vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix ssm_scan numerical error & others update

Browse source
This commit is contained in:
pidack 2024-08-27 16:51:21 +08:00
parent 8dd323b496
commit b423a6df5e
2 changed files with 36 additions and 38 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml/src/ggml-cuda/ssm_conv.cu
View file

@ -2,7 +2,7 @@
template <int block_size> template <int block_size>
static __global__ void ssm_conv_f32( static __global__ void ssm_conv_f32(
const float * src0, const float * src1, const float * __restrict__ src0, const float * __restrict__ src1,
const int src0_nb0, const int src0_nb1, const int src0_nb2, const int src0_nb0, const int src0_nb1, const int src0_nb2,
const int src1_nb1, const int src1_nb1,
float * dst, float * dst,
@ -32,7 +32,6 @@ static __global__ void ssm_conv_f32(
float * x = (float *) ((char *) dst + ir0*dst_nb0 + i2*dst_nb1 + i3*dst_nb2); // {d_inner, n_t, n_s} float * x = (float *) ((char *) dst + ir0*dst_nb0 + i2*dst_nb1 + i3*dst_nb2); // {d_inner, n_t, n_s}
// TODO: transpose the output for smaller strides for big batches? // TODO: transpose the output for smaller strides for big batches?
// d_inner // d_inner
#pragma unroll
for (int i1 = 0; i1 < ir; ++i1) { for (int i1 = 0; i1 < ir; ++i1) {
// rowwise dot product // rowwise dot product
// NOTE: not using ggml_vec_dot_f32, because its sum is in double precision // NOTE: not using ggml_vec_dot_f32, because its sum is in double precision
@ -56,7 +55,7 @@ static void ssm_conv_f32_cuda(
const dim3 block_dims(WARP_SIZE, n_s, 1); const dim3 block_dims(WARP_SIZE, n_s, 1);
const int nblocks = n_t; const int nblocks = n_t;
printf("size is %d\n",nr);
ssm_conv_f32<WARP_SIZE><<<nblocks, block_dims, 0, stream>>>( ssm_conv_f32<WARP_SIZE><<<nblocks, block_dims, 0, stream>>>(
src0, src1, src0, src1,
src0_nb0, src0_nb1, src0_nb2, src0_nb0, src0_nb1, src0_nb2,
@ -97,4 +96,3 @@ void ggml_cuda_op_ssm_conv(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
nc, ncs, nr, n_t, n_s, nc, ncs, nr, n_t, n_s,
stream); stream);
} }

