Skip to content
Snippets Groups Projects

GPU backend

Merged GPU backend
gpu-cg into main
Merged prathi3 requested to merge gpu-cg into main
Viewing commit 87b27f34
Prev Next
Show latest version
1 file
+ 0
− 752
Compare changes
  • Side-by-side
  • Inline
test.jn deleted 100644 → 0
+ 0
− 752
fn Flat1d<t:number, n:usize>(X: t[n]) -> t[n] {
return X;
}
fn Flat1dB0<t:number, n:usize>(X: t[1]) -> t[n] {
let Z : t[n];
for i = 0 to n {
Z[i] = X[0];
}
return Z;
}
fn Flat1dI0<t:number, n:usize, p:usize>(X: t[p]) -> t[n * p] {
let Z : t[n * p];
for i = 0 to n {
for j = 0 to p {
Z[i * p + j] = X[j];
}
}
return Z;
}
fn Flat1dI1<t:number, n:usize, p:usize>(X: t[n]) -> t[n * p] {
let Z : t[n * p];
for i = 0 to n {
for j = 0 to p {
Z[i * p + j] = X[i];
}
}
return Z;
}
fn Flat2d<t:number, n:usize, p:usize>(X: t[n, p]) -> t[n * p] {
let Z : t[n * p];
for i = 0 to n {
for j = 0 to p {
Z[i * p + j] = X[i, j];
}
}
return Z;
}
fn Flat2dB0<t:number, n:usize, p:usize>(X: t[1, p]) -> t[n * p] {
let Z : t[n * p];
for i = 0 to n {
for j = 0 to p {
Z[i * p + j] = X[0, j];
}
}
return Z;
}
fn Flat2dB1<t:number, n:usize, p:usize>(X: t[n, 1]) -> t[n * p] {
let Z : t[n * p];
for i = 0 to n {
for j = 0 to p {
Z[i * p + j] = X[i, 0];
}
}
return Z;
}
fn Flat2dI0<t:number, n:usize, p:usize, q:usize>(X: t[p, q]) -> t[n * p * q] {
let Z : t[n * p * q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i * p * q + j * q + k] = X[j, k];
}
}
}
return Z;
}
fn Flat2dI1<t:number, n:usize, p:usize, q:usize>(X: t[n, q]) -> t[n * p * q] {
let Z : t[n * p * q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i * p * q + j * q + k] = X[i, k];
}
}
}
return Z;
}
fn Flat2dI2<t:number, n:usize, p:usize, q:usize>(X: t[n, p]) -> t[n * p * q] {
let Z : t[n * p * q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i * p * q + j * q + k] = X[i, j];
}
}
}
return Z;
}
fn Flat3d<t:number, n:usize, p:usize, q:usize>(X: t[n, p, q]) -> t[n * p * q] {
let Z : t[n * p * q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i * p * q + j * q + k] = X[i, j, k];
}
}
}
return Z;
}
fn Flat3dB0<t:number, n:usize, p:usize, q:usize>(X: t[1, p, q]) -> t[n * p * q] {
let Z : t[n * p * q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i * p * q + j * q + k] = X[0, j, k];
}
}
}
return Z;
}
fn Flat3dB1<t:number, n:usize, p:usize, q:usize>(X: t[n, 1, q]) -> t[n * p * q] {
let Z : t[n * p * q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i * p * q + j * q + k] = X[i, 0, k];
}
}
}
return Z;
}
fn Flat3dB2<t:number, n:usize, p:usize, q:usize>(X: t[n, p, 1]) -> t[n * p * q] {
let Z : t[n * p * q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i * p * q + j * q + k] = X[i, j, 0];
}
}
}
return Z;
}
fn Flat3dI0<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[p, q, r]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[j, k, l];
}
}
}
}
return Z;
}
fn Flat3dI1<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, q, r]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[i, k, l];
}
}
}
}
return Z;
}
fn Flat3dI2<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, r]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[i, j, l];
}
}
}
}
return Z;
}
fn Flat3dI3<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[i, j, k];
}
}
}
}
return Z;
}
fn Flat4d<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, r]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[i, j, k, l];
}
}
}
}
return Z;
}
fn Flat4dB0<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[1, p, q, r]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[0, j, k, l];
}
}
}
}
return Z;
}
fn Flat4dB1<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, 1, q, r]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[i, 0, k, l];
}
}
}
}
return Z;
}
fn Flat4dB2<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, 1, r]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[i, j, 0, l];
}
}
}
}
return Z;
}
fn Flat4dB3<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, 1]) -> t[n * p * q * r] {
let Z : t[n * p * q * r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i * p * q * r + j * q * r + k * r + l] = X[i, j, k, 0];
}
}
}
}
return Z;
}
fn Unflat2d<t:number, n:usize, p:usize>(X: t[n * p]) -> t[n, p] {
let Z : t[n, p];
for i = 0 to n {
for j = 0 to p {
Z[i, j] = X[i * p + j];
}
}
return Z;
}
fn Unflat3d<t:number, n:usize, p:usize, q:usize>(X: t[n * p * q]) -> t[n, p, q] {
let Z : t[n, p, q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i, j, k] = X[i * p * q + j * q + k];
}
}
}
return Z;
}
fn Unflat4d<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n * p * q * r]) -> t[n, p, q, r] {
let Z : t[n, p, q, r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i, j, k, l] = X[i * p * q * r + j * q * r + k * r + l];
}
}
}
}
return Z;
}
fn Permute2d<t:number, n:usize, p:usize>(X: t[n, p]) -> t[p, n] {
let Z : t[p, n];
for i = 0 to n {
for j = 0 to p {
Z[j, i] = X[i, j];
}
}
return Z;
}
fn Permute3d01<t:number, n:usize, p:usize, q:usize>(X: t[n, p, q]) -> t[p, n, q] {
let Z : t[p, n, q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[j, i, k] = X[i, j, k];
}
}
}
return Z;
}
fn Permute3d02<t:number, n:usize, p:usize, q:usize>(X: t[n, p, q]) -> t[q, p, n] {
let Z : t[q, p, n];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[k, j, i] = X[i, j, k];
}
}
}
return Z;
}
fn Permute3d12<t:number, n:usize, p:usize, q:usize>(X: t[n, p, q]) -> t[n, q, p] {
let Z : t[n, q, p];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
Z[i, k, j] = X[i, j, k];
}
}
}
return Z;
}
fn Permute4d01<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, r]) -> t[p, n, q, r] {
let Z : t[p, n, q, r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[j, i, k, l] = X[i, j, k, l];
}
}
}
}
return Z;
}
fn Permute4d02<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, r]) -> t[q, p, n, r] {
let Z : t[q, p, n, r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[k, j, i, l] = X[i, j, k, l];
}
}
}
}
return Z;
}
fn Permute4d03<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, r]) -> t[r, p, q, n] {
let Z : t[r, p, q, n];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[l, j, k, i] = X[i, j, k, l];
}
}
}
}
return Z;
}
fn Permute4d12<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, r]) -> t[n, q, p, r] {
let Z : t[n, q, p, r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i, k, j, l] = X[i, j, k, l];
}
}
}
}
return Z;
}
fn Permute4d13<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, r]) -> t[n, r, q, p] {
let Z : t[n, r, q, p];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i, l, k, j] = X[i, j, k, l];
}
}
}
}
return Z;
}
fn Permute4d23<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q, r]) -> t[n, p, r, q] {
let Z : t[n, p, r, q];
for i = 0 to n {
for j = 0 to p {
for k = 0 to q {
for l = 0 to r {
Z[i, j, l, k] = X[i, j, k, l];
}
}
}
}
return Z;
}
fn Add<t:number, n:usize>(X: t[n], Y: t[n]) -> t[n] {
let Z : t[n];
for i = 0 to n {
Z[i] = X[i] + Y[i];
}
return Z;
}
fn AddConst<t:number, n:usize>(X: t[n], Y: t) -> t[n] {
let Z : t[n];
for i = 0 to n {
Z[i] = X[i] + Y;
}
return Z;
}
fn Mul<t:number, n:usize>(X: t[n], Y: t[n]) -> t[n] {
let Z : t[n];
for i = 0 to n {
Z[i] = X[i] * Y[i];
}
return Z;
}
fn MulConst<t:number, n:usize>(X: t[n], Y: t) -> t[n] {
let Z : t[n];
for i = 0 to n {
Z[i] = X[i] * Y;
}
return Z;
}
fn HardTanh<t:number, n:usize>(X: t[n], Min: t, Max: t) -> t[n] {
let Z : t[n];
for i = 0 to n {
Z[i] = max!::<t>(Min, min!::<t>(X[i], Max));
}
return Z;
}
fn Mean<t: number, n:usize, p:usize>(X: t[n, p]) -> t[n] {
let Z : t[n];
for i = 0 to n {
let sum : t = 0;
for j = 0 to p {
sum += X[i, j];
}
Z[i] = sum / (p as t);
}
return Z;
}
fn MeanKD<t: number, n:usize, p:usize>(X: t[n, p]) -> t[n, p] {
let Z : t[n, p];
for i = 0 to n {
let sum : t = 0;
for j = 0 to p {
sum += X[i, j];
}
for j = 0 to p {
Z[i, j] = sum / (p as t);
}
}
return Z;
}
fn BatchNorm<t: number, n:usize, p:usize>(X: t[n, p], W: t[n], B: t[n], M: t[n], V: t[n], E: t) -> t[n, p] {
let Z : t[n, p];
for i = 0 to n {
for j = 0 to p {
Z[i, j] = W[i] * (X[i, j] - M[i]) / (sqrt!::<f32>((V[i] + E) as f32) as t) + B[i];
}
}
return Z;
}
fn Conv2d<t:number, m:usize, n:usize, p:usize, q:usize, r:usize, s:usize, u:usize, v:usize, w:usize, stride1:usize, stride2:usize, padding1:usize, padding2:usize, dilation1:usize, dilation2:usize, output_padding1:usize, output_padding2:usize, groups:usize>(X: t[m, n, p, q], W: t[r, n, s, u]) -> t[m, r, v, w] {
let Z : t[m, r, v, w];
for i = 0 to m {
for j = 0 to r {
let C: t[v, w];
for k = 0 to v {
for l = 0 to w {
C[k, l] = 0;
}
}
let gid = (j * groups) / r;
for k = 0 to (n / groups) {
for l = 0 to s {
let real_l : usize = l * dilation1;
let ih : usize = (padding1 / stride1 + 1) * stride1 - padding1 + real_l;
let oh : usize = (padding1 / stride1 + 1);
let numh : usize = min!::<usize>(v - oh, (p - ih) / stride1 + 1);
for g = 0 to u {
let real_g : usize = g * dilation2;
let iw : usize = (padding2 / stride2 + 1) * stride2 - padding2 + real_g;
let ow : usize = (padding2 / stride2 + 1);
let numw : usize = min!::<usize>(w - ow, (q - iw) / stride2 + 1);
for iterh = 0 to numh {
for iterw = 0 to numw {
C[output_padding1 + oh + iterh, output_padding2 + ow + iterw] += X[i, (gid * n) / groups + k, ih + iterh * stride1, iw + iterw * stride2] * W[j, k, l, g];
}
}
}
}
}
for k = 0 to v {
for l = 0 to w {
Z[i, j, k, l] = C[k, l];
}
}
}
}
return Z;
}
fn BatchMatMul<t:number, n:usize, p:usize, q:usize, r:usize>(X: t[n, p, q], Y: t[n, q, r]) -> t[n, p, r] {
let Z : t[n, p, r];
for i = 0 to n {
for j = 0 to p {
for k = 0 to r {
Z[i, j, k] = 0;
for l = 0 to q {
Z[i, j, k] += X[i, j, l] * Y[i, l, k];
}
}
}
}
return Z;
}
fn MatMul<t:number, n:usize, p:usize, q:usize>(X: t[n, p], Y: t[p, q]) -> t[n, q] {
let Z : t[n, q];
for i = 0 to n {
for j = 0 to q {
Z[i, j] = 0;
for k = 0 to p {
Z[i, j] += X[i, k] * Y[k, j];
}
}
}
return Z;
}
#[entry]
fn main(arg0_1: f32[32, 1, 3, 3], arg1_1: f32[32], arg2_1: f32[32], arg3_1: f32[32, 1, 3, 3], arg4_1: f32[32], arg5_1: f32[32], arg6_1: f32[16, 32, 1, 1], arg7_1: f32[16], arg8_1: f32[16], arg9_1: f32[96, 16, 1, 1], arg10_1: f32[96], arg11_1: f32[96], arg12_1: f32[96, 1, 3, 3], arg13_1: f32[96], arg14_1: f32[96], arg15_1: f32[24, 96, 1, 1], arg16_1: f32[24], arg17_1: f32[24], arg18_1: f32[144, 24, 1, 1], arg19_1: f32[144], arg20_1: f32[144], arg21_1: f32[144, 1, 3, 3], arg22_1: f32[144], arg23_1: f32[144], arg24_1: f32[24, 144, 1, 1], arg25_1: f32[24], arg26_1: f32[24], arg27_1: f32[144, 24, 1, 1], arg28_1: f32[144], arg29_1: f32[144], arg30_1: f32[144, 1, 3, 3], arg31_1: f32[144], arg32_1: f32[144], arg33_1: f32[32, 144, 1, 1], arg34_1: f32[32], arg35_1: f32[32], arg36_1: f32[192, 32, 1, 1], arg37_1: f32[192], arg38_1: f32[192], arg39_1: f32[192, 1, 3, 3], arg40_1: f32[192], arg41_1: f32[192], arg42_1: f32[32, 192, 1, 1], arg43_1: f32[32], arg44_1: f32[32], arg45_1: f32[192, 32, 1, 1], arg46_1: f32[192], arg47_1: f32[192], arg48_1: f32[192, 1, 3, 3], arg49_1: f32[192], arg50_1: f32[192], arg51_1: f32[32, 192, 1, 1], arg52_1: f32[32], arg53_1: f32[32], arg54_1: f32[192, 32, 1, 1], arg55_1: f32[192], arg56_1: f32[192], arg57_1: f32[192, 1, 3, 3], arg58_1: f32[192], arg59_1: f32[192], arg60_1: f32[64, 192, 1, 1], arg61_1: f32[64], arg62_1: f32[64], arg63_1: f32[384, 64, 1, 1], arg64_1: f32[384], arg65_1: f32[384], arg66_1: f32[384, 1, 3, 3], arg67_1: f32[384], arg68_1: f32[384], arg69_1: f32[64, 384, 1, 1], arg70_1: f32[64], arg71_1: f32[64], arg72_1: f32[384, 64, 1, 1], arg73_1: f32[384], arg74_1: f32[384], arg75_1: f32[384, 1, 3, 3], arg76_1: f32[384], arg77_1: f32[384], arg78_1: f32[64, 384, 1, 1], arg79_1: f32[64], arg80_1: f32[64], arg81_1: f32[384, 64, 1, 1], arg82_1: f32[384], arg83_1: f32[384], arg84_1: f32[384, 1, 3, 3], arg85_1: f32[384], arg86_1: f32[384], arg87_1: f32[64, 384, 1, 1], arg88_1: f32[64], arg89_1: f32[64], arg90_1: f32[384, 64, 1, 1], arg91_1: f32[384], arg92_1: f32[384], arg93_1: f32[384, 1, 3, 3], arg94_1: f32[384], arg95_1: f32[384], arg96_1: f32[96, 384, 1, 1], arg97_1: f32[96], arg98_1: f32[96], arg99_1: f32[576, 96, 1, 1], arg100_1: f32[576], arg101_1: f32[576], arg102_1: f32[576, 1, 3, 3], arg103_1: f32[576], arg104_1: f32[576], arg105_1: f32[96, 576, 1, 1], arg106_1: f32[96], arg107_1: f32[96], arg108_1: f32[576, 96, 1, 1], arg109_1: f32[576], arg110_1: f32[576], arg111_1: f32[576, 1, 3, 3], arg112_1: f32[576], arg113_1: f32[576], arg114_1: f32[96, 576, 1, 1], arg115_1: f32[96], arg116_1: f32[96], arg117_1: f32[576, 96, 1, 1], arg118_1: f32[576], arg119_1: f32[576], arg120_1: f32[576, 1, 3, 3], arg121_1: f32[576], arg122_1: f32[576], arg123_1: f32[160, 576, 1, 1], arg124_1: f32[160], arg125_1: f32[160], arg126_1: f32[960, 160, 1, 1], arg127_1: f32[960], arg128_1: f32[960], arg129_1: f32[960, 1, 3, 3], arg130_1: f32[960], arg131_1: f32[960], arg132_1: f32[160, 960, 1, 1], arg133_1: f32[160], arg134_1: f32[160], arg135_1: f32[960, 160, 1, 1], arg136_1: f32[960], arg137_1: f32[960], arg138_1: f32[960, 1, 3, 3], arg139_1: f32[960], arg140_1: f32[960], arg141_1: f32[160, 960, 1, 1], arg142_1: f32[160], arg143_1: f32[160], arg144_1: f32[960, 160, 1, 1], arg145_1: f32[960], arg146_1: f32[960], arg147_1: f32[960, 1, 3, 3], arg148_1: f32[960], arg149_1: f32[960], arg150_1: f32[320, 960, 1, 1], arg151_1: f32[320], arg152_1: f32[320], arg153_1: f32[1280, 320, 1, 1], arg154_1: f32[1280], arg155_1: f32[1280], arg156_1: f32[10, 1280], arg157_1: f32[10], arg158_1: f32[32], arg159_1: f32[32], arg160_1: i64[], arg161_1: f32[32], arg162_1: f32[32], arg163_1: i64[], arg164_1: f32[16], arg165_1: f32[16], arg166_1: i64[], arg167_1: f32[96], arg168_1: f32[96], arg169_1: i64[], arg170_1: f32[96], arg171_1: f32[96], arg172_1: i64[], arg173_1: f32[24], arg174_1: f32[24], arg175_1: i64[], arg176_1: f32[144], arg177_1: f32[144], arg178_1: i64[], arg179_1: f32[144], arg180_1: f32[144], arg181_1: i64[], arg182_1: f32[24], arg183_1: f32[24], arg184_1: i64[], arg185_1: f32[144], arg186_1: f32[144], arg187_1: i64[], arg188_1: f32[144], arg189_1: f32[144], arg190_1: i64[], arg191_1: f32[32], arg192_1: f32[32], arg193_1: i64[], arg194_1: f32[192], arg195_1: f32[192], arg196_1: i64[], arg197_1: f32[192], arg198_1: f32[192], arg199_1: i64[], arg200_1: f32[32], arg201_1: f32[32], arg202_1: i64[], arg203_1: f32[192], arg204_1: f32[192], arg205_1: i64[], arg206_1: f32[192], arg207_1: f32[192], arg208_1: i64[], arg209_1: f32[32], arg210_1: f32[32], arg211_1: i64[], arg212_1: f32[192], arg213_1: f32[192], arg214_1: i64[], arg215_1: f32[192], arg216_1: f32[192], arg217_1: i64[], arg218_1: f32[64], arg219_1: f32[64], arg220_1: i64[], arg221_1: f32[384], arg222_1: f32[384], arg223_1: i64[], arg224_1: f32[384], arg225_1: f32[384], arg226_1: i64[], arg227_1: f32[64], arg228_1: f32[64], arg229_1: i64[], arg230_1: f32[384], arg231_1: f32[384], arg232_1: i64[], arg233_1: f32[384], arg234_1: f32[384], arg235_1: i64[], arg236_1: f32[64], arg237_1: f32[64], arg238_1: i64[], arg239_1: f32[384], arg240_1: f32[384], arg241_1: i64[], arg242_1: f32[384], arg243_1: f32[384], arg244_1: i64[], arg245_1: f32[64], arg246_1: f32[64], arg247_1: i64[], arg248_1: f32[384], arg249_1: f32[384], arg250_1: i64[], arg251_1: f32[384], arg252_1: f32[384], arg253_1: i64[], arg254_1: f32[96], arg255_1: f32[96], arg256_1: i64[], arg257_1: f32[576], arg258_1: f32[576], arg259_1: i64[], arg260_1: f32[576], arg261_1: f32[576], arg262_1: i64[], arg263_1: f32[96], arg264_1: f32[96], arg265_1: i64[], arg266_1: f32[576], arg267_1: f32[576], arg268_1: i64[], arg269_1: f32[576], arg270_1: f32[576], arg271_1: i64[], arg272_1: f32[96], arg273_1: f32[96], arg274_1: i64[], arg275_1: f32[576], arg276_1: f32[576], arg277_1: i64[], arg278_1: f32[576], arg279_1: f32[576], arg280_1: i64[], arg281_1: f32[160], arg282_1: f32[160], arg283_1: i64[], arg284_1: f32[960], arg285_1: f32[960], arg286_1: i64[], arg287_1: f32[960], arg288_1: f32[960], arg289_1: i64[], arg290_1: f32[160], arg291_1: f32[160], arg292_1: i64[], arg293_1: f32[960], arg294_1: f32[960], arg295_1: i64[], arg296_1: f32[960], arg297_1: f32[960], arg298_1: i64[], arg299_1: f32[160], arg300_1: f32[160], arg301_1: i64[], arg302_1: f32[960], arg303_1: f32[960], arg304_1: i64[], arg305_1: f32[960], arg306_1: f32[960], arg307_1: i64[], arg308_1: f32[320], arg309_1: f32[320], arg310_1: i64[], arg311_1: f32[1280], arg312_1: f32[1280], arg313_1: i64[], arg314_1: f32[1, 1, 28, 28]) -> (f32[1, 10]) {
let convolution = Conv2d::<f32, 1, 1, 28, 28, 32, 3, 3, 15, 15, 2, 2, 1, 1, 1, 1, 0, 0, 1>(arg314_1, arg0_1);
let getitem_0 = Flat4d::<f32, 1, 32, 14, 14>(getitem);
let hardtanh_0 = HardTanh::<f32, 6272>(getitem_0, 0, 6);
let hardtanh = Unflat4d::<f32, 1, 32, 14, 14>(hardtanh_0);
let convolution_1 = Conv2d::<f32, 1, 32, 14, 14, 32, 3, 3, 15, 15, 1, 1, 1, 1, 1, 1, 0, 0, 32>(hardtanh, arg3_1);
let getitem_1_0 = Flat4d::<f32, 1, 32, 14, 14>(getitem_1);
let hardtanh_1_0 = HardTanh::<f32, 6272>(getitem_1_0, 0, 6);
let hardtanh_1 = Unflat4d::<f32, 1, 32, 14, 14>(hardtanh_1_0);
let convolution_2 = Conv2d::<f32, 1, 32, 14, 14, 16, 1, 1, 15, 15, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_1, arg6_1);
let convolution_3 = Conv2d::<f32, 1, 16, 14, 14, 96, 1, 1, 15, 15, 1, 1, 0, 0, 1, 1, 0, 0, 1>(getitem_2, arg9_1);
let getitem_3_0 = Flat4d::<f32, 1, 96, 14, 14>(getitem_3);
let hardtanh_2_0 = HardTanh::<f32, 18816>(getitem_3_0, 0, 6);
let hardtanh_2 = Unflat4d::<f32, 1, 96, 14, 14>(hardtanh_2_0);
let convolution_4 = Conv2d::<f32, 1, 96, 14, 14, 96, 3, 3, 8, 8, 2, 2, 1, 1, 1, 1, 0, 0, 96>(hardtanh_2, arg12_1);
let getitem_4_0 = Flat4d::<f32, 1, 96, 7, 7>(getitem_4);
let hardtanh_3_0 = HardTanh::<f32, 4704>(getitem_4_0, 0, 6);
let hardtanh_3 = Unflat4d::<f32, 1, 96, 7, 7>(hardtanh_3_0);
let convolution_5 = Conv2d::<f32, 1, 96, 7, 7, 24, 1, 1, 8, 8, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_3, arg15_1);
let convolution_6 = Conv2d::<f32, 1, 24, 7, 7, 144, 1, 1, 8, 8, 1, 1, 0, 0, 1, 1, 0, 0, 1>(getitem_5, arg18_1);
let getitem_6_0 = Flat4d::<f32, 1, 144, 7, 7>(getitem_6);
let hardtanh_4_0 = HardTanh::<f32, 7056>(getitem_6_0, 0, 6);
let hardtanh_4 = Unflat4d::<f32, 1, 144, 7, 7>(hardtanh_4_0);
let convolution_7 = Conv2d::<f32, 1, 144, 7, 7, 144, 3, 3, 8, 8, 1, 1, 1, 1, 1, 1, 0, 0, 144>(hardtanh_4, arg21_1);
let getitem_7_0 = Flat4d::<f32, 1, 144, 7, 7>(getitem_7);
let hardtanh_5_0 = HardTanh::<f32, 7056>(getitem_7_0, 0, 6);
let hardtanh_5 = Unflat4d::<f32, 1, 144, 7, 7>(hardtanh_5_0);
let convolution_8 = Conv2d::<f32, 1, 144, 7, 7, 24, 1, 1, 8, 8, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_5, arg24_1);
let getitem_5_0 = Flat4d::<f32, 1, 24, 7, 7>(getitem_5);
let getitem_8_0 = Flat4d::<f32, 1, 24, 7, 7>(getitem_8);
let add_0 = Add::<f32, 1176>(getitem_5_0, getitem_8_0);
let add = Unflat4d::<f32, 1, 24, 7, 7>(add_0);
let convolution_9 = Conv2d::<f32, 1, 24, 7, 7, 144, 1, 1, 8, 8, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add, arg27_1);
let getitem_9_0 = Flat4d::<f32, 1, 144, 7, 7>(getitem_9);
let hardtanh_6_0 = HardTanh::<f32, 7056>(getitem_9_0, 0, 6);
let hardtanh_6 = Unflat4d::<f32, 1, 144, 7, 7>(hardtanh_6_0);
let convolution_10 = Conv2d::<f32, 1, 144, 7, 7, 144, 3, 3, 4, 4, 2, 2, 1, 1, 1, 1, 0, 0, 144>(hardtanh_6, arg30_1);
let getitem_10_0 = Flat4d::<f32, 1, 144, 4, 4>(getitem_10);
let hardtanh_7_0 = HardTanh::<f32, 2304>(getitem_10_0, 0, 6);
let hardtanh_7 = Unflat4d::<f32, 1, 144, 4, 4>(hardtanh_7_0);
let convolution_11 = Conv2d::<f32, 1, 144, 4, 4, 32, 1, 1, 5, 5, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_7, arg33_1);
let convolution_12 = Conv2d::<f32, 1, 32, 4, 4, 192, 1, 1, 5, 5, 1, 1, 0, 0, 1, 1, 0, 0, 1>(getitem_11, arg36_1);
let getitem_12_0 = Flat4d::<f32, 1, 192, 4, 4>(getitem_12);
let hardtanh_8_0 = HardTanh::<f32, 3072>(getitem_12_0, 0, 6);
let hardtanh_8 = Unflat4d::<f32, 1, 192, 4, 4>(hardtanh_8_0);
let convolution_13 = Conv2d::<f32, 1, 192, 4, 4, 192, 3, 3, 5, 5, 1, 1, 1, 1, 1, 1, 0, 0, 192>(hardtanh_8, arg39_1);
let getitem_13_0 = Flat4d::<f32, 1, 192, 4, 4>(getitem_13);
let hardtanh_9_0 = HardTanh::<f32, 3072>(getitem_13_0, 0, 6);
let hardtanh_9 = Unflat4d::<f32, 1, 192, 4, 4>(hardtanh_9_0);
let convolution_14 = Conv2d::<f32, 1, 192, 4, 4, 32, 1, 1, 5, 5, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_9, arg42_1);
let getitem_11_0 = Flat4d::<f32, 1, 32, 4, 4>(getitem_11);
let getitem_14_0 = Flat4d::<f32, 1, 32, 4, 4>(getitem_14);
let add_1_0 = Add::<f32, 512>(getitem_11_0, getitem_14_0);
let add_1 = Unflat4d::<f32, 1, 32, 4, 4>(add_1_0);
let convolution_15 = Conv2d::<f32, 1, 32, 4, 4, 192, 1, 1, 5, 5, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_1, arg45_1);
let getitem_15_0 = Flat4d::<f32, 1, 192, 4, 4>(getitem_15);
let hardtanh_10_0 = HardTanh::<f32, 3072>(getitem_15_0, 0, 6);
let hardtanh_10 = Unflat4d::<f32, 1, 192, 4, 4>(hardtanh_10_0);
let convolution_16 = Conv2d::<f32, 1, 192, 4, 4, 192, 3, 3, 5, 5, 1, 1, 1, 1, 1, 1, 0, 0, 192>(hardtanh_10, arg48_1);
let getitem_16_0 = Flat4d::<f32, 1, 192, 4, 4>(getitem_16);
let hardtanh_11_0 = HardTanh::<f32, 3072>(getitem_16_0, 0, 6);
let hardtanh_11 = Unflat4d::<f32, 1, 192, 4, 4>(hardtanh_11_0);
let convolution_17 = Conv2d::<f32, 1, 192, 4, 4, 32, 1, 1, 5, 5, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_11, arg51_1);
let add_1_0 = Flat4d::<f32, 1, 32, 4, 4>(add_1);
let getitem_17_0 = Flat4d::<f32, 1, 32, 4, 4>(getitem_17);
let add_2_0 = Add::<f32, 512>(add_1_0, getitem_17_0);
let add_2 = Unflat4d::<f32, 1, 32, 4, 4>(add_2_0);
let convolution_18 = Conv2d::<f32, 1, 32, 4, 4, 192, 1, 1, 5, 5, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_2, arg54_1);
let getitem_18_0 = Flat4d::<f32, 1, 192, 4, 4>(getitem_18);
let hardtanh_12_0 = HardTanh::<f32, 3072>(getitem_18_0, 0, 6);
let hardtanh_12 = Unflat4d::<f32, 1, 192, 4, 4>(hardtanh_12_0);
let convolution_19 = Conv2d::<f32, 1, 192, 4, 4, 192, 3, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 192>(hardtanh_12, arg57_1);
let getitem_19_0 = Flat4d::<f32, 1, 192, 2, 2>(getitem_19);
let hardtanh_13_0 = HardTanh::<f32, 768>(getitem_19_0, 0, 6);
let hardtanh_13 = Unflat4d::<f32, 1, 192, 2, 2>(hardtanh_13_0);
let convolution_20 = Conv2d::<f32, 1, 192, 2, 2, 64, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_13, arg60_1);
let convolution_21 = Conv2d::<f32, 1, 64, 2, 2, 384, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(getitem_20, arg63_1);
let getitem_21_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_21);
let hardtanh_14_0 = HardTanh::<f32, 1536>(getitem_21_0, 0, 6);
let hardtanh_14 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_14_0);
let convolution_22 = Conv2d::<f32, 1, 384, 2, 2, 384, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 0, 0, 384>(hardtanh_14, arg66_1);
let getitem_22_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_22);
let hardtanh_15_0 = HardTanh::<f32, 1536>(getitem_22_0, 0, 6);
let hardtanh_15 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_15_0);
let convolution_23 = Conv2d::<f32, 1, 384, 2, 2, 64, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_15, arg69_1);
let getitem_20_0 = Flat4d::<f32, 1, 64, 2, 2>(getitem_20);
let getitem_23_0 = Flat4d::<f32, 1, 64, 2, 2>(getitem_23);
let add_3_0 = Add::<f32, 256>(getitem_20_0, getitem_23_0);
let add_3 = Unflat4d::<f32, 1, 64, 2, 2>(add_3_0);
let convolution_24 = Conv2d::<f32, 1, 64, 2, 2, 384, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_3, arg72_1);
let getitem_24_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_24);
let hardtanh_16_0 = HardTanh::<f32, 1536>(getitem_24_0, 0, 6);
let hardtanh_16 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_16_0);
let convolution_25 = Conv2d::<f32, 1, 384, 2, 2, 384, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 0, 0, 384>(hardtanh_16, arg75_1);
let getitem_25_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_25);
let hardtanh_17_0 = HardTanh::<f32, 1536>(getitem_25_0, 0, 6);
let hardtanh_17 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_17_0);
let convolution_26 = Conv2d::<f32, 1, 384, 2, 2, 64, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_17, arg78_1);
let add_3_0 = Flat4d::<f32, 1, 64, 2, 2>(add_3);
let getitem_26_0 = Flat4d::<f32, 1, 64, 2, 2>(getitem_26);
let add_4_0 = Add::<f32, 256>(add_3_0, getitem_26_0);
let add_4 = Unflat4d::<f32, 1, 64, 2, 2>(add_4_0);
let convolution_27 = Conv2d::<f32, 1, 64, 2, 2, 384, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_4, arg81_1);
let getitem_27_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_27);
let hardtanh_18_0 = HardTanh::<f32, 1536>(getitem_27_0, 0, 6);
let hardtanh_18 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_18_0);
let convolution_28 = Conv2d::<f32, 1, 384, 2, 2, 384, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 0, 0, 384>(hardtanh_18, arg84_1);
let getitem_28_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_28);
let hardtanh_19_0 = HardTanh::<f32, 1536>(getitem_28_0, 0, 6);
let hardtanh_19 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_19_0);
let convolution_29 = Conv2d::<f32, 1, 384, 2, 2, 64, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_19, arg87_1);
let add_4_0 = Flat4d::<f32, 1, 64, 2, 2>(add_4);
let getitem_29_0 = Flat4d::<f32, 1, 64, 2, 2>(getitem_29);
let add_5_0 = Add::<f32, 256>(add_4_0, getitem_29_0);
let add_5 = Unflat4d::<f32, 1, 64, 2, 2>(add_5_0);
let convolution_30 = Conv2d::<f32, 1, 64, 2, 2, 384, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_5, arg90_1);
let getitem_30_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_30);
let hardtanh_20_0 = HardTanh::<f32, 1536>(getitem_30_0, 0, 6);
let hardtanh_20 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_20_0);
let convolution_31 = Conv2d::<f32, 1, 384, 2, 2, 384, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 0, 0, 384>(hardtanh_20, arg93_1);
let getitem_31_0 = Flat4d::<f32, 1, 384, 2, 2>(getitem_31);
let hardtanh_21_0 = HardTanh::<f32, 1536>(getitem_31_0, 0, 6);
let hardtanh_21 = Unflat4d::<f32, 1, 384, 2, 2>(hardtanh_21_0);
let convolution_32 = Conv2d::<f32, 1, 384, 2, 2, 96, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_21, arg96_1);
let convolution_33 = Conv2d::<f32, 1, 96, 2, 2, 576, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(getitem_32, arg99_1);
let getitem_33_0 = Flat4d::<f32, 1, 576, 2, 2>(getitem_33);
let hardtanh_22_0 = HardTanh::<f32, 2304>(getitem_33_0, 0, 6);
let hardtanh_22 = Unflat4d::<f32, 1, 576, 2, 2>(hardtanh_22_0);
let convolution_34 = Conv2d::<f32, 1, 576, 2, 2, 576, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 0, 0, 576>(hardtanh_22, arg102_1);
let getitem_34_0 = Flat4d::<f32, 1, 576, 2, 2>(getitem_34);
let hardtanh_23_0 = HardTanh::<f32, 2304>(getitem_34_0, 0, 6);
let hardtanh_23 = Unflat4d::<f32, 1, 576, 2, 2>(hardtanh_23_0);
let convolution_35 = Conv2d::<f32, 1, 576, 2, 2, 96, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_23, arg105_1);
let getitem_32_0 = Flat4d::<f32, 1, 96, 2, 2>(getitem_32);
let getitem_35_0 = Flat4d::<f32, 1, 96, 2, 2>(getitem_35);
let add_6_0 = Add::<f32, 384>(getitem_32_0, getitem_35_0);
let add_6 = Unflat4d::<f32, 1, 96, 2, 2>(add_6_0);
let convolution_36 = Conv2d::<f32, 1, 96, 2, 2, 576, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_6, arg108_1);
let getitem_36_0 = Flat4d::<f32, 1, 576, 2, 2>(getitem_36);
let hardtanh_24_0 = HardTanh::<f32, 2304>(getitem_36_0, 0, 6);
let hardtanh_24 = Unflat4d::<f32, 1, 576, 2, 2>(hardtanh_24_0);
let convolution_37 = Conv2d::<f32, 1, 576, 2, 2, 576, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 0, 0, 576>(hardtanh_24, arg111_1);
let getitem_37_0 = Flat4d::<f32, 1, 576, 2, 2>(getitem_37);
let hardtanh_25_0 = HardTanh::<f32, 2304>(getitem_37_0, 0, 6);
let hardtanh_25 = Unflat4d::<f32, 1, 576, 2, 2>(hardtanh_25_0);
let convolution_38 = Conv2d::<f32, 1, 576, 2, 2, 96, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_25, arg114_1);
let add_6_0 = Flat4d::<f32, 1, 96, 2, 2>(add_6);
let getitem_38_0 = Flat4d::<f32, 1, 96, 2, 2>(getitem_38);
let add_7_0 = Add::<f32, 384>(add_6_0, getitem_38_0);
let add_7 = Unflat4d::<f32, 1, 96, 2, 2>(add_7_0);
let convolution_39 = Conv2d::<f32, 1, 96, 2, 2, 576, 1, 1, 3, 3, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_7, arg117_1);
let getitem_39_0 = Flat4d::<f32, 1, 576, 2, 2>(getitem_39);
let hardtanh_26_0 = HardTanh::<f32, 2304>(getitem_39_0, 0, 6);
let hardtanh_26 = Unflat4d::<f32, 1, 576, 2, 2>(hardtanh_26_0);
let convolution_40 = Conv2d::<f32, 1, 576, 2, 2, 576, 3, 3, 2, 2, 2, 2, 1, 1, 1, 1, 0, 0, 576>(hardtanh_26, arg120_1);
