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

ggml : update unary asserts and "supports_op"

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-06-10 16:17:51 +03:00
parent ebf95c2225
commit 8412561c4b
No known key found for this signature in database
GPG key ID: BF970631944C16B7
7 changed files with 70 additions and 57 deletions
Download patch file Download diff file Expand all files Collapse all files

20
ggml-cuda/unary.cu
View file

@ -148,6 +148,8 @@ void ggml_cuda_op_gelu(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -160,6 +162,8 @@ void ggml_cuda_op_silu(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -172,6 +176,8 @@ void ggml_cuda_op_gelu_quick(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -184,6 +190,8 @@ void ggml_cuda_op_tanh(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -196,6 +204,8 @@ void ggml_cuda_op_relu(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -208,6 +218,8 @@ void ggml_cuda_op_sigmoid(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -220,6 +232,8 @@ void ggml_cuda_op_hardsigmoid(ggml_backend_cuda_context & ctx, ggml_tensor * dst
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -232,6 +246,8 @@ void ggml_cuda_op_hardswish(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -244,6 +260,8 @@ void ggml_cuda_op_leaky_relu(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);
@ -259,6 +277,8 @@ void ggml_cuda_op_sqr(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
float * dst_d = (float *)dst->data;
cudaStream_t stream = ctx.stream();
GGML_ASSERT(ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F32);
GGML_ASSERT( dst->type == GGML_TYPE_F32);