16
ggml/src/ggml-cuda/ssm_scan.cu
View file

@ -2,8 +2,8 @@
template <int block_size> template <int block_size>
static __global__ void ssm_scan_f32( static __global__ void ssm_scan_f32(
const float * src0, const float * src1, const float * src2, const float * src3, const float * __restrict__ src0, const float * __restrict__ src1, const float * __restrict__ src2, const float * __restrict__ src3,
const float * src4, const float * src5, const float * __restrict__ src4, const float * __restrict__ src5,
const int src0_nb1, const int src0_nb2, const int src0_nb1, const int src0_nb2,
const int src1_nb0, const int src1_nb1, const int src1_nb2, const int src1_nb3, const int src1_nb0, const int src1_nb1, const int src1_nb2, const int src1_nb3,
const int src2_nb0, const int src2_nb1, const int src2_nb2, const int src2_nb0, const int src2_nb1, const int src2_nb2,
@ -11,10 +11,10 @@ static __global__ void ssm_scan_f32(
const int src4_nb1, const int src4_nb2, const int src4_nb1, const int src4_nb2,
const int src5_nb1, const int src5_nb2, const int src5_nb1, const int src5_nb2,
float * dst, float * dst,
const int nc, const int nr) { const int nc, const int nr, const int n_t, const int n_s) {
// const int row = blockIdx.x*blockDim.y + threadIdx.y;
const int tid = threadIdx.x; const int tid = threadIdx.x;
const int i2 = blockIdx.x;
const int i3 = threadIdx.y; const int i3 = threadIdx.y;
const int ith = tid; const int ith = tid;
@ -27,7 +27,7 @@ static __global__ void ssm_scan_f32(
const int ir0 = dr*ith; const int ir0 = dr*ith;
const int ir1 = min(ir0 + dr, nr); const int ir1 = min(ir0 + dr, nr);
const int ir = ir1 - ir0; const int ir = ir1 - ir0;
for (int i2 = 0; i2 < n_t; ++i2) {
const float * s0 = (const float *) ((const char *) src0 + ir0*src0_nb1 + i3*src0_nb2); // {d_state, d_inner, n_s} const float * s0 = (const float *) ((const char *) src0 + ir0*src0_nb1 + i3*src0_nb2); // {d_state, d_inner, n_s}
const float * x = (const float *) ((const char *) src1 + ir0*src1_nb0 + i2*src1_nb1 + i3*src1_nb2); // {d_inner, n_t, n_s} const float * x = (const float *) ((const char *) src1 + ir0*src1_nb0 + i2*src1_nb1 + i3*src1_nb2); // {d_inner, n_t, n_s}
const float * dt = (const float *) ((const char *) src2 + ir0*src2_nb0 + i2*src2_nb1 + i3*src2_nb2); // {d_inner, n_t, n_s} const float * dt = (const float *) ((const char *) src2 + ir0*src2_nb0 + i2*src2_nb1 + i3*src2_nb2); // {d_inner, n_t, n_s}
@ -41,7 +41,6 @@ static __global__ void ssm_scan_f32(
if (i2 > 0) { s0 = s; } if (i2 > 0) { s0 = s; }
// d_inner // d_inner
#pragma unroll
for (int i1 = 0; i1 < ir; ++i1) { for (int i1 = 0; i1 < ir; ++i1) {
// ref: https://github.com/state-spaces/mamba/blob/34076d664838588a3c97727b263478ab9f621a07/mamba_ssm/ops/triton/selective_state_update.py#L78 // ref: https://github.com/state-spaces/mamba/blob/34076d664838588a3c97727b263478ab9f621a07/mamba_ssm/ops/triton/selective_state_update.py#L78
float dt_soft_plus = dt[i1] <= 20.0f ? log1pf(expf(dt[i1])) : dt[i1]; float dt_soft_plus = dt[i1] <= 20.0f ? log1pf(expf(dt[i1])) : dt[i1];
@ -60,6 +59,7 @@ static __global__ void ssm_scan_f32(
y[i1] = sumf; y[i1] = sumf;
} }
} }
}
static void ssm_scan_f32_cuda( static void ssm_scan_f32_cuda(
const float * src0, const float * src1, const float * src2, const float * src3, const float * src0, const float * src1, const float * src2, const float * src3,
@ -75,7 +75,7 @@ static void ssm_scan_f32_cuda(
cudaStream_t stream) { cudaStream_t stream) {
const dim3 block_dims(WARP_SIZE, n_s, 1); const dim3 block_dims(WARP_SIZE, n_s, 1);
const int nblocks = n_t; const int nblocks = 1; // TODO
ssm_scan_f32<WARP_SIZE><<<nblocks, block_dims, 0, stream>>>( ssm_scan_f32<WARP_SIZE><<<nblocks, block_dims, 0, stream>>>(
src0, src1, src2, src3, src0, src1, src2, src3,
@ -87,7 +87,7 @@ static void ssm_scan_f32_cuda(
src4_nb1, src4_nb2, src4_nb1, src4_nb2,
src5_nb1, src5_nb2, src5_nb1, src5_nb2,
dst, dst,
nc, nr); nc, nr, n_t, n_s);
} }
void ggml_cuda_op_ssm_scan(ggml_backend_cuda_context & ctx, ggml_tensor * dst) { void ggml_cuda_op_ssm_scan(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {