Transformer architecture implemented in TypeScript with WebGPU.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
nn/shader/embed.wgsl

30 lines
815 B

struct MatrixBatch {
size: vec4<u32>,
data: array<f32>,
}
@group(0) @binding(0)
var<storage, read> ilogits: MatrixBatch;
@group(0) @binding(1)
var<storage, read> inp: MatrixBatch;
@group(0) @binding(2)
var<storage, read> pos: MatrixBatch;
@group(0) @binding(3)
var<storage, read_write> r: MatrixBatch;
@compute @workgroup_size(1)
fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
r.size = pos.size;
var s = 0f;
var li = (global_id.z * ilogits.size.y + global_id.y) * ilogits.size.x;
var ri = global_id.z * inp.size.x * inp.size.y + global_id.x;
for (var i = 0u; i < ilogits.size.x; i++) {
s += ilogits.data[li] * inp.data[ri];
li++;
ri += inp.size.x;
}
var resi = (global_id.z * r.size.y + global_id.y) * r.size.x + global_id.x;
r.data[resi] = s + pos.data[resi];
}