64,12,unk__0] "p2o.pd_op.hardswish.6.0"
float[batch,128,12,unk__0] "p2o.pd_op.hardswish.7.0"
float[batch,128,12,unk__0] "p2o.pd_op.hardswish.8.0"
float[batch,128,12,unk__0] "p2o.pd_op.hardswish.9.0"
float[batch,seq,360] "p2o.pd_op.matmul.0.0"
float[batch,8,seq,seq] "p2o.pd_op.matmul.1.0"
float[batch,seq,240] "p2o.pd_op.matmul.10.0"
float[batch,seq,120] "p2o.pd_op.matmul.11.0"
float[batch,seq,519] "p2o.pd_op.matmul.12.0"
float[batch,8,seq,15] "p2o.pd_op.matmul.2.0"
float[batch,seq,120] "p2o.pd_op.matmul.3.0"
float[batch,seq,240] "p2o.pd_op.matmul.4.0"
float[batch,seq,120] "p2o.pd_op.matmul.5.0"
float[batch,seq,360] "p2o.pd_op.matmul.6.0"
float[batch,8,seq,seq] "p2o.pd_op.matmul.7.0"
float[batch,8,seq,15] "p2o.pd_op.matmul.8.0"
float[batch,seq,120] "p2o.pd_op.matmul.9.0"
float[batch,240,1,1] "p2o.pd_op.pool2d.0.0"
float[batch,480,1,1] "p2o.pd_op.pool2d.1.0"
float[batch,480,1,seq] "p2o.pd_op.pool2d.2.0"
float[batch,60,1,1] "p2o.pd_op.relu.0.0"
float[batch,120,1,1] "p2o.pd_op.relu.1.0"
float[batch,seq,120] "p2o.pd_op.reshape.33.0"
float[batch,seq,120] "p2o.pd_op.reshape.35.0"
float[batch,1,seq,120] "p2o.pd_op.reshape.36.0"
float[batch,8,seq,seq] "p2o.pd_op.softmax.0.0"
float[batch,8,seq,seq] "p2o.pd_op.softmax.1.0"
float[batch,120,seq] "p2o.pd_op.squeeze.0.0"
float[batch,60,1,seq] "p2o.pd_op.swish.0.0"
float[batch,120,1,seq] "p2o.pd_op.swish.1.0"
float[batch,seq,240] "p2o.pd_op.swish.2.0"
float[batch,seq,240] "p2o.pd_op.swish.3.0"
float[batch,480,1,seq] "p2o.pd_op.swish.4.0"
float[batch,60,1,seq] "p2o.pd_op.swish.5.0"
float[batch,120,1,seq] "p2o.pd_op.swish.6.0"
float[batch,seq,120] "p2o.pd_op.transpose.0.0"
float[batch,8,15,seq] "p2o.pd_op.transpose.2.0"
float[batch,seq,8,15] "p2o.pd_op.transpose.3.0"
float[batch,8,15,seq] "p2o.pd_op.transpose.5.0"
float[batch,seq,8,15] "p2o.pd_op.transpose.6.0"
float[batch,120,1,seq] "p2o.pd_op.transpose.7.0"
float[batch,seq,120] "p2o.pd_op.transpose.8.0"
float[batch,seq,1] max_logits
float[batch,48,width,3] tmp
float[batch,3,48,width] tmp_0
float[batch,seq,519] tmp_0_2
float[batch,seq,519] tmp_1
int64[batch,seq] tmp_2
float[batch,seq] tmp_3
float[batch,16,24,unk__0] val_0
float[batch,32,24,unk__0] val_1
float[batch,128,12,unk__0] val_10
float[batch,128,12,unk__0] val_11
float[batch,128,12,unk__0] val_12
float[batch,128,12,unk__0] val_13
float[batch,128,12,unk__0] val_14
float[batch,128,12,unk__11] val_15
float[batch,240,12,unk__11] val_16
float[batch,240,12,unk__11] val_17
float[batch,240,12,unk__11] val_18
float[batch,240,12,unk__11] val_19
float[batch,32,24,unk__0] val_2
float[batch,240,12,unk__11] val_20
float[batch,240,12,unk__11] val_21
float[batch,240,12,unk__11] val_22
float[batch,240,12,unk__11] val_23
float[batch,240,12,unk__11] val_24
float[batch,240,12,unk__11] val_25
float[batch,240,12,unk__11] val_26
float[batch,240,12,unk__11] val_27
float[batch,240,12,unk__11] val_28
float[batch,240,12,unk__11] val_29
float[batch,32,24,unk__0] val_3
float[batch,240,6,unk__11] val_30
float[batch,480,6,unk__11] val_31
float[batch,480,6,unk__11] val_32
float[batch,480,6,unk__11] val_33
float[batch,480,6,unk__11] val_34
float[batch,480,6,unk__11] val_35
float[batch,480,3,unk__11] val_36
float[batch,480,3,unk__11] val_37
float[batch,480,3,unk__11] val_38
float[batch,480,3,unk__11] val_39
float[batch,64,24,unk__0] val_4
float[batch,480,3,unk__11] val_40
float[batch,480,1,seq] val_41
float[batch,480,1,seq] val_42
float[batch,64,24,unk__0] val_5
float[batch,64,24,unk__0] val_6
float[batch,64,24,unk__0] val_7
float[batch,64,24,unk__0] val_8
float[batch,64,12,unk__0] val_9