let getitem_40_0 = Flat4d::<f32, 1, 576, 1, 1>(getitem_40);
let hardtanh_27_0 = HardTanh::<f32, 576>(getitem_40_0, 0, 6);
let hardtanh_27 = Unflat4d::<f32, 1, 576, 1, 1>(hardtanh_27_0);
let convolution_41 = Conv2d::<f32, 1, 576, 1, 1, 160, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_27, arg123_1);
let convolution_42 = Conv2d::<f32, 1, 160, 1, 1, 960, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(getitem_41, arg126_1);
let getitem_42_0 = Flat4d::<f32, 1, 960, 1, 1>(getitem_42);
let hardtanh_28_0 = HardTanh::<f32, 960>(getitem_42_0, 0, 6);
let hardtanh_28 = Unflat4d::<f32, 1, 960, 1, 1>(hardtanh_28_0);
let convolution_43 = Conv2d::<f32, 1, 960, 1, 1, 960, 3, 3, 2, 2, 1, 1, 1, 1, 1, 1, 0, 0, 960>(hardtanh_28, arg129_1);
let getitem_43_0 = Flat4d::<f32, 1, 960, 1, 1>(getitem_43);
let hardtanh_29_0 = HardTanh::<f32, 960>(getitem_43_0, 0, 6);
let hardtanh_29 = Unflat4d::<f32, 1, 960, 1, 1>(hardtanh_29_0);
let convolution_44 = Conv2d::<f32, 1, 960, 1, 1, 160, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_29, arg132_1);
let getitem_41_0 = Flat4d::<f32, 1, 160, 1, 1>(getitem_41);
let getitem_44_0 = Flat4d::<f32, 1, 160, 1, 1>(getitem_44);
let add_8_0 = Add::<f32, 160>(getitem_41_0, getitem_44_0);
let add_8 = Unflat4d::<f32, 1, 160, 1, 1>(add_8_0);
let convolution_45 = Conv2d::<f32, 1, 160, 1, 1, 960, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_8, arg135_1);
let getitem_45_0 = Flat4d::<f32, 1, 960, 1, 1>(getitem_45);
let hardtanh_30_0 = HardTanh::<f32, 960>(getitem_45_0, 0, 6);
let hardtanh_30 = Unflat4d::<f32, 1, 960, 1, 1>(hardtanh_30_0);
let convolution_46 = Conv2d::<f32, 1, 960, 1, 1, 960, 3, 3, 2, 2, 1, 1, 1, 1, 1, 1, 0, 0, 960>(hardtanh_30, arg138_1);
let getitem_46_0 = Flat4d::<f32, 1, 960, 1, 1>(getitem_46);
let hardtanh_31_0 = HardTanh::<f32, 960>(getitem_46_0, 0, 6);
let hardtanh_31 = Unflat4d::<f32, 1, 960, 1, 1>(hardtanh_31_0);
let convolution_47 = Conv2d::<f32, 1, 960, 1, 1, 160, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_31, arg141_1);
let add_8_0 = Flat4d::<f32, 1, 160, 1, 1>(add_8);
let getitem_47_0 = Flat4d::<f32, 1, 160, 1, 1>(getitem_47);
let add_9_0 = Add::<f32, 160>(add_8_0, getitem_47_0);
let add_9 = Unflat4d::<f32, 1, 160, 1, 1>(add_9_0);
let convolution_48 = Conv2d::<f32, 1, 160, 1, 1, 960, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(add_9, arg144_1);
let getitem_48_0 = Flat4d::<f32, 1, 960, 1, 1>(getitem_48);
let hardtanh_32_0 = HardTanh::<f32, 960>(getitem_48_0, 0, 6);
let hardtanh_32 = Unflat4d::<f32, 1, 960, 1, 1>(hardtanh_32_0);
let convolution_49 = Conv2d::<f32, 1, 960, 1, 1, 960, 3, 3, 2, 2, 1, 1, 1, 1, 1, 1, 0, 0, 960>(hardtanh_32, arg147_1);
let getitem_49_0 = Flat4d::<f32, 1, 960, 1, 1>(getitem_49);
let hardtanh_33_0 = HardTanh::<f32, 960>(getitem_49_0, 0, 6);
let hardtanh_33 = Unflat4d::<f32, 1, 960, 1, 1>(hardtanh_33_0);
let convolution_50 = Conv2d::<f32, 1, 960, 1, 1, 320, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(hardtanh_33, arg150_1);
let convolution_51 = Conv2d::<f32, 1, 320, 1, 1, 1280, 1, 1, 2, 2, 1, 1, 0, 0, 1, 1, 0, 0, 1>(getitem_50, arg153_1);
let getitem_51_0 = Flat4d::<f32, 1, 1280, 1, 1>(getitem_51);
let hardtanh_34_0 = HardTanh::<f32, 1280>(getitem_51_0, 0, 6);
let hardtanh_34 = Unflat4d::<f32, 1, 1280, 1, 1>(hardtanh_34_0);
let hardtanh_34_1_0 = Permute4d01::<f32, 1, 1280, 1, 1>(hardtanh_34);
let hardtanh_34_1_1 = Permute4d12::<f32, 1280, 1, 1, 1>(hardtanh_34_1_0);
let hardtanh_34_1_1_0 = Flat4d::<f32, 1280, 1, 1, 1>(hardtanh_34_1_1);
let hardtanh_34_1_1_1 = Unflat2d::<f32, 1280, 1>(hardtanh_34_1_1_0);
let mean_0 = MeanKD::<f32, 1280, 1>(hardtanh_34_1_1_1);
let mean_1 = Flat2d::<f32, 1, 1280, 1, 1>(mean_0);
let mean = Unflat4d::<f32, 1, 1280, 1, 1>(mean_1);
let clone_0 = Flat2d::<f32, 1, 1280>(clone);
let clone_1 = Unflat2d::<f32, 1, 1280>(clone_0);
let clone_1_0 = Flat2d::<f32, 1, 10>(clone_1);
let arg157_1_0 = Flat1dI0::<f32, 1, 10>(arg157_1);
let arg157_1_0_0 = Flat2d::<f32, 1, 10>(arg157_1_0);
let addmm_0 = Add::<f32, 10>(arg157_1_0_0, clone_1_0);
let addmm = Unflat2d::<f32, 1, 10>(addmm_0);
return (addmm);
}