gahow/xserv
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

phase 5: naive multi-head attention

Browse Source 
- Batched GEMM via cublasGemmStridedBatchedEx
- Causal mask CUDA kernel (F32 + BF16)
- Element-wise scale CUDA kernel (F32 + BF16)
- attention() composing: batched_matmul + scale + causal_mask + softmax
- Fixed to_device/contiguous infinite recursion (GPU contiguous via CPU round-trip)
- 5 attention tests passing (max_err < 3e-7 F32)
- Total: 61 tests passing across all crates

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Gahow Wang
2026-05-21 21:17:23 +08:00
parent c8e8153702
commit 6035ffdc0b
10 changed files with 550 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
csrc/activation/activations.cu
View File

@@ -35,6 +35,16 @@ __global__ void silu_bf16(const __nv_bfloat16* x, __nv_bfloat16* out, int n) {
if (idx < n) out[idx] = __float2bfloat16(silu_f(__bfloat162float(x[idx])));
}
__global__ void scale_f32_kernel(const float* x, float* out, float scale, int n) {
int idx = blockIdx.x * blockDim.x + threadIdx.x;
if (idx < n) out[idx] = x[idx] * scale;
}
__global__ void scale_bf16_kernel(const __nv_bfloat16* x, __nv_bfloat16* out, float scale, int n) {
int idx = blockIdx.x * blockDim.x + threadIdx.x;
if (idx < n) out[idx] = __float2bfloat16(__bfloat162float(x[idx]) * scale);
}
extern "C" {
void launch_gelu_f32(const void* x, void* out, int n, void* stream) {
@@ -63,4 +73,18 @@ void launch_silu_bf16(const void* x, void* out, int n, void* stream) {
(const __nv_bfloat16*)x, (__nv_bfloat16*)out, n);
}
void launch_scale_f32(const void* x, void* out, float scale, int n, void* stream) {
int block = 256;
int grid = (n + block - 1) / block;
scale_f32_kernel<<<grid, block, 0, (cudaStream_t)stream>>>(
(const float*)x, (float*)out, scale, n);
}
void launch_scale_bf16(const void* x, void* out, float scale, int n, void* stream) {
int block = 256;
int grid = (n + block - 1) / block;
scale_bf16_kernel<<<grid, block, 0, (cudaStream_t)stream>>>(
(const __nv_bfloat16*)x, (__nv_bfloat16*)out, scale, n);
}
}