float[batch,3,48,width] x
>
{
[n0] tmp = Cast <to: int = 1> (image)
[n1] tmp_0 = Transpose <perm: ints = [0, 3, 1, 2]> (tmp)
[n4] x = Sub (tmp_0, const_cast)
"p2o.pd_op.batch_norm_.0.0" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [3, 3], pads: ints = [1, 1, 1, 1], strides: ints = [2, 2]> (x, "conv2d_0.w_0", "conv2d_0.w_0_bias")
"Add.3" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 16, kernel_shape: ints = [3, 3], pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> ("p2o.pd_op.batch_norm_.0.0", "conv2d_136.w_0_lab", "p2o.pd_op.depthwise_conv2d.0.0_bias_lab")
val_0 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.3")
"p2o.pd_op.hardswish.0.0" = Mul ("Add.3", val_0)
"Add.9" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.0.0", "conv2d_137.w_0_lab_lab", "p2o.pd_op.conv2d.1.0_bias_lab_lab")
val_1 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.9")
"p2o.pd_op.hardswish.1.0" = Mul ("Add.9", val_1)
val_2 = Add ("p2o.pd_op.hardswish.1.0", "conv2d_138.w_0_lab_laboffset")
"Add.15" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 32, kernel_shape: ints = [3, 3], pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (val_2, "conv2d_138.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.1.0_bias_lab")
val_3 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.15")
"p2o.pd_op.hardswish.2.0" = Mul ("Add.15", val_3)
"Add.21" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.2.0", "conv2d_139.w_0_lab_lab", "p2o.pd_op.conv2d.2.0_bias_lab_lab")
val_4 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.21")
"p2o.pd_op.hardswish.3.0" = Mul ("Add.21", val_4)
val_5 = Add ("p2o.pd_op.hardswish.3.0", "conv2d_140.w_0_lab_laboffset")
"Add.27" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 64, kernel_shape: ints = [3, 3], pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (val_5, "conv2d_140.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.2.0_bias_lab")
val_6 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.27")
"p2o.pd_op.hardswish.4.0" = Mul ("Add.27", val_6)
"Add.33" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.4.0", "conv2d_141.w_0_lab_lab", "p2o.pd_op.conv2d.3.0_bias_lab_lab")
val_7 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.33")
"p2o.pd_op.hardswish.5.0" = Mul ("Add.33", val_7)
val_8 = Add ("p2o.pd_op.hardswish.5.0", "conv2d_142.w_0_lab_laboffset")
"Add.39" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 64, kernel_shape: ints = [3, 3], pads: ints = [1, 1, 1, 1], strides: ints = [2, 1]> (val_8, "conv2d_142.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.3.0_bias_lab")
val_9 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.39")
"p2o.pd_op.hardswish.6.0" = Mul ("Add.39", val_9)
"Add.45" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.6.0", "conv2d_143.w_0_lab_lab", "p2o.pd_op.conv2d.4.0_bias_lab_lab")
val_10 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.45")
"p2o.pd_op.hardswish.7.0" = Mul ("Add.45", val_10)
val_11 = Add ("p2o.pd_op.hardswish.7.0", "conv2d_144.w_0_lab_laboffset")
"Add.51" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 128, kernel_shape: ints = [3, 3], pads: ints = [1, 1, 1, 1], strides: ints = [1, 1]> (val_11, "conv2d_144.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.4.0_bias_lab")
val_12 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.51")
"p2o.pd_op.hardswish.8.0" = Mul ("Add.51", val_12)
"Add.57" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.8.0", "conv2d_145.w_0_lab_lab", "p2o.pd_op.conv2d.5.0_bias_lab_lab")
val_13 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.57")
"p2o.pd_op.hardswish.9.0" = Mul ("Add.57", val_13)
val_14 = Add ("p2o.pd_op.hardswish.9.0", "conv2d_146.w_0_lab_laboffset")
"Add.63" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 128, kernel_shape: ints = [3, 3], pads: ints = [1, 1, 1, 1], strides: ints = [1, 2]> (val_14, "conv2d_146.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.5.0_bias_lab")
val_15 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.63")
"p2o.pd_op.hardswish.10.0" = Mul ("Add.63", val_15)
"Add.69" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.10.0", "conv2d_147.w_0_lab_lab", "p2o.pd_op.conv2d.6.0_bias_lab_lab")
val_16 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.69")
"p2o.pd_op.hardswish.11.0" = Mul ("Add.69", val_16)
val_17 = Add ("p2o.pd_op.hardswish.11.0", "conv2d_148.w_0_lab_laboffset")
"Add.75" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 240, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [1, 1]> (val_17, "conv2d_148.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.6.0_bias_lab")
val_18 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.75")
"p2o.pd_op.hardswish.12.0" = Mul ("Add.75", val_18)
"Add.81" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.12.0", "conv2d_149.w_0_lab_lab", "p2o.pd_op.conv2d.7.0_bias_lab_lab")
val_19 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.81")
"p2o.pd_op.hardswish.13.0" = Mul ("Add.81", val_19)
val_20 = Add ("p2o.pd_op.hardswish.13.0", "conv2d_150.w_0_lab_laboffset")
"Add.87" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 240, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [1, 1]> (val_20, "conv2d_150.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.7.0_bias_lab")
val_21 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.87")
"p2o.pd_op.hardswish.14.0" = Mul ("Add.87", val_21)
"Add.93" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.14.0", "conv2d_151.w_0_lab_lab", "p2o.pd_op.conv2d.8.0_bias_lab_lab")
val_22 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.93")
"p2o.pd_op.hardswish.15.0" = Mul ("Add.93", val_22)
val_23 = Add ("p2o.pd_op.hardswish.15.0", "conv2d_152.w_0_lab_laboffset")
"Add.99" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 240, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [1, 1]> (val_23, "conv2d_152.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.8.0_bias_lab")
val_24 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.99")
"p2o.pd_op.hardswish.16.0" = Mul ("Add.99", val_24)
"Add.105" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.16.0", "conv2d_153.w_0_lab_lab", "p2o.pd_op.conv2d.9.0_bias_lab_lab")
val_25 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.105")
"p2o.pd_op.hardswish.17.0" = Mul ("Add.105", val_25)
val_26 = Add ("p2o.pd_op.hardswish.17.0", "conv2d_154.w_0_lab_laboffset")
"Add.111" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 240, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [1, 1]> (val_26, "conv2d_154.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.9.0_bias_lab")
val_27 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.111")
"p2o.pd_op.hardswish.18.0" = Mul ("Add.111", val_27)
"Add.117" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.18.0", "conv2d_155.w_0_lab_lab", "p2o.pd_op.conv2d.10.0_bias_lab_lab")
val_28 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.117")
"p2o.pd_op.hardswish.19.0" = Mul ("Add.117", val_28)
val_29 = Add ("p2o.pd_op.hardswish.19.0", "conv2d_156.w_0_lab_laboffset")
"Add.123" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 240, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [2, 1]> (val_29, "conv2d_156.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.10.0_bias_lab")
val_30 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.123")
"p2o.pd_op.hardswish.20.0" = Mul ("Add.123", val_30)
["Mul.82"] "Mul.83" = Mul ("learnable_affine_block_41.w_0", "p2o.pd_op.hardswish.20.0")
["Add.124"] "Add.125" = Add ("Mul.83", "learnable_affine_block_41.w_1")
["GlobalAveragePool.0"] "p2o.pd_op.pool2d.0.0" = GlobalAveragePool ("Add.125")
"Add.127" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.pool2d.0.0", "conv2d_96.w_0", "p2o.pd_op.conv2d.11.0_bias")
["Relu.0"] "p2o.pd_op.relu.0.0" = Relu ("Add.127")
"Add.129" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.relu.0.0", "conv2d_97.w_0", "p2o.pd_op.conv2d.12.0_bias")
["HardSigmoid.0"] "p2o.pd_op.hardsigmoid.0.0" = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.129")
["Mul.84"] "Mul.85" = Mul ("Add.125", "p2o.pd_op.hardsigmoid.0.0")
"Add.133" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("Mul.85", "conv2d_157.w_0_lab", "p2o.pd_op.conv2d.13.0_bias_lab")
val_31 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.133")
"p2o.pd_op.hardswish.21.0" = Mul ("Add.133", val_31)
val_32 = Add ("p2o.pd_op.hardswish.21.0", "conv2d_158.w_0_lab_laboffset")
"Add.139" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 480, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [1, 1]> (val_32, "conv2d_158.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.11.0_bias_lab")
val_33 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.139")
"p2o.pd_op.hardswish.22.0" = Mul ("Add.139", val_33)
["Mul.92"] "Mul.93" = Mul ("learnable_affine_block_45.w_0", "p2o.pd_op.hardswish.22.0")
["Add.140"] "Add.141" = Add ("Mul.93", "learnable_affine_block_45.w_1")
["GlobalAveragePool.1"] "p2o.pd_op.pool2d.1.0" = GlobalAveragePool ("Add.141")
"Add.143" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.pool2d.1.0", "conv2d_107.w_0", "p2o.pd_op.conv2d.14.0_bias")
["Relu.1"] "p2o.pd_op.relu.1.0" = Relu ("Add.143")
"Add.145" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.relu.1.0", "conv2d_108.w_0", "p2o.pd_op.conv2d.15.0_bias")
["HardSigmoid.1"] "p2o.pd_op.hardsigmoid.1.0" = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.145")
["Mul.94"] "Mul.95" = Mul ("Add.141", "p2o.pd_op.hardsigmoid.1.0")
"Add.149" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("Mul.95", "conv2d_159.w_0_lab", "p2o.pd_op.conv2d.16.0_bias_lab")
val_34 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.149")
"p2o.pd_op.hardswish.23.0" = Mul ("Add.149", val_34)
val_35 = Add ("p2o.pd_op.hardswish.23.0", "conv2d_160.w_0_lab_laboffset")
"Add.155" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 480, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [2, 1]> (val_35, "conv2d_160.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.12.0_bias_lab")
val_36 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.155")
"p2o.pd_op.hardswish.24.0" = Mul ("Add.155", val_36)
"Add.161" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.24.0", "conv2d_161.w_0_lab_lab", "p2o.pd_op.conv2d.17.0_bias_lab_lab")
val_37 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.161")
"p2o.pd_op.hardswish.25.0" = Mul ("Add.161", val_37)
val_38 = Add ("p2o.pd_op.hardswish.25.0", "conv2d_162.w_0_lab_laboffset")
"Add.167" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 480, kernel_shape: ints = [5, 5], pads: ints = [2, 2, 2, 2], strides: ints = [1, 1]> (val_38, "conv2d_162.w_0_lab_labscale", "p2o.pd_op.depthwise_conv2d.13.0_bias_lab")
val_39 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.167")
"p2o.pd_op.hardswish.26.0" = Mul ("Add.167", val_39)
"Add.173" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.hardswish.26.0", "conv2d_163.w_0_lab_lab", "p2o.pd_op.conv2d.18.0_bias_lab_lab")
val_40 = HardSigmoid <alpha: float = 0.166667, beta: float = 0.5> ("Add.173")
"p2o.pd_op.hardswish.27.0" = Mul ("Add.173", val_40)
val_41 = AveragePool <auto_pad: string = "NOTSET", ceil_mode: int = 0, count_include_pad: int = 0, kernel_shape: ints = [3, 2], pads: ints = [0, 0, 0, 0], strides: ints = [3, 2]> ("p2o.pd_op.hardswish.27.0")
val_42 = Mul (val_41, "learnable_affine_block_55.w_0")
"p2o.pd_op.pool2d.2.0" = Add (val_42, "learnable_affine_block_55.w_1")
"p2o.pd_op.batch_norm_.1.0" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 3], pads: ints = [0, 1, 0, 1], strides: ints = [1, 1]> ("p2o.pd_op.pool2d.2.0", "conv2d_131.w_0", "conv2d_131.w_0_bias")
["Sigmoid.0"] "Sigmoid.1" = Sigmoid ("p2o.pd_op.batch_norm_.1.0")
["Mul.116"] "p2o.pd_op.swish.0.0" = Mul ("p2o.pd_op.batch_norm_.1.0", "Sigmoid.1")
"p2o.pd_op.batch_norm_.2.0" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.swish.0.0", "conv2d_132.w_0", "conv2d_132.w_0_bias")
["Sigmoid.2"] "Sigmoid.3" = Sigmoid ("p2o.pd_op.batch_norm_.2.0")
["Mul.117"] "p2o.pd_op.swish.1.0" = Mul ("p2o.pd_op.batch_norm_.2.0", "Sigmoid.3")
"p2o.pd_op.flatten.0.0" = Squeeze ("p2o.pd_op.swish.1.0", "helper.constant.9")
["Transpose.0"] "p2o.pd_op.transpose.0.0" = Transpose <perm: ints = [0, 2, 1]> ("p2o.pd_op.flatten.0.0")
"Add.181" = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> ("p2o.pd_op.transpose.0.0", "helper.reshape.0", "helper.reshape.1")
["MatMul.0"] "p2o.pd_op.matmul.0.0" = MatMul ("Add.181", "linear_0.w_0_qkv_scaled")
["Add.182"] "Add.183" = Add ("p2o.pd_op.matmul.0.0", "linear_0.b_0_qkv_scaled")
"Add.183_qkv" = Reshape <allowzero: int = 0> ("Add.183", "Add.183_qkv_shape")
"Add.183_qkv_heads" = Transpose <perm: ints = [0, 2, 1, 3]> ("Add.183_qkv")
"Add.183_q", "Add.183_k", "Add.183_v" = Split <axis: int = 1> ("Add.183_qkv_heads", "Add.183_qkv_sizes")
["Transpose.2"] "p2o.pd_op.transpose.2.0" = Transpose <perm: ints = [0, 1, 3, 2]> ("Add.183_k")
["MatMul.1"] "p2o.pd_op.matmul.1.0" = MatMul ("Add.183_q", "p2o.pd_op.transpose.2.0")
["Softmax.0"] "p2o.pd_op.softmax.0.0" = Softmax <axis: int = -1> ("p2o.pd_op.matmul.1.0")
["MatMul.2"] "p2o.pd_op.matmul.2.0" = MatMul ("p2o.pd_op.softmax.0.0", "Add.183_v")
["Transpose.3"] "p2o.pd_op.transpose.3.0" = Transpose <perm: ints = [0, 2, 1, 3]> ("p2o.pd_op.matmul.2.0")
["Reshape.38"] "p2o.pd_op.reshape.33.0" = Reshape <allowzero: int = 0> ("p2o.pd_op.transpose.3.0", "auto.cast.38")
["MatMul.3"] "p2o.pd_op.matmul.3.0" = MatMul ("p2o.pd_op.reshape.33.0", "linear_1.w_0")
["Add.184"] "Add.185" = Add ("p2o.pd_op.matmul.3.0", "linear_1.b_0")
["Add.186"] "Add.187" = Add ("p2o.pd_op.transpose.0.0", "Add.185")
"Add.191" = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> ("Add.187", "helper.reshape.2", "helper.reshape.3")
["MatMul.4"] "p2o.pd_op.matmul.4.0" = MatMul ("Add.191", "linear_2.w_0")
["Add.192"] "Add.193" = Add ("p2o.pd_op.matmul.4.0", "linear_2.b_0")
["Sigmoid.4"] "Sigmoid.5" = Sigmoid ("Add.193")
["Mul.124"] "p2o.pd_op.swish.2.0" = Mul ("Add.193", "Sigmoid.5")
["MatMul.5"] "p2o.pd_op.matmul.5.0" = MatMul ("p2o.pd_op.swish.2.0", "linear_3.w_0")
["Add.194"] "Add.195" = Add ("p2o.pd_op.matmul.5.0", "linear_3.b_0")
["Add.196"] "Add.197" = Add ("Add.187", "Add.195")
"Add.201" = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> ("Add.197", "helper.reshape.4", "helper.reshape.5")
["MatMul.6"] "p2o.pd_op.matmul.6.0" = MatMul ("Add.201", "linear_4.w_0_qkv_scaled")
["Add.202"] "Add.203" = Add ("p2o.pd_op.matmul.6.0", "linear_4.b_0_qkv_scaled")
"Add.203_qkv" = Reshape <allowzero: int = 0> ("Add.203", "Add.183_qkv_shape")
"Add.203_qkv_heads" = Transpose <perm: ints = [0, 2, 1, 3]> ("Add.203_qkv")
"Add.203_q", "Add.203_k", "Add.203_v" = Split <axis: int = 1> ("Add.203_qkv_heads", "Add.183_qkv_sizes")
["Transpose.5"] "p2o.pd_op.transpose.5.0" = Transpose <perm: ints = [0, 1, 3, 2]> ("Add.203_k")
["MatMul.7"] "p2o.pd_op.matmul.7.0" = MatMul ("Add.203_q", "p2o.pd_op.transpose.5.0")
["Softmax.1"] "p2o.pd_op.softmax.1.0" = Softmax <axis: int = -1> ("p2o.pd_op.matmul.7.0")
["MatMul.8"] "p2o.pd_op.matmul.8.0" = MatMul ("p2o.pd_op.softmax.1.0", "Add.203_v")
["Transpose.6"] "p2o.pd_op.transpose.6.0" = Transpose <perm: ints = [0, 2, 1, 3]> ("p2o.pd_op.matmul.8.0")
["Reshape.46"] "p2o.pd_op.reshape.35.0" = Reshape <allowzero: int = 0> ("p2o.pd_op.transpose.6.0", "auto.cast.38")
["MatMul.9"] "p2o.pd_op.matmul.9.0" = MatMul ("p2o.pd_op.reshape.35.0", "linear_5.w_0")
["Add.204"] "Add.205" = Add ("p2o.pd_op.matmul.9.0", "linear_5.b_0")
["Add.206"] "Add.207" = Add ("Add.197", "Add.205")
"Add.211" = LayerNormalization <axis: int = -1, epsilon: float = 1e-05, stash_type: int = 1> ("Add.207", "helper.reshape.6", "helper.reshape.7")
["MatMul.10"] "p2o.pd_op.matmul.10.0" = MatMul ("Add.211", "linear_6.w_0")
["Add.212"] "Add.213" = Add ("p2o.pd_op.matmul.10.0", "linear_6.b_0")
["Sigmoid.6"] "Sigmoid.7" = Sigmoid ("Add.213")
["Mul.131"] "p2o.pd_op.swish.3.0" = Mul ("Add.213", "Sigmoid.7")
["MatMul.11"] "p2o.pd_op.matmul.11.0" = MatMul ("p2o.pd_op.swish.3.0", "linear_7.w_0")
["Add.214"] "Add.215" = Add ("p2o.pd_op.matmul.11.0", "linear_7.b_0")
["Add.216"] "Add.217" = Add ("Add.207", "Add.215")
"Add.221" = LayerNormalization <axis: int = -1, epsilon: float = 1e-06, stash_type: int = 1> ("Add.217", "helper.reshape.8", "helper.reshape.9")
"p2o.pd_op.reshape.36.0" = Unsqueeze ("Add.221", rec_unsqueeze_axis)
["Transpose.7"] "p2o.pd_op.transpose.7.0" = Transpose <perm: ints = [0, 3, 1, 2]> ("p2o.pd_op.reshape.36.0")
"p2o.pd_op.batch_norm_.3.0" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.transpose.7.0", "conv2d_133.w_0", "conv2d_133.w_0_bias")
["Sigmoid.8"] "Sigmoid.9" = Sigmoid ("p2o.pd_op.batch_norm_.3.0")
["Mul.134"] "p2o.pd_op.swish.4.0" = Mul ("p2o.pd_op.batch_norm_.3.0", "Sigmoid.9")
["Concat.3"] "Concat.4" = Concat <axis: int = 1> ("p2o.pd_op.pool2d.2.0", "p2o.pd_op.swish.4.0")
"p2o.pd_op.batch_norm_.4.0" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 3], pads: ints = [0, 1, 0, 1], strides: ints = [1, 1]> ("Concat.4", "conv2d_134.w_0", "conv2d_134.w_0_bias")
["Sigmoid.10"] "Sigmoid.11" = Sigmoid ("p2o.pd_op.batch_norm_.4.0")
["Mul.135"] "p2o.pd_op.swish.5.0" = Mul ("p2o.pd_op.batch_norm_.4.0", "Sigmoid.11")
"p2o.pd_op.batch_norm_.5.0" = Conv <auto_pad: string = "NOTSET", dilations: ints = [1, 1], group: int = 1, kernel_shape: ints = [1, 1], pads: ints = [0, 0, 0, 0], strides: ints = [1, 1]> ("p2o.pd_op.swish.5.0", "conv2d_135.w_0", "conv2d_135.w_0_bias")
["Sigmoid.12"] "Sigmoid.13" = Sigmoid ("p2o.pd_op.batch_norm_.5.0")
["Mul.136"] "p2o.pd_op.swish.6.0" = Mul ("p2o.pd_op.batch_norm_.5.0", "Sigmoid.13")
["Squeeze.8"] "p2o.pd_op.squeeze.0.0" = Squeeze ("p2o.pd_op.swish.6.0", "helper.constant.9")
["Transpose.8"] "p2o.pd_op.transpose.8.0" = Transpose <perm: ints = [0, 2, 1]> ("p2o.pd_op.squeeze.0.0")
["MatMul.12"] "p2o.pd_op.matmul.12.0" = MatMul ("p2o.pd_op.transpose.8.0", "linear_8.w_0")
["Add.222"] "Add.223" = Add ("p2o.pd_op.matmul.12.0", "linear_8.b_0")
[n0_2] tmp_2 = ArgMax <axis: int = 2, keepdims: int = 0> ("Add.223")
[n1_2] ctc_indices = Cast <to: int = 6> (tmp_2)
[n3_2] max_logits = ReduceMax <keepdims: int = 1> ("Add.223", "helper.constant.9")
[n4_2] tmp_0_2 = Sub ("Add.223", max_logits)
[n5] tmp_1 = Exp (tmp_0_2)
[n7] tmp_3 = ReduceSum <keepdims: int = 0> (tmp_1, "helper.constant.9")
[n8] ctc_confidence = Reciprocal (tmp_3)
}
weights:
Add.183_qkv_shape INT64[4] 003762987fc4
Add.183_qkv_sizes INT64[3] 49f92c4a88e0
auto.cast.38 INT64[3] b190f2a09873
const_cast FLOAT[] 9fd754fbfd83
conv2d_0.w_0 FLOAT[16,3,3,3] a93d2a513737
conv2d_0.w_0_bias FLOAT[16] d58e07852b5d
conv2d_107.w_0 FLOAT[120,480,1,1] d285ed78f24e
conv2d_108.w_0 FLOAT[480,120,1,1] 89cbf82f1ef6
conv2d_131.w_0 FLOAT[60,480,1,3] 3c42358e7f9f
conv2d_131.w_0_bias FLOAT[60] 13f7a5429fda
conv2d_132.w_0 FLOAT[120,60,1,1] b8bef0667a08
conv2d_132.w_0_bias FLOAT[120] 5e5f651b146a
conv2d_133.w_0 FLOAT[480,120,1,1] 3238cafb10a9
conv2d_133.w_0_bias FLOAT[480] b06073df2c2b
conv2d_134.w_0 FLOAT[60,960,1,3] 374e0566e87d
conv2d_134.w_0_bias FLOAT[60] ff64978f261f
conv2d_135.w_0 FLOAT[120,60,1,1] 168f23d8322d
conv2d_135.w_0_bias FLOAT[120] af001e17855e
conv2d_136.w_0_lab FLOAT[16,1,3,3] ede0bc053725
conv2d_137.w_0_lab_lab FLOAT[32,16,1,1] 656c4832c084
conv2d_138.w_0_lab_laboffset FLOAT[] 8d1d5f0af9f9
conv2d_138.w_0_lab_labscale FLOAT[32,1,3,3] 4acaa107ebc9
conv2d_139.w_0_lab_lab FLOAT[64,32,1,1] 635a5769d3ee
conv2d_140.w_0_lab_laboffset FLOAT[] 561a12c05d24
conv2d_140.w_0_lab_labscale FLOAT[64,1,3,3] a6b126a9af99
conv2d_141.w_0_lab_lab FLOAT[64,64,1,1] fa0143fe0470
conv2d_142.w_0_lab_laboffset FLOAT[] 71c7653ee4a5
conv2d_142.w_0_lab_labscale FLOAT[64,1,3,3] c167504f3425
conv2d_143.w_0_lab_lab FLOAT[128,64,1,1] 1aa9c3111a47
conv2d_144.w_0_lab_laboffset FLOAT[] 85ff9271354f
conv2d_144.w_0_lab_labscale FLOAT[128,1,3,3] 51da1c5b7ec0
conv2d_145.w_0_lab_lab FLOAT[128,128,1,1] 0ecd5c8d5ae0
conv2d_146.w_0_lab_laboffset FLOAT[] db12b6695ca5
conv2d_146.w_0_lab_labscale FLOAT[128,1,3,3] d0f5a1ce5deb
conv2d_147.w_0_lab_lab FLOAT[240,128,1,1] 473bcb68b3f3
conv2d_148.w_0_lab_laboffset FLOAT[] 386f72b369cc
conv2d_148.w_0_lab_labscale FLOAT[240,1,5,5] 474a885d43f5
conv2d_149.w_0_lab_lab FLOAT[240,240,1,1] d5d35291d572
conv2d_150.w_0_lab_laboffset FLOAT[] be956dc02ca2
conv2d_150.w_0_lab_labscale FLOAT[240,1,5,5] 89f3d39d8d83
conv2d_151.w_0_lab_lab FLOAT[240,240,1,1] b3aa6a8c623d
conv2d_152.w_0_lab_laboffset FLOAT[] 70161ccb53cd
conv2d_152.w_0_lab_labscale FLOAT[240,1,5,5] 2eaf582b4f08
conv2d_153.w_0_lab_lab FLOAT[240,240,1,1] 4db6af902890
conv2d_154.w_0_lab_laboffset FLOAT[] 8b3fd8f3d2b6
conv2d_154.w_0_lab_labscale FLOAT[240,1,5,5] 1f32b42756bf
conv2d_155.w_0_lab_lab FLOAT[240,240,1,1] bbd5c768af3a
conv2d_156.w_0_lab_laboffset FLOAT[] 257d1e3c44a8
conv2d_156.w_0_lab_labscale FLOAT[240,1,5,5] 42a1188db96a
conv2d_157.w_0_lab FLOAT[480,240,1,1] 6d9c4565c413
conv2d_158.w_0_lab_laboffset FLOAT[] 27bf10f54046
conv2d_158.w_0_lab_labscale FLOAT[480,1,5,5] 20cdd2c6a288
conv2d_159.w_0_lab FLOAT[480,480,1,1] 1378b14ce1f5
conv2d_160.w_0_lab_laboffset FLOAT[] ea7a2c1e111f
conv2d_160.w_0_lab_labscale FLOAT[480,1,5,5] 386b4836bc44
conv2d_161.w_0_lab_lab FLOAT[480,480,1,1] 66f7ce693445
conv2d_162.w_0_lab_laboffset FLOAT[] 3a37dd3e85b6
conv2d_162.w_0_lab_labscale FLOAT[480,1,5,5] 02b200e74d26
conv2d_163.w_0_lab_lab FLOAT[480,480,1,1] 65a615717979
conv2d_96.w_0 FLOAT[60,240,1,1] 72c514e97575
conv2d_97.w_0 FLOAT[240,60,1,1] 605da763c133
helper.constant.9 INT64[1] d86e8112f3c4
helper.reshape.0 FLOAT[120] 626968f0a836
helper.reshape.1 FLOAT[120] 874962b7ed74
helper.reshape.2 FLOAT[120] b0c164918fad
helper.reshape.3 FLOAT[120] 9756fa8d9997
helper.reshape.4 FLOAT[120] 5b5aae6c3b2d
helper.reshape.5 FLOAT[120] 89885e9c16b7
helper.reshape.6 FLOAT[120] b34a1b323a6c
helper.reshape.7 FLOAT[120] 973fc0c82d64
helper.reshape.8 FLOAT[120] 75a61d3d0774
helper.reshape.9 FLOAT[120] d6ccc781d652
learnable_affine_block_41.w_0 FLOAT[1] a260fa134c86
learnable_affine_block_41.w_1 FLOAT[1] 6d9f030a30d2
learnable_affine_block_45.w_0 FLOAT[1] fc22eb6ddd89
learnable_affine_block_45.w_1 FLOAT[1] 6d9537a59ec4
learnable_affine_block_55.w_0 FLOAT[1] 7197224fc319
learnable_affine_block_55.w_1 FLOAT[1] f58c89c7725b
linear_0.b_0_qkv_scaled FLOAT[360] ead1e9e9b19e
linear_0.w_0_qkv_scaled FLOAT[120,360] 1a7d997f5a57
linear_1.b_0 FLOAT[120] 880cddbf2c6e
linear_1.w_0 FLOAT[120,120] c9470a150504
linear_2.b_0 FLOAT[240] d81ef6327693
linear_2.w_0 FLOAT[120,240] 47e0ba8e4352
linear_3.b_0 FLOAT[120] b1ffbe179d9f
linear_3.w_0 FLOAT[240,120] c5e7c0e2831a
linear_4.b_0_qkv_scaled FLOAT[360] 208c615e9c55
linear_4.w_0_qkv_scaled FLOAT[120,360] 5c49783e2c00
linear_5.b_0 FLOAT[120] 374a600bd9ef
linear_5.w_0 FLOAT[120,120] b8629393a5f0
linear_6.b_0 FLOAT[240] 35e161071603
linear_6.w_0 FLOAT[120,240] ab85dcf999a9
linear_7.b_0 FLOAT[120] c34b9e042705
linear_7.w_0 FLOAT[240,120] e02cd81d3f67
linear_8.b_0 FLOAT[519] ee35ae5f080c
linear_8.w_0 FLOAT[120,519] c2ac4c825a18
p2o.pd_op.conv2d.1.0_bias_lab_lab FLOAT[32] cc147e45bbc7
p2o.pd_op.conv2d.10.0_bias_lab_lab FLOAT[240] 4451fee6b310
p2o.pd_op.conv2d.11.0_bias FLOAT[60] e7a9a924957c
p2o.pd_op.conv2d.12.0_bias FLOAT[240] 65487d896eca
p2o.pd_op.conv2d.13.0_bias_lab FLOAT[480] 7884a758a49f
p2o.pd_op.conv2d.14.0_bias FLOAT[120] 62ac95f9c0e5
p2o.pd_op.conv2d.15.0_bias FLOAT[480] 30a7bc9eb01f
p2o.pd_op.conv2d.16.0_bias_lab FLOAT[480] 03aa5337a8bf
p2o.pd_op.conv2d.17.0_bias_lab_lab FLOAT[480] 2763d73b8750
p2o.pd_op.conv2d.18.0_bias_lab_lab FLOAT[480] c85032c984da
p2o.pd_op.conv2d.2.0_bias_lab_lab FLOAT[64] 962445bd5233
p2o.pd_op.conv2d.3.0_bias_lab_lab FLOAT[64] 53e51523850e
p2o.pd_op.conv2d.4.0_bias_lab_lab FLOAT[128] 516abbd2e368
p2o.pd_op.conv2d.5.0_bias_lab_lab FLOAT[128] 772d2b4d680a
p2o.pd_op.conv2d.6.0_bias_lab_lab FLOAT[240] 88bb9879c022
p2o.pd_op.conv2d.7.0_bias_lab_lab FLOAT[240] e518a758bf6c
p2o.pd_op.conv2d.8.0_bias_lab_lab FLOAT[240] 592d6046f74a
p2o.pd_op.conv2d.9.0_bias_lab_lab FLOAT[240] 81d52209eeab
p2o.pd_op.depthwise_conv2d.0.0_bias_lab FLOAT[16] 30c3a093a416
p2o.pd_op.depthwise_conv2d.1.0_bias_lab FLOAT[32] 94dde1e19ce0
p2o.pd_op.depthwise_conv2d.10.0_bias_lab FLOAT[240] 0d124d4378e6
p2o.pd_op.depthwise_conv2d.11.0_bias_lab FLOAT[480] b1428c867d72
p2o.pd_op.depthwise_conv2d.12.0_bias_lab FLOAT[480] db2fe2e96a34
p2o.pd_op.depthwise_conv2d.13.0_bias_lab FLOAT[480] 298fb1b13b32
p2o.pd_op.depthwise_conv2d.2.0_bias_lab FLOAT[64] e375d0c77c91
p2o.pd_op.depthwise_conv2d.3.0_bias_lab FLOAT[64] 286f73c9a5ac
p2o.pd_op.depthwise_conv2d.4.0_bias_lab FLOAT[128] 9cb8e6c47ef6
p2o.pd_op.depthwise_conv2d.5.0_bias_lab FLOAT[128] 5cc9a70d387e
p2o.pd_op.depthwise_conv2d.6.0_bias_lab FLOAT[240] 2545e7b1a852
p2o.pd_op.depthwise_conv2d.7.0_bias_lab FLOAT[240] c915fe3596e9
p2o.pd_op.depthwise_conv2d.8.0_bias_lab FLOAT[240] 3ee33fa28bfe
p2o.pd_op.depthwise_conv2d.9.0_bias_lab FLOAT[240] 6d0a712103cd
rec_unsqueeze_axis INT64[1] 7c9fa136d441
Reference in New Issue View Git Blame Copy Permalink